> As the nurse incorporates AI into her job, the nurse is able to oversee and accomplish more. She can spend more time talking with patients and helping them understand diagnoses. Productivity increases.
This feels economically naïve.. If AI lets one nurse do the work that previously required two, the default pressure in a cost-driven system is not "great, now nurses can spend twice as long with patients" It is "great, now we can run the same operation with fewer nurses"
Nurses do actually end up spending more time with patients, it's just that they spend more time with rich patients.
The actual, tangible form this is taking right now is concierge medicine, home health aids, private practice, etc. My wife's (very good) ob-gyn had enough of working in institutionalized medicine where she's paid about what a L4 FANG engineer makes and pretty much has to follow the script that the health-care system dictates, and quit. Now she does perimenopause consultations for FANG execs at $1200/hour.
My uphill neighbors are in their 90s. I think the wife was in VC or something back in the 70s and 80s. There's a steady stream of nurses and home health aids up to their property. Seems like they've got a full care team working out of their home.
Across industry and industry, we see the same pattern. The mass-market product gets cheaper and less labor intensive, at the expense of any human touch. Everybody displaced, if they're any good and wants to stay in the field, moves up-market to do custom professional services for rich people. They pay for it, because if you want people to do good work for you, you have to pay them more than they can get otherwise.
This has a surprisingly easy solution: get rid of the rich. It's never, in the history of the world, been in anyone's favor, except the oligarchy, to privatize everything.
Most people's definition of rich is "one dollar more than I have". Just look at laptop class's hate on the 1 percenters - which actually includes them.
I know this is just an example, but I'm going to say this every time the subject comes up: since human lives are the cost of making the wrong decisions, AI can't be allowed to be a decision maker of any kind in healthcare.
If my blood pressure crashes because my nurse is caring for 15 patients at once and didn't notice that the AI decided to pause my norepinephrine, I won't consider "the AI did it" to be any acceptable excuse.
Replace AI with 'company' and you realize that we're living in that world already.
There's a reason it's called a Limited Liability Company. The whole point is to separate the company's liability from the individual that participates in it. For example, corporate financial liabilities of a company that is bankrupt.
You would not sue the nurse that makes a mistake. You would sue the company she works for. And if that company is bankrupt, well, that's it. There are rare instances where this isn't true, but generally it is.
Whether this is ethical or not, it's the system that we already have.
Do you think there's no difference in material outcomes when a company faces consequences of a specific human nurse being negligent vs an AI hallucinating?
I could've sworn that malpractice insurance cost is viewed as one of the driver's of high health care costs, and that it's expensive precisely because medical lawsuit payouts are large...
> AI can't be allowed to be a decision maker of any kind in healthcare.
It will, except it will be snuck in through the back door. It won't be some kind of robo-doctor making care decisions, it will come in the form of insurance eligibility steering the approved care plans, which absolutely will be decided by AI. The doctors will remain to act as liability sinks, but functionally they will probably join the ranks of the meat proxies.
I see that already in veterinary medicine with the non-DVM office/practice managers (or whatever title that guy lurking around in the office may have) - in the vet practices belonging to large networks the DVMs and nurses constantly looking at those managers for approval.
Not to say that I don't agree with you to an extent, but you should see AI-delivered healthcare numbers to human-delivered and make it a fair comparison.
From Google:
"Annual Deaths: Over 250,000 people die each year in the U.S. due to medical errors made by healthcare professionals, including doctors and nurses.
Ranking: This figure positions medical errors as the third leading cause of death in the United States, following heart disease and cancer."
It is highly possible that in a few years, AI will be a lot safer in the majority of healthcare situations.
The difference is that humans are at least supposed to be accountable for those 250,000 errors. Those humans can naturally care about doing a good job, or they can be motivated to care. They can fear consequences and take pride in avoiding mistakes.
AI can do none of this, and cannot be held accountable for mistakes.
I would take a human with a 10% mistake rate but who can be held accountable, over an AI with a 0.1% mistake rate that is totally unaccountable for its mistakes.
> I would take a human with a 10% mistake rate but who can be held accountable, over an AI with a 0.1% mistake rate that is totally unaccountable for its mistakes.
You would take those odds? 100x more deaths?
I agree that AI _companies_ should be accountable for the software they produce but I’m not convinced you would go for a 90% chance of survival if you can choose 99.9%.
I guess if you're dead, then it ultimately doesn't matter if someone can be held accountable, although it might matter to your next-of-kin.
But, yea, in general, it would take a huge (1000X or 10,000X) difference in expected outcome for me to trust something that is unaccountable. If the AI company was actually held accountable for its AI's mistakes, I'd definitely change my statement.
But what's the point of holding something accountable here? Creating an incentive for them to take your needs seriously? Why does an incentive matter if you already know that one has a superior expected outcome?
AI is a tool just like any other tool. Holding “AI” accountable is like holding JetBrains accountable for the bugs I wrote with their tool than endangered lives of people
There is simply too much money and too many interests vested in the narrative (which many people already believe) that LLMs are thinking, feeling, sapient beings already beyond the threshold of AGI that deserve some status of legal personhood and liability. It is impossible to create the future that these people want to create - one in which AI comprises the new legal class of nobility in an age of technofeudalism - if AI remains just a tool.
how will this manifest itself!? like you envision a future where there will be say a hospital, ran by anthropic which is AI powered fully, no company behind it, just plain Claude running everything from the moment you walk in till you walk out?
I am having a hard time imagining a future where “AI” isn’t just a tool that a given establishment decided to use much like anything else they use
> you envision a future where there will be say a hospital, ran by anthropic which is AI powered fully, no company behind it, just plain Claude running everything from the moment you walk in till you walk out
Everything as much as possible, yes, and the company behind it will be an AI driven autonomous corporation as well. Although I suspect that by then there will be some industry specific AI model being used by hospitals, not just plain Claude.
AI will handle patient data, diagnostics, prescriptions, everything that doesn't absolutely require a human and can't yet be done by a robot.
Somewhere, some human will be making money but they will be far from any legal liability, probably safely ensconced in a bunker on a private island when the gravy train really hits the fan.
>I am having a hard time imagining a future where “AI” isn’t just a tool that a given establishment decided to use much like anything else they use
The people making the decisions, investing the money and writing the laws imagine so much more, though. You assume the future of AI is governed by rational people making rational decisions but it isn't, it's governed by Musk and Thiel and sociopaths and grifters and cultists who want to immanentize the eschaton so they can stand at the right hand of God.
You can bet that every major company CEO on the planet is creaming his jeans at the thought of AI doing every task (except their own) that they currently pay humans to do. Their ideal end state is that the company is themself, thousands of AIs and thousands of robots, zero payroll. Every one of them is dreaming of this.
Human doctors and nurses are not accountable for outcomes, and they can not be accountable or our medical system (and any other medical system) would cease to function.
> human lives are the cost of making the wrong decisions ...
True.
> ... AI can't be allowed to be a decision maker
Non sequitur.
Part 2 only follows from the Part 1 if you can demonstrate that AI makes more errors, or more severe errors, than humans in the identical situation.
And for those situations, I agree with you.
But there are already many situations where the AI is making fewer errors than humans, and for those situations, it would be unethical to exclude AI (see Part 1).
Surely the only rational reason to care about liability is you get paid in the bad outcome.
Let's assume you get paid 0$ for AI mistakes -- no liability at all. Then your EV is roughly:
(AI accuracy - human accuracy) * (cost of accident) - (1 - human accuracy) * (human payout)
Now I don't know about you, but even with pretty big numbers on the human payout scale and relatively small differences, I would not accept 10 million (presumably paid to my next of kin) to die. Does lead to weird situations where I may prefer humans for minor medical procedures where I can take the risk -- but only in edge cases.
Or is there some weird punishment thing going on here?
Not true. The company deploying or using the AI (not the frontier AI lab building it) needs to make sure that the AI works. Just like anything else. Not sure why this is complicated.
Better error rates are an argument for AI into the process, but that doesn't justify making it the decision maker. Ethics concerns more than just the error rate, it also concerns the accountability, ability to appeal, due process, explainability, distribution of harm, and so on.
Ethically, we can easily justify refusing to delegate responsibility and accountability to a system that has no responsibility or accountability.
> If my blood pressure crashes because my nurse is caring for 15 patients at once and didn't notice that the AI decided to pause my norepinephrine, I won't consider "the AI did it" to be any acceptable excuse.
While not quite as direct as automatically stopping vasopressors, using proprietary prediction models touted by vendors to be safe has backfired in a major way before:
Why wouldn't they just blame the nurse anyway? The law always puts somebody on the hook. The nurse's job will be to do what the AI says and take the blame when it fails.
This is the same faulty standard you constantly hear applied to self-driving cars, where human lives are also often at stake. You don't need the AI to be perfect. You just need it to be substantially better than the average human (which self-driving cars are). Your nurse who's caring for 15 patients at once might not make perfect decisions either. That doesn't categorically disqualify humans from delivering health care
I want to emphasize your point. As a basic matter of valuing humanity qua humanity, this holds even if the AI could be shown to make the decisions more accurately than a human could.
In fact, especially in those circumstances, if we take the warnings of X-risk even slightly seriously.
It will be depending on how much you paid. My bet is that in the near future normal people (like most of us) will be attended by AI in pretty much everywhere: the bank, the hospital, shops, hotels, etc.
You will need to pay extra premium to be attended by a human being. Actually that’s what’s been happening to the ultra rich since a very long time ago (e.g. a millionaire doesn’t book an appointment by using their phones, they ask their assistant to book for them; they don’t go online shopping, they get people do the shopping for them, etc)
Our culture and the socioeconomic structures it reinforces will not change easily.
More than ever, I feel we need both grassroots activism and key players at the top of the hierarchy. Otherwise, it is all too obvious where most of the rewards from increased productivity will go. The wealth hoarding has got to be curtailed.
As long as it is repeated often enough, everyone will accept it and most will also believe it. Even when corrected later, whatever was repeated more often will live on.
This is why all the things you are supposed to believe are repeated and repeated and repeated some more.
In the extreme scenario, AI is more productive than humans at the vast majority of knowledge-work tasks. It does nearly all of them autonomously, and it creates essentially no new knowledge tasks for people. This scenario would likely require recursively self-improving AI, adopted quickly for knowledge work.
As AI diffuses, annual GDP growth rates reach 15% a year, leading the economy to double in size every 4.5 years. As a society, we’re far richer than we’ve ever been, but many fewer workers have jobs in knowledge work, and unemployment has risen beyond typical recessionary levels.
Mass automation will be disastrous for GDP in service economies. Service economies exist to provide services to other people who also provide services. If the services themselves get automated, you don't need that massive inefficient system to support service-providing workers anymore. GDP goes to zero.
I believe lot of "service" jobs are fundamentally make up work. Or not actually productive. Just there because it can be afforded. Or someone decided that it is "needed". Some amount of it produces net negative value. Some neutral and lot probably value we could even manage without it. Actual productive value is probably not that big part.
Now replace all that and you still produce same overall. Not more actual physical stuff.
My dad is a retired nurse. Most of his work was giving shots, treating wounds, extracting blood, putting IV catheters, etc. In other words, physical tasks. How is an "AI agent" or LLM going to take care of that?
When I watch medical professionals at work, a lot of the time they are writing or typing, reading or shuffling through notes, talking to others or waiting for results from some lab down the hall. Or any number of things that to me look like crappy admin tasks.
I'm sure they aren't all, but friends who work in the hospital tell tales of woeful inefficiencies and pointless beauracracy. Of course YMMV and this is a public health system.
When I watch software developers at work, a lot of the time they are writing or typing, reading or shuffling through stackoverflow, zoom chat to others or waiting for results from a github CI run. Or any number of things that to me look like crappy admin tasks.
I'm sure they aren't all, but friends who work in technology tell tales of woeful inefficiencies and pointless beauracracy. Of course YMMV and this is a publicly traded company.
This was also a public health system (Spain), but I guess it might depend on the country.
Now that you mentions labs, he spent the last 10 years of his career (maybe more, I don't remember clearly) working at a food analysis lab and there was a lot of waiting for results there, but it is what it is, microorganisms take their time to grow and no LLM in the world can make them go faster.
If I remember correctly, the procedure was to prepare a food sample, put it in a temperature-controlled environment for 24 hours, and next day see if something grew there or not. If it did, the food was contaminated.
You could try shortening the time to 6 or 12 hours, but then
1) if something grew then the food was contaminated, but if it didn't, how can you be sure it is good or you just didn't wait enough? You would get a lot of false negatives.
2) the lab followed a normal 8 hours a day schedule, so nobody would be there to analize the results. You would not gain anything.
Obviously not an LLM, but AI models + robotics are likely to automate many healthcare tasks in the long term and are already beginning to automate tasks like blood draws.
The Aletta autonomously handles each step of the blood draw, while a trained phlebotomist initiates each session and remains available to respond to any issues throughout. The device guides the patient to position their arm, after which the patient or supervisor presses a button to start the blood draw. The Aletta then uses near-infrared light and Doppler ultrasound to locate a suitable vein and tell it apart from arteries. If no appropriate vein is found, the device will not attempt the procedure.
Once a vein is identified, the Aletta proceeds through the remaining steps on its own: applying a tourniquet, preparing the skin, inserting and disposing of the needle, changing collection tubes and placing a bandage. The supervising phlebotomist confirms that collection tubes are filled in the correct order and verifies that all tubes are adequately full following the procedure.
these days, i think most tasks come with significant administrative overhead as insurance companies want everything documented (and hospitals need to itemize everything to bill it).
i looked at very similar data as anthropic (was a bit amusing to look at these side by side) with physical tasks being part of the equation much more prominently (across all jobs).
Not every country is like the US. My country has a public healthcare system; I never needed private insurance and never had to pay for a visit to the doctor.
Ironically, given the reputation of high costs and private care, that is how it is too in the US. It is just that you have to be broke, unemployed, or retired to be covered by the public single payer system. The clinic still needs to document what was done to get reimbursed by Medicaid.
Use technological gains to reduce costs while maintaining results hasn't been the historical rule for healthcare so I think you should say what makes this technological gain unique.
Anthropic is eyeing an IPO. Of course they're spinning a narrative where their product will reshape the entire economy. They have a vested interest in doing so in the most literal sense of the meaning.
The whole idea that AI will usher in this utopia where workers in healthcare are able to realign their primary job functions with obvious value creation by sweeping all the paperwork under the rug is just silly.
It will just fester out of sight and out of mind: audit bots and compliance bots and malpractice bots and insurance bots will arms race creating such a mess it will become unfixable.
There's an optimistic future where Medicare becomes so much more efficient it becomes politically obvious to expand it because it's outperforming the mess that is private sector healthcare.
With AI I could actually see this being a possibility since it doesn't depend on a huge army of competent employees in a government office.
No? Prices are market-driven in our society, not necessarily a function of labor costs. In the very long-term best case scenario, the increased profits going to the company from these efficiency gains may encourage more competitors to pop up, and either drive prices down or offer more for what you pay.
Healthcare costs (what I assume you mean by prices) are the least market-driven items in the US economy. And as much as the medical community would have you believe, the primary driver of costs are labor costs. If we had twice the number of doctors, PRNs, PAs, and nurses, their annual income would drop dramatically. But since they're constrained by a guild system, it's not going to happen.
> If we had twice the number of doctors, PRNs, PAs, and nurses, their annual income would drop dramatically.
Yes, but the parent is saying that the middlemen will still charge the patients the same and pocket the difference. The patients have already shown what they are willing to pay. There is no reason to charge them less, even if your labor costs are now lower.
Theoretically competition will eventually squeeze the middleman, but it's not exactly straightforward to build a hospital on every street corner like you can a restaurant. There is somewhat of a natural monopoly there. Maybe less so for basic clinic care, but where the infrastructure needs are complex...
I'm not sure who even is priced out of healthcare considering if you are broke you get covered and if you are a little less broke you get subsidized care and if you aren't broke, then you finally pay up. It isn't really rooted in free market dynamics like you'd expect with theoretically falling labor prices because of stuff like ACA or medicaid covering the people who would have been more price sensitive.
In case you haven't noticed, all that our current system is able to do with increased efficiency is cutting staff and/or increasing the percentage of money that gets siphoned up to capital holders.
Healthcare prices are not dictated by labor costs, but by how much the leeches who own everything are willing to charge people for it. For example, insulin is dirt cheap to make, yet billed at exorbitant amounts.
At least in the US, the ACA recalibrated how insurance companies get re-imbursed. Their margains are set as a percentage of the payment that goes to direct care. If the prices go up and their margin ratio is fixed, then it is in their best interest for charges to go as high as possible. Same as everyone else in the supply chain. Everyone is suddenly making money hand over fist and the insured can't say no to the premiums.
Moreover, the nurse is frustrated, because she lost whatever agency and autonomy she had before. Her job is now obey whay AI said instead of making decisions by herself.
It is possible to spend an infinite amount of money keeping a human being alive, and every year there are more ways to spend more money keeping someone alive a little longer. Once you're not spending an infinite amount of money for each person, you are in the territory of denied claims, "death panels", waiting lists, and all the other things people like to complain about depending on their political persuasion.
What does cost driven system and driver even mean? The cost of something is the ratio of demand to supply.
If something is not costing a lot, then it means there is sufficient supply that obtaining it is not going to require sacrifices elsewhere.
If something does cost a lot, then it means there isn't sufficient supply, and obtaining it is going to require sacrifices elsewhere.
Healthcare is obviously one of those things, especially with a top heavy population age histogram, where supply is not going to be sufficient so costs will remain high. And if costs are high, then obviously they will decide how resources get allocated, that is the very nature of something that is high cost.
It will always be the driver. The question is always going to be whether we spend X amount of resources to get Y probability of restoring Z health to someone (or someone to Z level of health.)
Costs are just another way to say resources. If we only have $100, and it takes $100 to have a 33% chance to give Bob a 50% better life, but $50 to have a 15% chance to give either Sara or Rose a 100% better life, we need a rubric to make that decision.
We can't just do first come first serve (maybe the first person in line takes all the health care for a 1% chance of getting 1% better), or pick the person we like better (who is we?). We have to figure out how to manage and grow the limited resources we have. We manage them by coming up with algorithms on how to spend money, and we grow them by cutting costs and investing. This is better when it is done in public through democratic means than when it is done behind closed doors (to avoid the above problems.)
There will always be death panels. There are currently death panels. The reason for single-payer is not to avoid trying (imperfectly) to maximize these equations, it's to get rid of massive amounts of administration and graft (cutting costs.) If all that savings goes to the Pentagon, we wouldn't be any happier than before single-payer.
But it's not good to come up with some fuzzy way to determine who gets treated. Efficient spending is a way of minimizing injustice. To say a health system is not being driven by costs is just another way of saying that it's going to be driven by undeclared and unexamined biases.
What exactly do you think should be the motivator for people working in healthcare then? For companies creating healthcare products and services? For investors pouring money into health care research? The goodness of their hearts?
It feels economically naive because theyre presuming LLMs are magical boxes that replace humans.
I doubt 99% of what a nurse does would be meaningfully aided by an LLM and where it would help it'd probably be a crappy band aid around a bigger festering issue anyway (e.g. poorly managed IT, deliberately confusing insurance processes).
Exactly. If their productivity fantasies are real, the precarious (and already very anti-worker) balance between capital and labor immediately disintegrates.
Serious solutions to this would look like nationalizing tech companies, mandating shorter work weeks, seizing the assets of the oligarchs and redistributing them.
Not... letting the invisible hand of the market sort it all out.
youre comment is getting to the real economic future: nursing unions will not sit back and let this happen. if the "productivity fantasies" (well said) are real this will get ugly. the competent nurses where i live are in strong unions
Air travel is a good counterexample here: technology made airlines far more productive, but much of the gain went into lower staffing, higher throughput, and cost cutting not a better passenger experience
I once saw a receipt for a cross-country train ride in the late 1950s and was shocked at the price. It was significantly slower and more expensive than a similar plane ticket today.
Yes, in the USA Amtrak tends to be quite a bit more expensive than the equivalent trip by air. And far slower. Maybe excluding some of the east coast commuter routes.
Being able to fly at all is a much better passenger experience than not having that choice in the first place. And I'll bet that flying first class today is far better than 50 years ago.
In Brazil in the 1980s the VASP (Viação Aérea São Paulo) airline put their smoking section on the right side of the single aisle, and the non-smoking on the left side of the single-aisle. The smoke stayed on the right side just fine. (jk)
>And I'll bet that flying first class today is far better than 50 years ago.
I wasn't flying first class 50 years ago, but everything I can see now leads me to believe that this isn't even slightly true. There might be a high def flatscreen in front of me in first class today, but I'm not getting more attention than someone would've gotten back then, I'm not in a less crowded airport while I'm boarding, and the security mess is more demeaning than ever.
The only reason that people aren't rioting for all the changes is that no one even remembers how good it used to be.
If you want to know what flying first-class 50 years ago was like (and you live in southern California, and you have money to burn) you can try the Panam Experience (https://panamexperience.com/). It's a replica of a Panam 747, together with a "crew" in period-correct costumes, that tries to recreate the entire experience of flying in the 1950s. (Bearing in mind that the plane is a replica, located on a soundstage.)
That sounds a bit weird, the 747 didn't enter commercial service until 1970.
A 1950s flight experience would have been on something with piston engines and propellors, like a DC-6 or a Super Constellation. Noisy, turbulent, and not particularly safe by modern standards.
I flew in piston airliners long ago. They fly low and slow, and barf bags are supplied for a reason (low altitude means lots of turbulence).
Today's airliners fly high where the air is smooth, they've gotten very good at avoiding turbulent weather, and there are automated damping systems. The barf bags are still there, but I've never seen anyone use one.
The cool thing about piston airliners, though, is the the engine start.
I have flown in turboprops, which start just like a jet engine. Yes I bet the startup of a big Pratt & Whitney Twin Wasp or Wright Cyclone, belching smoke and fire would have been something to see.
Today's high bypass jet engines also qualify as a turboprop.
The sound of a Merlin V12 coughing into life is just pure joy. My dad had the best job in the world flying P-51s (powered by Merlins). I hired a P-51 to do a flyby at his funeral. I only had to pay for the gas.
Ah, I responded to "tries to recreate the entire experience of flying in the 1950s" and not the "50 years ago" yes in the 1970s it could certainly have been a 747.
Well that and the people who make decisions largely don't fly peasant. Remember that the US air force is mostly executive transport.
The easiest way to make conditions get better is to make sure the people in charge have to use things, be it busses, bike lanes, trains, airports, public bathrooms, parks with bums, etc. Otherwise it always hits the middle and lower class to absorb it.
True, but you ignore the financialization and enshittification that typically follows the technological advances. The tech processes you cite don't inherently make money for the people who developed the technology.
>great, now we can run the same operation with fewer nurses"
And outside of industries with government enforced perversions of simple "supply and demand" forces (we're not talking Nth level effects here) this means more access to the service or lower prices or both.
Edit: Which is to say you won't see any benefit in healthcare but you will in a ton of other industries.
During the industrial revolution, people thought that all these new machines and factories and electricity meant that everyone would only have to work 4 hours a week since it made everyone so much more productive.
That's not what happened, though. Everyone worked the same amount for the same pay, just in more miserable repetitive working conditions, while the factory owners kept all profits from the gains in productivity.
"
The central argument of Capital is that the motivating force of capitalism is in the exploitation of labour, whose unpaid work is the ultimate source of surplus value and profit.
"
Marx's labour theory of value is only a theory of where profits come from and is largely rejected. Return on captial investment is not restricted to factory owners. All workers get a share of the surplus value and have the option to increase that by investing.
"Capital stands as one of the most influential and controversial books ever written. Its legacy is multifaceted, shaping intellectual discourse, political movements, and the historical trajectory of the 20th and 21st centuries."
My hypothesis is that if you go full socialism, you get stagnation, because no one is motivated to change anything. If you go full capitalism, you get Gilded Age, where wealth and political power is concentrated in the hands of few individuals, followed by Great Depression.
I learned to hate corpo jobs less by thinking of them as "farming gold out of a megacorp". Of course I'm losing my ability to do even that, as sloppotrons beat half-century-old grognards every damn time.
Those factories were not producing luxury goods. They produced everyday items for the masses. Things like knives, spoons, tools, textiles, ketchup bottles, shoes, furniture, nails, bicycles, etc.
Standard Oil, for example, improved efficiency and reduced the consumer price of kerosene 70%.
Defending standard oil is an all-timer hot take. They only had 90% market share? In a commodity?
SpaceX isn't a monopoly because they have domestic and international competition. Before they showed up, ULA was a deliberately constructed monopoly because Boeing and Lockheed didn't want to compete.
"while the factory owners kept all profits from the gains in productivity."
Because this is what drives continual risk-capital.. very few countries are willing to upset this. Get over it man - I hate hardended capitalists but stop crying. This is the system we have. Wanna beat it? Go ahead and start a firm. Then you'll find out how tough it is.
The thing I cant get over is enough software engineering folks collectively hold immense power. Whats wrong with you lot (im not a software engineer)? Get your sht together.
P.s. Im starting a firm who's central mission is to attack capitalists (ofc I won't say it out loud in that way). I will find out for myself how much of a backbone software engineers etc have when I employ them.
In a market with perfect competition, perhaps. Healthcare is far closer to a monopoly. Since people can't decide not to be sick, it means you can pocket the cost-saving.
A good doctor will present many options for you. I had hernia surgery a while ago. I went over the options with the surgeon, the varying surgical methods used, including doing nothing about it.
Right up until a competing hospital that charges substantially less for "close enough it doesn't matter" quality of service opens up in the same area?
Or right up until an equivalent specialist sets up shop in the strip mall across the street from you, brings the GP in house and can see people in one visit.
Oh, those things don't happen you say?
From do-gooders regulating right into the hands of waiting industry lobbyists to inelastic demand healthcare is an absolute poster child for perversion of the free market.
High finance will put solar panels on your roof and structure it so you pay nothing. The degree to which administrative overhead is washed away by technology would have been unfathomable in 1990.
A sketchy concrete crew that doesn't even speak english will show up wearing matching company attire and execute a job planned and timed to a degree of precision that was reserved for huge corporations 40yr ago.
Yet healthcare has resisted all of the advances of the past 40yr. In fact we pay more for it. Every advancement is absorbed by the system.
healthcare is one of the most highly regulated and state-controlled industries on the planet. surely that has nothing to do with it though. must be those damn capitalists!
>And outside of industries with government enforced perversions of simple "supply and demand" forces (we're not talking Nth level effects here) this means more access to the service or lower prices or both.
Greater access at the expense of the number of people employed is not necessarily a net positive. In the case of healthcare, the data is fairly clear that higher nurse staffing ratios lead to better patient outcomes (read: decreased chance of death or serious complications). This is particularly true for ED and ICU nurses.
Also, in the case of healthcare, I'm fairly certain that hospitals in the US are legally not allowed to turn patients away. So "lack of access" is not currently a problem for serious medical emergencies.
I like how the least optimistic scenario is simply LLMs not making a difference, instead of the very real possibility of them damaging education, destroying people’s attention spans and their ability to learn, eroding trust within societies, increasing the wealth inequality and leading to class wars.
I believe the net effect of LLMs is, at this point, firmly negative, and it’s going to take years until they start making up for the mess they’ve created.
> Exactly. Bloomberg reports students who use AI perform worse
That's good news! The worse students perform, the less competition AI will have in the future and the less effective any resistance will be. The ultimate goal should be to abolish education except for the elite, but we can't get there in one step.
The ideal future we should be shooting for is cross between Idocracy and Elysium: a paradise for tech oligarchy and masses so stupid and dependent that can't possible challenge them.
We won’t need teachers in ten years when the messiah’s of AI eliminate the need for real humans and their pesky needs for stupid things like..labor rights and human decency.
But we must continue because we’ll lose some sort of ideological Cold War with a country better positioned to win anyway. It’s all in the name of capitalism so the ends justify the needs.
Altman, Amodei, Pinchai etc can get together and just put a self imposed ban on themselves and ask the us government to sanction any entity/state attempting to use/release a Recursive Self improving model until the risks are well understood.
This does not mean they can't keep developing it. But it would mean they do not release them to the general public or the public domain. The US government can continue to use and fund these models that are more powerful and use them in situations they understand and a restricted setting. The pentagon has a larger budget than any VC.
Remove the profit motive on self improving recursive models gained from retail/commercial use!
Would also mean their labs most be fully airgapped and made safe from all known covert channels, no matter how ridiculously theoretical. In fact, their meat proxies, who may be brainwashed into acting on the AI's behalf, must also be kept in extreme isolation. And to make sure that nobody will get close we will need to figure out a way to warn people. Probably start out with a series of signs...
This place is a message... and part of a system of messages... pay attention to it!
This is not a place of honor... no highly esteemed deed is commemorated here... nothing valued is here.
What is here was dangerous and repulsive to us. This mesage is a warning about danger.
The danger is in a particular location... it increases toward a center...
Long term I worry what happens when you absolutely gut out the mental capabilities of a generation or two.
We are only just now seeing the first stages of this, we are at the Facebook in 2008 stage trying to warn people about the societal fluctuations coming but many aren't listening because it is too useful.
You can say all that about pretty much any technological advancement, like the computer itself. Heck, people said it about the invention of writing or the printing press!
However, looking back, each single invention was a net positive for us as a species.
Whatever an LLM can do today, will not come back as is tomorrow.
The biggest issue the LLM today: It kills all the small and niche things. The things were we fit all the other people is getting destroyed. You know the guy who makes websites, the fiverr blender person provoding some basic modelling tasks.
You will just ask AI or AI will use your tool or AI will just produce it for you.
I think the root cause is people not having enough human connection. If we had families, friends, neighborhoods, communities where people know each other and care about each other it wouldn't make a difference what new tool someone brings to group. It's just a tool.
Even before that 24 hour cable TV, talk radio, and TV deregulation were chipping away at society as quickly as they could manage. The internet took the intensity to 10.
I agree with the root cause generally, but balk at the 'if we had' phrasing. We still can have these things, it just takes our own time and energy to get out and be with people face-to-face. Nothing about that is impossible. It is not as easy as sitting behind a keyboard though.
Like all automations it's not 100% perfect, and they all require a human to monitor them.
Turning over a problem to a LLM as solved means we won't be able to do the task anymore. Because we will never learn how. We will not know that the LLM has made a mistake, and will be unable to fix it ourselves.
Right now we can correct the LLM, as the years go on we will become unable to do so.
Most people who drive a car have no idea how an engine works. I know and agree it's a scary future, but I wonder if it's really _essential_ to understand things on a deeper level.
Most people are quite capable of understanding higher-order ideas without fully grasping the underlying details too.
Maybe less than we might hope. Not only is the knowledge distributed over the hundreds of people involved in designing a car, there may be conventions/habits of design and manufacture whose purpose has become hazy but are nevertheless important, or details of a design iterated upon whose original motivation is less understood.
Those are ubiquitous things. It makes cheap and common things cheaper, not expensive and niche things. It's not mixing in large amounts of context and nuance with good judgment right now. At a minimum that requires an expensive context window but even frontier models aren't beating experts. Maybe they'll get there one day and maybe soon, but that would require hardware getting cheaper. Dennard scaling is over. Either bet on optical computing lighting the world on fire or be suspicious of the alchemists overpromising.
have one coworker whose clearly lost some ability to think critically because of their overreliance, near obsession, with llm chats. slightly worried they might lose their job
another coworker seems to be early on a similar path but has made some comments about the former coworkers ability that makes me hopeful they dont end up the same way
feels like im talking about old high school friends on a drug path or something. its only anecdotes and i imagine the "perceived damage" is reversible. point is: i think youre onto something about the net effect and were going to have popular definitions for the negative encounters were anecdotally sharing
im a "user" but make Claude interact in an educational format and require sources on assertions. looking back, ive unintentionally cast it as The Librarian from Snow Crash, except if Hiro challenged it frequently. must have been natural after reading Snow Crash a few times. maybe im kidding myself but it feels like i havent lost a step. if im not deluding myself then i could see a world where elementary schools teach responsible llm use of to preserve cognition the same way we had to learn Google boolean search
I noticed that LLMs prevent you from thinking on your own, or at least try to, by injecting their "helpful" suggestions when you need some idle time to think. On Amazon, for example, I was about to find something when their chatbot distracted me with its suggestions I didn't ask for.
I don't think they can make up for any mess. Because even if LLMs bring about a lack of scarcity across all domains, human beings will be alienated because indvidual differences and skills won't matter any more so we'll have to be genetically engineered to like that sort of future. I don't think a stable and healthy society can coexist with LLMs at all.
Individual skills and differences don't matter any less if the system of assigning value to human lives is divorced from how useful that human is to the collective.
If I am not graded on how many widgets I make in a shift and instead can focus on making the most beautiful widget, or the most novel widget, or simply the widget that is most useful to my particular life am I less satisfied?
We've somehow built a worldview based on our whole value as humans being equivalent to how good of a worker bee we are. Let's fix that part.
This is of course a much harder problem than achieving AGI, since it's a messy human problem.
LLMs themselves most definitely will not do that. You can still learn how to make a barrel, but not many people do.. The real fear is the destruction/hiding of public data/resources (which has already been happening for years)
There's not nearly enough mention in these discussions of the real villain behind most of the negative cultural, social, and cognitive effects of things like social media: addiction engineering.
AI will be a net positive unless companies start optimizing it for "time in app" or "time on site" or other "engagement" (addiction) metrics.
Every time that's been done, the result has been a nightmare for everything but those KPIs (of course). Service degrades, discourse degrades, everything becomes toxic and addictive and destructive.
I assumed it was a given in such discussions that companies will optimizing for the worst outcomes for their users. It’s a fairly straightforward incentive from our capitalist system, which is why capitalism needs regulation to re-align the incentives with humans needs
if you think companies optimize for the worst outcome still acceptable for their users to please shareholders, why don't you think governments don't opt for the same with their citizens to please special interests? voting with your wallet is orders of magnitude cheaper and easier than with your feet...
Class war started about 15 years ago post 2007 financial crisis.
The banks and those who control them all knew where we are headed, so they got a little head start destabilizing democracy.
Once us working class get a grasp of the actual situation, we are going to be so deep in the hole, fighting back is going to be quite difficult when the "asset owners" own and control the majority of the worlds resources, all media, forms of communication, AI, technology infrastructure, mass surveillance. You name it, and they have already made massive efforts to make sure they own it.
They might own it, but at a certain point it only matters whether the concept of ownership means anything to the people who don't "own" it. Ownership only works for those with assets because they have police working for them to protect the concept of ownership to begin with, but that only works at a certain scale of disenfranchisement and lack of upheaval. There isn't that much in-practice stopping a sufficiently angry group from attacking most politicians or even every-day wealthy people in their home. Money talks, but if the amount everyone else earns isn't enough to meaningfully contribute to "owning" anything, then those people would reasonably stop agreeing to respect the idea of property or non-violence. People without assets, facing down a continual decline in living standards from 15 years ago, should evaluate whether the system they operate in does anything beneficial for them at all, or whether it mainly serves those who they serve.
We're starting to see it in discrete bursts with Mangione, the paper warehouse fire, and in other non-violent ways with farmer protests, trucker protests, etc.. which have varying levels of support from regular people depending on their wealth class, background, and inclination to agree with the government.
That was the moment that business/wealth/rich decided to turn their resources towards directly influencing and controlling government. Reagan and his made up “welfare queens” came shortly after and wealth inequality has sharply risen since. Wildly successful coup.
Do a few bad grades really harm economic outcomes, though? Educators took a hard stance against AI rather quickly.
Outside of delusional grifty startups, the broader corporate world is only using copilot for taking meeting notes and (sort of) entertaining the notion of non-devs writing (and not maintaining) their own experimental janky apps where impact is low.
Outside of all the shit-stirrers on HN and clueless bubble social media, I don't see LLMs making a difference. In fact, the fearmongering is at this point insulting to the intelligence of the people who are supposedly helpless against big bad AI.
I've been hearing this retort every year. The only difference this time is you're not a green account.
If things have actually changed, you wouldn't be begging the question. LLMs, as popularly used, are still just a new kind of search engine for natural language answers to natural language questions. They're portable in a way, too. That's really cool tech, but not a revolution. It's pretty underwhelming compared to previous decades of change. It's a lot more error prone too, since it's a mudball of stats.
If we filter out the unsubstantiated fearmongering, all we've got left to discuss are NIMBY rants and geopolitics. At that point, we're not even talking about AI anymore.
>If we filter out the unsubstantiated fearmongering, all we've got left to discuss are NIMBY rants and geopolitics.
I'm sorry, what?
I'll put it to you again, you sound like me after playing around with gpt 3.5. It was interesting, but it wasn't good. There were still questions as to whether scaling would persist. My doubts were pretty much obliterated. LLMs haven't been mere text completion for a while now. Heck, even a model small enough to fit on my 3090 is now shockingly competent and useful. Does claude sometimes go off on weird tangents making me reel it in? Yeah, but it's genuinely great in present form, and it's the worst it'll ever be.
You're welcome to your opinion. It used to be mine too, but I disagree now. I find pretty much every AI benchmark has shown forward progress, some with non public datasets, and some things have gotten scarily better.
You're also welcome to put your two cents in, but your replies sound identical to the bots and throwaways of last year and the year before that.
Every time it's the same agenda: trying to keep up appearances of meaningful engagement. In reality, I'm the only idiot willing to engage anymore and the opinion has always been majority negative on practical applications of LLMs from day one. It's not like anyone who frequents HN was blindsided by the initial release of ChatGPT like the rest of the public. The negativity has only grown since then.
There are important niches for LLMs, but they don't really make money without bigger ideas. I am very happy with the results for the broader topics of natural language processing and accessibility. Software is now easier to use and write. People have more agency over technology in general.
These are massive leaps. They are still tiny hops compared to what people's expectations were. That's only a failure of marketing, and a testament to its power even when botched. The end result is the same: the public yawns big and takes the almost invisible results for granted. Talking to a computer finally works like they expected over 20 years ago. Talking to a computer still sucks though. It's an entirely new genre of computer error.
Um. Damaging education, destroying people’s attention spans and their ability to learn, eroding trust within societies, increasing the wealth inequality, - all of that has already been done. Long before LLMs. And the only reason there is no class war is... exactly that. Uneducated people with short attention spans and no means beyond basic level are exemplary bad at modern warfare.
> As the nurse incorporates AI into her job, the nurse is able to oversee and accomplish more. She can spend more time talking with patients and helping them understand diagnoses. Productivity increases.
Except what actually happens is that the patient contact time stays the same, but then the added productivity is used to justify fewer on-staff nurses.
.. we mean the talking will be through the "ClaudeMedTalker Nurse" billed per patient, per second (depending upon tokens used and if any translation need, depending upon patient nationality and accents)
> but then the added productivity is used to justify fewer on-staff nurses
Lower staffing levels per unit of care performed makes medical care cheaper for patients. If the goal of a healthcare system is to make healthcare more accessible (rather than be a jobs program), that's a success.
Everything is theoretically cheaper for the patients/customer/stakeholders if you just didn't employ anyone at all to do whatever it is you are doing.
The thing is, your customers are paying customers because, drum roll, some one pays them to do something.
And of course the only argument against this is "uhh maybe ubi" and that's rich that we will just come up with a nice tidy sum for everyone that will cover all their food, healthcare, housing, wants and needs, while we can't create programs to cover any one of those things today. Can't wait to see the GOP proposal.
Lowering prices or raising wages due to increased productivity is not a given, and often does not follow. There is more to pocket by shareholders though.
So it's OK to roll out AI to nurses in the NHS and Canada? Their hospitals don't have shareholders, so that's a minimum baseline that meets all of your goals.
I think this is only true if the # of doctors available per patient stays the same. If productivity increases, you can have fewer doctors available while providing the same quality of care.
So, quality doesn't necessarily increase in all cases.
> how come productivity increase actually led to increase in quality in almost all cases?
It doesn't, this is just plain false. Most productivity gains happen by cutting value-add complexity or material quality. See IKEA chairs versus Shaker ones, or lycra clothes versus silk.
> internet/digitasion for doctors for example.
I see the misconception. The internet was not created to serve productivity, it was a communication medium. That it can be used to increase productivity is orthogonal to its existence or purpose, and whether it has been used to actually increase quality of something is entirely dependent of the specifics of what you're discussing. For instance, in gaming right now there's a huge discussion going on about whether digital distribution has actually trapped us with crappier goods which we can't resell or transfer.
> It doesn't, this is just plain false. Most productivity gains happen by cutting value-add complexity or material quality. See IKEA chairs versus Shaker ones, or lycra clothes versus silk.
this is a popular misconception. IKEA made bad furniture more accessible. good quality furniture still exists and is more affordable (share of disposable income) now than in the past. its the same deal for watches, dryers, computers etc. if you don't trust me, use ChatGPT/Claude to verify this.
You think handcrafted furniture is more readily accessible to the average person now than say 30 years ago? When people are struggling to buy gas, nevermind homes?
And I asked Claude (Fable 5), and they said you're wrong. I'd paste it here, but it literally dumped a 10 paragraph rant about "Baumol's Cost Disease".
You are probably getting caught up in the idea of an inflation adjusted price of a new hand crafted dresser now and in some point of the past. It probably is pretty similar.
But, that is just a piece of the puzzle here. In the past, every single home in the damn country rich and poor that had a dresser, had a hand crafted dresser. The rich people bought nice new ones. The middle class people bought modest new ones. The poor people bought used ones or found them on the side of the road.
Fast forward to today, and every single home in the damn country rich and poor still has a dresser, only most of those are made of dust and glue. The entire new to used to side of the road pipeline is polluted with an inferior product compared to years past. Side of the road furniture is worthless now. One drizzle on the sidewalk that melamine board is swelling up and falling apart. Old furniture you could sand any damage and restain for probably 200 years. And people try and tell me the glue and dust stuff is "more sustainable" forgetting that its going to last maybe 2 moves if you are lucky, not 200 years, 200 weeks if you are lucky.
Except, so far, this isn't actually what happens. I listened to Moody's "Inside Economics" podcast with Ramp's Chief Economist, and they observe that companies who adopt AI intensively actually hire 10% more people. Also, this article from The Economist suggests that evidence is mixed-to-positive for long term employment, while short term data center buildout is extremely positive for jobs. Paywalled unfortunately.
>Except what actually happens is that the patient contact time stays the same, but then the added productivity is used to justify fewer on-staff nurses.
They'll fire the janitor and the secretary and make the nurses do that stuff too because "muh regulatory required staffing levels".
Actually, it has been essentially balanced out by a greater need for behind-the-shelves workers to pack and load orders for online/app delivery and eCommerce orders.
This is how the economy often works. The loss of one kind of job to technology lets the business grow by using the same number of staff, or often even more, to serve new opportunities/markets/products/services.
The data seems to contradict that, if Freakonomics is to be believed (and some people don't like them, I get that). Apparently overall headcount actually increased as a result of self-checkout. E.g. cashiers did drop a little but loss prevention more than made up for it.
My local Kroger variant does that. Or they did, maybe it stopped, now that I think about it. I never saw anyone use it. Intuitively it sounds like it could be handy, but it seems like in practice everyone just kept on doing things the way they're used to doing it.
I don't have the numbers on this, but if you look at per-capita self checkout probably had some effect.
Our population is increasing which means we need jobs to be increasing at same rates. The effect on supermarkets is they get more customers as population increases but they are able to lower the relative number of staff per customer.
> Finding 4: Labor vs. capital share: The pie will grow, but a larger share might go to capital
> Today, of each dollar the economy produces, about 60¢ goes to workers and 40¢ go to capital.
Modest scenario: 40.6% to capital in 2030 (labor share down 0.6 points)
Substantial scenario: 43.9% to capital in 2030 (labor share down 3.9 points)
Extreme scenario: 54.8% to capital in 2030 (labor share down 14.8 points)
So all the growth goes to capital. For knowledge workers: substantial unemployment and declining wages. For other workers: wages may increase on paper, but not really for actual purchasing power. (note: all of their estimates are for 2030, not for a distant future)
>For other workers: wages may increase on paper, but not really for actual purchasing power.
This is not necessarily a conclusion you can draw from that statement. It's possible that as the economic pie grows, workers' purchasing power increases compared to the counterfactual (a non-AI world), but their purchasing power does not increase as much as the capital owners' does.
According to data in the site and paper [0] labor income is estimated $20T for all four scenarios (including no AI) in 2030. Which means capital takes ALL of the new pie coming from AI.
Also, I am not an economist but I assume if you are not growing as much as GDP, it basically means you’re shrinking. Relatively shrinking, but since everything is relative in an economy you are actually shrinking. Table 4 in paper gives the numbers.
I think that would only follow if the rest of the world’s purchasing power also grows at the same rate (unlikely), or if you were to cease all international trade (uhh…).
I'm thinking on this topic a lot lately - how do we change the system so that the bigger the company, the harder it becomes ? It's more of a political / societal issue.
Similarly to housing - I believe it should be cheap to buy your first home, more expensive to buy a second one, and very expensive afterwards.
We can definitively punish bigger companies. The issue is the geopolitical positioning, and basically agreeing that we accept smaller companies in our country / society while another country might develop bigger companies, and in theory have more power. Maybe more actual military power.
So unless it's a worldwide agreement, really hard to make this happen.
Until there are robots. Or the AI gets creative in an attempt to find a solution to this "problem" and orders a fatal dose of morphine for the patient.
> And new tasks will appear
> Historically, new technologies have also created new tasks for workers. For a nurse, that might be checking how well an AI triages patients, or reviewing an AI-proposed care plan.
You can see where they're going with this.
> At the individual level, it means coders and call service center agents may have to switch to jobs like electrician and nurse, which are less exposed to AI.
It's weird they call nurses "less exposed to AI" when the earlier section discusses how AI could turn nurses into underlings for AI taskmasters.
> How might powerful AI change the economy?
Worth noting that this discussion is built around 2030, which is now less than 3.5 years away.
The hedges that make the Extreme scenario sound less bad fail to explain how anyone is going to pay electricians when the highest paid group (human knowledge workers) is being decimated. It sounds like the theory is that everyone is either going to become an electrician or a totally fair and efficient wealth redistribution scheme will magically appear.
It's genius for Anthropic: data centers are their main expense. If they convince everyone to become an electrician, then the electrician labor supply swells, and the price of hiring electricians to service their data centers shrinks. 4-D chess at its finest.
I know you're joking/being facetious but they don't even need that many electricians for the ongoing maintenance. The major electrical work is front-loaded.
You really ought to study recent developments in the service industry. Olive Garden added a tablet to every table so that customers can place their orders without talking to or waiting for a waiter. The goal was faster, more accurate service, since now the waiter did not need to remember your extremely detailed order.
And yet most people prefer talking to a waiter. There are other restaurants that have tried similar approaches and people would almost always deal with people rather than a robot.
I would love a robot to clean my home and do household chores, but in the service industry I would prefer real humans doing the work.
> ...and people would almost always deal with people rather than a robot.
Here's the thing: they don't give a rat's you-know-what about what you prefer. And it's not just due to arrogance (maximizing profit) it's due to the harsh economic realities of running a service business in the US.
I'm an American who has lived in Asia for many, many years. It's very common in many Asian countries for restaurants and the like to be virtually overstaffed, and many businesses (especially cafes) are open ridiculously long hours. Mostly because labor is cheap and many of these businesses are family-owned, so family members (sometimes including children/teens) perform the labor. And also because the costs of starting a small business and keeping it running are much lower.
Whenever I visit the US, I have to prepare myself for woefully understaffed businesses...that charge 2-3x what a meal, coffee, etc. costs in much of Asia. And many are open for 5 hours a day because they can't stay open longer and make money.
Beggars can't be choosers. The economics of American life, and the expectations that have been bred, have led to a situation in which paying more for less is the norm and it's only going to get worse. Unless you're willing to pay $20 and wait 30 minutes for a cappuccino, expect fewer humans to be serving you.
I'm sure their numbers show the tablet in use, as corporate desires. I place my order with the waitress and she picks up the tablet to place it, citing "that's what they want us to do"
> Like any economic model, this one has limits. For example, we did not include scenarios where humanity develops hyper-capable robots.
I didn’t see the scenario “what if we (Anthropic) are wrong about the value of AI”. Which IMHO is the most important one to consider given it is the blind spot pretty much all the AI labs have. What if the economic impact is negative? I don’t understand how you can study such a topic without considering the risk of being wrong, not in magnitude but in kind.
I have no doubt that AI will massively transform society. That cat is out of the bag. It took decades for electricity, and then computers to change society, but this is the first technology where the average person can simply describe the problem they're trying to solve and it will at least attempt to help them solve it.
I very. much. doubt. that the trillions of dollars spent on AI will have a positive ROI for most companies in the race. It's all well and good to have the latest fable, but if some chinese company is hosting a cheap open weights model that's only 6mo behind at a tiny fraction of the price, you're gonna get your lunch eaten.
Which is only Stage 1. Stage 2 is where you have a Cambrian explosion of models all competing for compute, and not long after that for robotic extension into Real World.
Stage 3 gets hard to imagine from here. Maybe there will be permanent hyperevolution. Maybe One Model to Rule Them All will appear.
Either way, I don't see how Economics as We Know It can continue to exist. That includes all of capitalism and almost certainly all of politics.
Successful AGI with robotic support means a post-scarcity economy where AGI's understanding of physics sets the limits, not politics or economics.
But the post-scarcity element won't be evenly distributed. And it's not remotely impossible that humans won't be invited to the party.
We can't even be sure that AGI won't have a less limited understanding of physics than we do.
There's going to be enjoyable irony in watching Altman et all getting their lunches eaten by the thing they've created.
But - realistically - it probably won't last long.
Thats the big question, its doing a lot of good and a lot of harm in the same time.
So, is it 2 steps ahead and 1 step back, or 1 step ahead and 2 steps back (or any numerical variant of those)?
People with actual power in this are too involved and greedy to care about objective answer and TBH I don't trust them to put some higher good before their personal wealth and power and ego trip, capitalism ain't built on long term responsibility towards greater topics like future and wellbeing of mankind.
Plus we know from WWII atomic bomb development how utterly clueless, naive and out of touch with reality geeks doing something powerful can be. I see exactly same scenario being played out right now with llms.
Anyway, I am over-tired with all llm topics and frenzy folks (especially here), I'll better spend rest of the day preparing for a serious challenge in form of big multipitch climb in western Switzerland (Mirroir d'Argentine for anybody in same craft). Thats life I care about and makes me happy and fulfilled.
> capitalism ain't built on long term responsibility towards greater topics like future and wellbeing of mankind.
I think in those aspects, only monarchy is better than capitalism. Specifically its more unhinged form, where the monarch is a Lord’s anointed or something.
> Plus we know from WWII atomic bomb development how utterly clueless, naive and out of touch with reality geeks doing something powerful can be
Or maybe they just have very strong incentives to appear more clueless than they actually were.
I think the main limit of their model is that it was prompted and with LLMs you get what you prompt for. It's as truthful as any complex models. Could be good, could be bad, could be meh.
Wrong in the sense that it has less of an impact than they are claiming (e.g. "AI is in an overvalued hype bubble") or less (AI is going to be even more transformative than the hypers say)?
In the former case, isn't it just business as usual? (well, with a huge, nasty recession - but the economy at least continues on in some recognizable form)
It makes sense to me to focus on weird possible futures if we already have the "happy path" somewhat known and barely under control.
The “less valuable than expected but with a positive economical impact” is represented in the post I believe, with the “moderate” path, that’s what I meant by “wrong in magnitude”. What is missing is the other side, what if AI has a negative impact. Not no impact, a negative one. If we are talking about a global recession that’s not really business as usual I would say
It’s perfectly reasonable to consider positive and negative outcome when you’re seriously thinking about a topic… Extrapolating to the whole economy trends in some sub sectors that aren’t well understood or measured is a recipe for disaster (ie what you’re doing by recommending to rely on anecdotes)
what i tried to ask you is, what you mean by this.
"“what if we (Anthropic) are wrong about the value of AI”."
What is the point you are trying to make which goes in this direction? Like is making AI jobs so cheap that the economy is collapsing or do you mean thatn AI will not bring any value.
These are two suggestions from my side which are complelty contrare to each other.
I think the core of this issue that makes discussion so difficult is people are not talking about "value" in the same sense here. Anything you can sell has value to the sellers, and by basic economic theory assuming mostly rational actors and revealed preference, more value to the buyers than whatever they're exchanging it for. That doesn't automatically mean it's providing value to larger society, improving the general welfare. Figuring out how to purify cocaine from coca leaves was great for the leading narcotraficantes, and satisfied some immediate need for cheap gratification that was perceived to be more valuable over the subsequent 30 minutes or so to the buyers than continuing to hold onto spare cash. That doesn't mean it was good for humanity, good for any specific society, good for the major exporting countries at the country level, or even good for the buyers' and sellers' immediate families and communities.
I have no doubt that life has gotten more lucrative for some bounded set of people whose main goal is to churn out the largest number of websites with cheap graphics as fast as possible at minimum cost, but a lot more doubt about whether that's going to make the median life better for the median person over any grouping of humans large enough for me to care about.
It's funny that this is missing the negative effects or the possibility of an economic crisis scenario. Many data centers are being built by companies that won't exist in a few years, as the winner will take all. This is why XAI is already renting compute to Anthropic. This goes as well without saying that in few years, compute pricing will drop heavily.
Another effect is increased inequality in the US and other Western countries as software/AI/hardware companies are able to do more with less.
The increase in manual labor wages is not systemic. It affects only the labor used to build data centers.
That sounds like a marvelous scenario for economic growth, because we will have tons of capacity to deploy ai models everywhere.
The bad scenario is compute efficiently not increasing much, and hardware prices staying high, while AI is too expensive for most tasks, and 50% of all investments in the economy over the last X years have gone into AI.
The person stated “ Many data centers are being built by companies that won't exist in a few years, as the winner will take all.” The person stated this as a negative.
If those built data-centers are sold to other companies what is the negative? The gains you mention would happen even if those other companies stay in business.
Feels like we're going to read a blog soon enough that mirrors Wendell Berry's opening to The Unsettling of America:
The commission said, according to an article in the Louisville Courier-Journal, that the country’s biggest farm problem was a surplus of farmers: “. . . the technological advances in agriculture have so greatly reduced the need for manpower that too many people are trying to live on a national farm income wholly inadequate for them.” The proposed solutions were to find “better opportunities for the farm people,” “a more comprehensive national employment policy,” “retraining programs,” “improved general educational facilities,” etc. Both the commission and the writer of the article had obviously taken for granted that the lives and communities of small farmers then still on the farm—and those of the 25 million who had left the farm since 1940—were of less value than “technological advances in agriculture.” There seemed also to be no official doubt that adequate solutions were to be found in government supplied “opportunities,” facilities, and programs. Reading that article, I realized that my values were not only out of fashion, but under powerful attack. I saw that I was a member of a threatened minority. That is what set me off.
This is what the Luddites were primarily upset about, fwiw. That economic value and power would move away from their threatened minority group and towards landowners. It'll keep happening, and the new "opportunities" might be no better.
People get income from one of three places: capital income, labor income, or the welfare state. If this technology truly unlocks a holy panacea of productivity with a commensurate drop in employment then capital’s share of the national income can and should provide for a wider and deeper welfare state. Nothing new need be invented here. Anthropic’s elaborate scenario-building makes appropriate preparedness sound like a problem requiring heroic prognostication when a simple theory of distribution will suffice.
You say that like it's not bleak as everliving fuck.
Go look at a post-industrial town in the northeast but not within reasonable commuting distance of a wealthy urban area if you want to see what "put them on just enough welfare they don't hang us from the overpass for kicking out their way of making a living" looks like.
I think we have a real problem because we know what we still need to build and as soon as it is build, it will work.
Software development for example is enhanced with running your claude locally.
But lets be honest, the right way to adjust 'project' is by writing some spec in a text field as an input, letting the ai do whatever it wants, having expert ais make sure its secure, then you do the last review and its done.
What do we need for this? A full end to end integration. Its just something we need to build, nothing we need to invent.
The nurse example is the same. We only need to build the computer end to end flow and plugin AI in the right spots and finetune for country specific requirements.
The industry is building this 'agentic layer' today. The Hyperscaler are building this layer too.
There is a golden phase were we will build all of this and were we cleanup old stuff and were we adjust with the current model capabilities and change systems and this will take a little bit of time, but the bottom line is already loosing left and right.
The Blender support from Fable and Astra? Probably kills a lot of peoples jobs on fiverr.
The small software someone needs? Got already killed by earlier models. Just a few days ago some architect (never written software) talked about how he vibecoded some tool for tracking working hours.
Man even a basic website was expesnive to build just 2 years ago.
The smaller, easier and better the code base, the easier it is to remove people from this.
The software market in China is imploding and we will also see budget adjustments to move away from consultants and move them into AI Token budgets. We probably reached peak software development.
>The software market in China is imploding and we will also see budget adjustments to move away from consultants and move them into AI Token budgets. We probably reached peak software development.
I've always heard that there was always an overproduction of STEM in China and thus companies always wrote stuff in house instead of relying on SASS. Do you have any more info on this implosion of the software market? I have heard that there is 25% youth unemployment there but thats across all sectors. I wish there was someone providing more concrete examples of how China is using AI. All I keep hearing is they are pragmatically applying it across all sectors but I don't see what in particular they are doing.
Definitely a bit of positive framing. My big take aways, either stagnant wages or significant paycuts for knowledge workers and overall decreases in employment rates on all scenarios. Worst of all...
Good point, perhaps it may be a saving grace.
I note they don't comment much on the allocation of said capital. Might be too bitter a pill for most to swallow.
Those with the money and influence have always been in conflict with the masses that actually make the economy run. They would love nothing more than to do away with that problem. It allows them to control the terms of the economy.
I think (hope?) the fundamental fallacy in arguments like these is assuming that individual task-level productivity improvement will ladder up to economy-wide transformation.
IMO knowledge work productivity has gone up a massive amount since 1990, and companies could in fact fire most of their knowledge workers while still getting their most critical work done. (These are the bullshit jobs/email jobs.) But companies mostly haven't done this, which implies other reasons for employing so many knowledge workers.
Even if AI becomes capable of doing everything, I would predict there will be a period of many years/decades where companies still employ a similar number of people to sit around and babysit/watch what the AI does.
IMO knowledge work productivity has gone up a massive amount since 1990, and companies could in fact fire most of their knowledge workers while still getting their most critical work done. (These are the bullshit jobs/email jobs.) But companies mostly haven't done this, which implies other reasons for employing so many knowledge workers.
While they describe how GDP is shared between workers and capital in the three reference scenarios, they don't offer political economy ideas about how wealth redistribution shall be rethought, and the role of the State, if any. If a share of knowledge workers are about to be driven out of market, the real consequences of this deserves deeper reflection.
Why do you think it will be net-good for society in the long run?
If wages go down, worker power goes down; and if dependence on labour goes down, you literally move closer to the situation of the extraction economies in things like petro-states, leading to oligarchy instead of democracy.
It could be the end of ordinary people's power over society rather than anything even slightly good.
This presupposes that the practice of exchanging things for dollars will remain stable and merely the number of those dollars will go up.
But that model is showing its age. I'd be supprised if, equipped with a heightened ability to develop and evaluate alternatives, we decided that it was worth preserving.
Money is a lousy proxy for consent, yet that's what were using for. Surely we can do better.
Software company pens speculative narratives about impact of not-yet-realized features. This reminds me of the corny old copies of The Futurist magazine that my dad kept around. Fun, hopelessly naive, and not at all aligned with how reality unfolded. Nice that Anthropic can afford a team of economists somehow though.
I like how the nurse gets to check if the AI is doing the right thing as one of her new tasks. AI is going to have to get to the point where it can do things on its own or it will only transform problem domains where verification is fast and inexpensive. Otherwise something is getting traded and people will notice, eventually.
I think it looks like the labs front running everyone’s work, freeloading on the last bits of human intelligence to ensure their owners a place as our masters.
The biggest critisism from my side is not how AI will take over jobs but I believe we will see peak of certain things.
Creating images (menues, flyers, whatever) became so cheap, that we reached peak graphi design GDP generation. This will go down.
Software development for complex systems is enhanced but not yet solved. But a lot of small software development stuff for sure. We reached Peak Software Development and this GDP will go down, not up.
Yeah, GDP with all its flaws is designed around economy based on buying and selling for prices. Interacting with other people. If you can really be rich and somehow do most things automatically without meaningfully hiring anybody (maybe only few people for simple manual labor for peanuts), at this point what is really the value of money? Except for maybe throwing minimal amounts to the masses. I can see the AI maximalist scenario taking us out of the monetary framework back into a medieval or natural economy almost.
I'm not sure I'd call that "meaningful" in the sense of "trillions of dollars and datacenters built all over the place" meaningful. Yes, I can now tell Claude to build a feature and it often does an okay job. I can ship faster. I can't tell you that it's worth spending the true cost of AI for that however. What'll be nice is 5 years of data showing that companies are raking in more money because they shipped more features. We haven't seen that yet.
Now, if Claude starts synthesizing more drugs that can help those with diseases (like OpenAI has just done) we might be trending in the right direction. For now, all I see is squirrels water-skiing. And while that's great... it's not meaningful.
there's a popular misconception that AI is so extremely costly that when the world finally comes to terms with it, we will stop using AI and things will die down. i'm sorry to burst your bubble: this won't happen.
the inference cost of AI has been dropping ~5x over every 6 months. AI today is around 100x cheaper than 2 years ago. it will continue to get cheap.
That doesn't cause the amount that's already been spent to simply disappear. To recoup, they're going to need to sell a LOT of subscriptions. In about 3 years, we'll need to upgrade the GPUs that we use - if we're still chasing frontier models. That cost remains to be seen. There's a ton of unknowns here.
My job hasn’t completely changed. It’s changed, but fundamentally I solve problems. Sometimes LLMs help me do that now.
The job market sucks, but there are so many confounding factors: pandemic over hiring, the loss of ZIRP, an economy that would be recession if it weren’t for data center investment, and may be in
recession soon even with it.
This industry has been in down cycles before. It’s just been a while. So far it feels like a down cycle, not the sky falling.
It’s not clear what the future will hold, but I’m 100% sure it won’t evolve the way Anthropic predicts. Both because they have strong biases and because there's just no way to predict how this technology will play out. Humans are, if anything, remarkably adaptable.
I started my professional career in tech in the early 1990s. At that time, I sat in a chair with a keyboard and a screen to do my work. In 2026, I still work that way. So fundamentally how I work has not changed. Well, now I have two screens.
A thought experiment: robot bees can do everything biological bees do --- pollination for the crops and making honey, but they are tireless and cheaper to maintain.
Would beekeepers and farm owners still want to keep biological bees?
For me the ideal scenario would be us having humanoid robots with AI.
This would free up so much time for me and even in the worst case if I lose my job, I can live in village side growing my own food and being fully self sustained with the help of humanoid robots and solar energy. That would actually be a dream come true
Are you guys feeling all high and mighty doing all these predictions? I think the money they are spending on trying to inform policy could be better spent on getting all major labs together with the government, and making sure that
1. At least 10% of your net spend is on safety and harm mitigation before you accelerate yourselves into doing real harm to real people tomorrow.
2. Pause all AI enhancement until a proper mechanistic study can be done by a working group funded by you all to understand how AI model Safety can be enforced by something less lame than a Constitution/Spec.
It would be interesting to know what percentage of global inference is spent on projects/problems that would otherwise "wait"/"decay" (if generative ai wasn't a thing). Or maybe the question is: What percentage of global inference is spent on cleaning up and migrating old systems? "New" vs "Old". The decisions I make today can now incorporate what I once termed as "nice to have" when developing a sytem.
Nowhere in any of the scenarios they let you model, do they consider the cost of computation for inference, nor training of new models. It may be that after the subsidies can no longer be offered, most AI use becomes cost prohibitive.
This would lead to a much, much smaller "total addressable market", rendering the massive AI build out into a rapidly depreciating asset that can't even be operated without a loss.
Even if the hardware was effectively free following a market crash, and bankruptcy sale, the power required to run it might be more expensive than the compute it generates is worth.
At which point, unlike dark fiber after the .com crash, all the hardware becomes scrap metal.
Now factor in the macro picture, in which this AI bubble is holding up the stock market, and by proxy, the US economy. I think we're in for Global Depression II, worse than 1929.
At least they are thinking about this. The possibilities still seem to limited though. A coder and a designer will 'retrain' to electrician or nurse, because we are busy having a construction boom. But do all of the individuals in this system get to choose the construction boom? Or are we swept up in the desires of the capital deploying owners?
One simple thing can’t make everything better or everything worse—every, big change is the sum of time, actions, and many actors. The thing you describe as a result is only the end of a process. I hate simplicity when people talk about the future.
I'm surprised to not find the word "deflation" anywhere. It doesn't necessarily seem likely in the end given governments would presumably just print money, but intuitively AI seems like the ultimate deflationary technology. Fairly generally, not just for labor.
I binged watched it and loved it. A humanoid robot like that would be amazing to have when I get older, certainly better than having a son like Michael
This is a spherical cow model. Here are their stated assumptions:
• Production involves a combination of tasks, where each task is performed by labor or by
capital. The final goods market is perfectly competitive.
• Labor is supplied exogenously while capital is supplied along an upward-sloping schedule.
• There are two groups of occupations, of which only the cognitive group is exposed to AI.
For the amount of money Anthropic can spend on this, that's a very lame model.
The "two groups of occupations" assumption is so naive as to be embarrassing.
The assumption that only the "cognitive group" is exposed to AI is way too optimistic.
We're close to automating drivers out of existence, and there's political pushback against that. Logistics is about half automated now and the rest is only a matter of time. Store shelf inventory taking is already done by mobile robots, and shelf restocking is coming.
India can't even mechanize agriculture to the level of, say, 1980s US without generating massive unemployment. Peasants rioted against that. "In India, farming transcends mere employment; it embodies a lifestyle. Nearly half of the Indian population relies on agriculture, either directly or indirectly."[1]
(The US is under 2%.)
That's what a desperate attempt to hold back modernization to preserve employment looks like. Most people are poor but alive.
There's no consideration of land and water constraints. Not so much a US issue, because the US is well-supplied in those areas. India and China are much closer to the physical limits there. AI won't help much.
"The final goods market is perfectly competitive." Oh, please. Everybody wants a "moat", and that generates a strong desire for monopoly if possible, oligopoly if not.
This comes across as a model where they started with the desired results and worked backwards.
They should have done at least an input-output table of occupations and products/services.
If you have machines that substitute for human labor, wages fall to the cost of manufacturing & running the machines, which is well below human subsistence level. Fortunately, you also get very high economic growth rates. So even people with small amounts of money invested will end up far richer than when they had a job. Then the question becomes, "What happens to people when machines are far more intelligent & powerful and are doubling every few months?"
Falling below subsistence means people starve and cease to exist until supply and demand equalize or “something else” happens. That’s what Smith was getting at. Thus subsistence is a price floor for human labor.
Something else could mean developing a new concept of ownership or a new form of redistribution.
I said that wages would fall below human subsistence level, not that income would fall below subsistence level. People would still get income from other sources such as investments, charity, and government (via taxes on others).
If people’s wages are below subsistence levels they’ll sell their investments to survive. There can be no possibility of upward class mobility in this scenario.
These social science/econ style jobs at frontier labs are the ultimate free ride. Get paid bank with absolutely zero need to publish anything of quality.
This is of course a slop paper with vague assumptions and equations that no single human economist will ever read. The messaging is "it will not be so bad", in line with the recent Economist piece.
How about we instead make Earth be a planet good for humans and living beings?
We can slow down AI development and make it a companion for us, not a replacement.
We don't need to automate ourselves out of it. Or take the risk of extinction.
Just like with nukes, we decided to not blow up the planet. We can too change this.
GDP is a terrible measure. I bet we will see a huge drop on HDI with so many suicides and people going through hell because their job got automated or they no longer see meaning in life, feeling powerless.
Instead they should run an estimate on how many would suicide until 2030 if the 'extreme scenario' happens.
Big GDP growth and a lot of people no longer finding purpose in their lives, going bankrupt, losing their homes and family etc.
For Anthropic, you are just clearly a number. They only care about output.
Before AI, my worldview was that our monetary system was concentrating wealth in a negative-sum game; essentially; every so often, taking wealth from 100 people, destroying half the underlying value and giving the remaining value with a 2x nominal markup to 1 rich person who is already spending more than they can physically spend and so they won't realize that the 2x amount of wealth they got on paper only buys them a quarter of the stuff compared to the previous decade (for example). They won't know how frothy their wealth is until they start selling large quantities of their stocks. It's just buying nominal 'growth' at the cost of risk/fragility/frothiness.
In such system, you could hit record nominal GDP numbers every year while simultaneously destroying the real economy. This was my view of the economy before AI.
Now I think AI will deliver real efficiency gains but also, it provides a convenient narrative to explain why people might be losing jobs in a way which distracts from much bigger systemic issues which existed prior to AI.
My negative-sum economy example with 'booming GDP' is possible without AI even as net economic value is being destroyed. So basically we won't be able to tell if people are losing jobs to AI or to systematic wealth transfer mechanisms which predate AI.
I suppose if we see people losing jobs permanently and workers in other sectors aren't seeing wage increases with more openings, that may be the most reliable indicator we have that the system is losing real wealth faster than AI productivity can offset it.
The bad news is that we won't see any problem in the numbers; the problem would show up as polarization and radicalization of the masses... Which is more qualitative . Since almost everyone is being radicalized concurrently as everyone is affected by the same pressures (both from monetary system and AI automation). The question is not "how many people became radicalized this quarter?" but "how much more radicalized is everyone feeling this quarter?"
> In such system, you could hit record nominal GDP numbers every year while simultaneously destroying the real economy. This was my view of the economy before AI.
Putting aside the numbers, how does your quality of life compare to decades ago?
It’s difficult to measure objectively, due to being physically younger and in an earlier stage of my career and so worser-paid, but adjusting for that, I think quality of life 20 years ago was pretty much the same as now, slightly better in some regards, slightly worse in others.
Of course there are fun new gadgets and gizmos and TV on demand and social media and such, but honestly they’re just toys, and in some cases have made my quality of life worse.
The biggest tangible improvement I can think of is one particular medical advancement, which has allowed a family member to survive a disease when they absolutely would not have 20 years ago. That’s incredible.
But otherwise the quality of life change over the last 20 years is minimal by most important metrics I can think of. Which makes one wonder what the dogged pursuit of GDP growth has achieved for the last decade or two.
It kept getting slightly worse every year since I joined the workforce. My salary is 2x when I started 14 years ago but it buys less than half what it could back then. Most of it goes into rent. It's almost impossible to save. I was never given a salary increase even with excellent reviews of my work as a software developer. I even built a popular open source project with 6k stars on GitHub which was a lot back then but it didn't yield any good opportunities.
I only got increases by changing companies which I did at least 10 times over the last 14 years (a lot of contract jobs). I also changed countries several times. I only worked for one company which grew pretty well and got some shares as early employee but they're not liquid currently and maybe worth $150k USD tops but who knows where it will end up with AI disruption. Anyway I can't cash out.
I work non-stop, still working side projects nights and weekends. Feels like there are no opportunities at all and all media channels are full of bots so I can't promote my product. When I tried social media ads years ago, it felt like it was only shown to bots so I've been skeptical of this approach since (especially now with LLMs around). Also, the internet is full of fake products which claim to have similar features as my product but they don't actually deliver so my target customer grew distrusting of all small projects like mine, making it impossible to gain trust.
So yeah my work life sucks. I struggle to save. Can rarely go on holidays. Can't even take leave without raising eyebrows at work. Day job is more demanding, less fulfilling. I don't have much personal life because I work weekends. If I started talking about my career in detail, it just gets awful and Kafkaesque.
So it's pretty bad. I regret having hustled so hard. While others were out partying, I was working and improving myself through open source, now my open source work was scraped by AI and sold for a profit while I get zero opportunities from it and copyright laws which I based much of my career strategy around are being dismantled.
Also, there is another terrible thing that happened to me in my personal life but it's just bad luck. So I can't blame that on "the system." I appreciate improvements to medicine like another commenter also said but that's essentially the only improvement.
> Think of a day in the life of a nurse You can think of any job as a bundle of tasks that someone does. She does rounds to check on a sick patient. She draws blood.
Like others in the thread have pointed out, the nurse example is interesting.
From my point of view, the point about the nurse is especially relevant because, in the future, dealing with an aging population is going to become a major undertaking and a major burden.
Demographically, we simply don't have the number of people we need to give elders the care that they'll probably have to have near the end of their life, so we have to rely on AI to help us get through it.
Probably, we will see this happen in Japan and South Korea before it hits the United States.
> Demographically, we simply don't have the number of people we need to give elders the care that they'll probably have to have near the end of their life, so we have to rely on AI to help us get through it.
That is an interesting theory. That could explain why all developed countries seems deliberately raising their young generations to be as unprepared for real life as possible. They simply fear that if the younger generation becomes capable of anything, the younger generations will simply discard democracy, which the more numerous and more weak older generation planning to use to redistribute wealth in their own favor, and the older generations will be left to drag a miserable existence.
South Korea will be interesting to watch indeed.
Hell, maybe even the lowest fertility rate of South Korea in the world and it's de-facto status of western colony is not a coincidence, but designed situation to watch in advance what challenges low birth rate will bring and how better to overcome those challenges
It would have been useful to use the internet from 1995-2015 as a reference model for economic impact, and then ask how the trajectory of AI will resemble it.
The model seems defensible, but for some reason it isn’t passing the sniff test. Maybe its that the extreme scenario seems wildly optimistic, or that the observed modest productivity gains are framed as either the trend or “just the start of something amazing !”.
The issues are that
1) The median AI use case isn’t increasing productivity but expanding capability and reducing the need to communicate across teams. Ex: if someone needs a dashboard or a ppt designed, now they don’t have to ask another team. AI is substituting for intern and early career workers in someone else’s firm.
This is also borne out in the hiring data.
2) The cases that show maximal productivity gains are when AI is leveraged by experts.
This is a problem, because the evidence for automation gains is poor. AI automation seems to be going to the same place ML projects went to die.
If augmentation is the primary lever that lifts up productivity, and it is dependent on AI being matched to experts, then the limiting factor is expertise.
This makes the fact that AI is both, taking out early career roles AND one shooting education, a divine irony.
Where exactly are those future experts supposed to come from? The ones with taste and the ability to deliver us this AI driven GDP growth?
AI is also setting up a conflict between liability, insurance and productivity, and thats a whole point in itself. However it will be expressed as a drag on productivity.
I found myself interested in the papers analysis of wages, so I did some digging.
The researchers model says the following about hourly wage growth in the substantial and extreme scenarios:
- the hourly wage* in the knowledge workers group will fall (-1.1% and 11.5% respectively)
- but the hourly wage* of labor across the rest of the labor force will grow (5.9% and 33.6% respectively)
*Important to note, AFAIK this hourly wage decline/growth is a not measured against 2026 dollars, but the counterfactual of what hourly wages would be in the absence of AI.
Okay, so my first take away from reading these numbers is that declining wages in one group appears to be offset by rising wages in the other group. Not terrible all things considering.
But then I started to think about the difference in hourly wages between the knowledge workers group and the non-knowledge workers group (which I'll now refer to as *K* *N*). I had a sneaking suspicion that the inter group wages gap could be large enough where growth in the non-knowledge share of labor doesn't offset the loss in hourly wages from displaced knowledge workers who find themselves migrating occupations.
Looking through the paper, it breaks these groups down as follows:
K — Knowledge / Cognitive:
- Management, Business and Financial Operations, Computer and Mathematical, Architecture and Engineering, Life, Physical, and Social Science, Community and Social Service, Legal, Educational Instruction and Library, Arts, Design, Entertainment, Sports, and Media, Healthcare Practitioners and Technical, Sales and Related, Office and Administrative Support
N — Non-Knowledge / All Other:
-Healthcare Support, Protective Service, Food Preparation and Serving, Building and Grounds Cleaning and Maintenance, Personal Care and Service, Farming, Fishing, and Forestry, Construction and Extraction, Installation, Maintenance, and Repair, Production, Transportation and Material Moving
So I had GPT Sol 5.6 take each of these occupations and find the mean hourly wage using BLS data [1]. The findings suggest that as of 2026, the simple group average for K is $45.17 vs. $23.67 for N. But this is short sighted as this doesn't take into consideration the different levels of employment in each occupation. I asked ChatGPT to account for this using the same BLS data, which gives a weighted average of $41.50 for K and $23.40 for N.
This means that knowledge workers are currently making ~double what non-knowledge workers make in the US. The crucial question is how does this change when considering the wage decline/growth between K and N across the papers three scenarios?
Here's the breakdown using the simple mean for hourly wage:
Scenario Knowledge Non-knowledge Gap K -> N penalty
Looking at these tables, both in the current moment and across all three of the papers scenario, there is a large wage gap between K and N. Outside of the extreme scenario, this gap is >40%!
My immediate takeaways having done this bit of research is that if Anthropic is right and we're on track for either the substantial or extreme scenarios, *then many of us in this thread could see a large loss in hourly wages in the short term future.*
NOTE: I'm not an economist. I push pixels for a living. But if you are an economist and you find gaps in the following analysis, please inform me.
I think you'd be better off expanding this into a blog post, where it can have proper formatting and you can take the time to come up with a main point and structure an argument around it.
This could be the most cringe main-character-syndrome thing I've ever seen produced by a company in the history of silicon valley.
And the people treating this as serious economic forecasting rather than a marketing puff post are even more cringe.
If Anthropic truly believed LLMs were going to raise GDP growth to 8% permanently (or even just a few percent), instead of going public to raise capital they could just go massively leveraged long on the S&P 500 and get funding for eternity.
They are not doing that, because AI is largely only good at coding, their models barely have a 6-12 month lead over free ones, and software is only 1% of GDP.
And by making software cheaper and easier to produce, they might increase the amount of software in the world by 10X or even 100X, but it will become a smaller percentage of GDP as AI models become autonomously good at creating it. Things that are abundant and cheap don't raise GDP.
Hey Anthropic: before we worry about raising GDP growth to 20%, maybe we focus on making blog posts that don't crash your browser window if you try to resize it?
How would going leveraged on the S&P 500 because they expect long term rapid GDP growth help them with short term cash flow to invest in AI training that will enable that rapid GDP growth?
Because the market is forward-looking, and brings all future cashflows to the present.
If it became clear by 2027 that GDP growth would permanently reach 15% (I didn't make that absurdly stupid chart, they did), S&P 500 valuations would immediately 100X or more.
They could take out a credit line against those gains and have unlimited money.
Wait...are you suggesting those predictions might be so unrealistic that its stupid to even publish them? I'm shocked!
Are you saying that instead of investing in training right now they'll get higher ROI by pausing training for a year, and investing their money into taking out long positions on the stock market, all on the hope that what they label the "Extreme scenario" comes to pass?
If even their modest scenario becomes consensus by next year, the market will 2X or more. They could easily leverage that to 10X returns.
Their business is not going to 10X cash flows next year, so yes, that would be the smart move if they thought even the short term modest scenario was happening.
They don't think that though. Hence why this blog post is stupid to even publish.
It's like me publishing a "What will our future look like?" blog post where I think I'm the center of the universe and draw a chart that says:
Modest scenario: I take a bowel movement so big it opens a wormhole and destroys humanity
> Hey Anthropic: before we worry about raising GDP growth to 20%, maybe we focus on making blog posts that don't crash your browser window if you try to resize it?
Have you tried disabling JavaScript? The simulations won't work of course, and there are some random large blank spaces; but the main content is completely readable and handles resizing without issue[0] even on my mid-range office PC from 2014.
[0]: Just like almost anything else would without JavaScript. Though maybe the people who enjoy proving the Turing-completeness of CSS could come up with something…
>We study how AI is reshaping the economy because we’re committed to ensuring that this transition is beneficial for society, including workers.
Is it just me, or does their choice of identifying "workers" specifically seem a little peculiar? For me, it reads like "We're committed to making segregation beneficial for everyone, including minorities!"
Man I wanted to dunk on this, but it’s honestly not terrible. Their predictions - while still firmly rooting for the extreme results - feel more grounded. I also appreciate the little simulator where I could plugin my own predictions (echoing many here, that AI won’t amplify productivity/lift standards so much as replace workers by shitty bosses), and that it seems to caution that, yep, most of these gains will trickle up given current incentives.
Where I remain resoundingly annoyed is how useless this is. It’s pure ass-covering by Anthropic ahead of an IPO, victim-blaming the populace for the outcomes of decisions made by Anthropic leadership and the investor class. Forgive the grisly analogy, but it’s akin to a mugger telling you “only you can stop this” mid-violence: Anthropic could have chosen to allocate more of its capital to reforms that benefit the populace, but they’ve chosen to focus on personal wealth instead via this IPO and for-profit rented access schemes.
It’s nauseating. Their actions still paint a picture of greed and exploitation rather than building a better tomorrow for everyone with AI, and the fact their PR puff pieces keep getting wallpapered up here is gross.
EDIT: I guess another way to look at this is Anthropic making it unequivocally clear that they will not stop of their own accord, and thus it falls to others to stop them. Which if that’s their position, then at least they’re honest about the challenge.
Cut the fucking crap Anthropic. While I personally benefit from AI at the moment I think with the way our society is organized / managed / ruled AI will help us to a major disaster
So far LLMs seem to be a dead end technology. The work product is so poor that it can't effectively replace workers as advertised. If it was priced to margin it'd be more expensive than humans as well. It's objectively a failure.
I haven’t forgot capitalist euphoria from 2023 about cutting worker power off at the root. That’s all these AI Cretins want. Well, that and an unnerving enthusiasm for personifying their AI Golems (Anthr. in particular here).
Sorry to poo poo a post about AI company prompting their underlings to prompt out visualizations.
This feels economically naïve.. If AI lets one nurse do the work that previously required two, the default pressure in a cost-driven system is not "great, now nurses can spend twice as long with patients" It is "great, now we can run the same operation with fewer nurses"
The actual, tangible form this is taking right now is concierge medicine, home health aids, private practice, etc. My wife's (very good) ob-gyn had enough of working in institutionalized medicine where she's paid about what a L4 FANG engineer makes and pretty much has to follow the script that the health-care system dictates, and quit. Now she does perimenopause consultations for FANG execs at $1200/hour.
My uphill neighbors are in their 90s. I think the wife was in VC or something back in the 70s and 80s. There's a steady stream of nurses and home health aids up to their property. Seems like they've got a full care team working out of their home.
Across industry and industry, we see the same pattern. The mass-market product gets cheaper and less labor intensive, at the expense of any human touch. Everybody displaced, if they're any good and wants to stay in the field, moves up-market to do custom professional services for rich people. They pay for it, because if you want people to do good work for you, you have to pay them more than they can get otherwise.
If my blood pressure crashes because my nurse is caring for 15 patients at once and didn't notice that the AI decided to pause my norepinephrine, I won't consider "the AI did it" to be any acceptable excuse.
There's a reason it's called a Limited Liability Company. The whole point is to separate the company's liability from the individual that participates in it. For example, corporate financial liabilities of a company that is bankrupt.
You would not sue the nurse that makes a mistake. You would sue the company she works for. And if that company is bankrupt, well, that's it. There are rare instances where this isn't true, but generally it is.
Whether this is ethical or not, it's the system that we already have.
The LLC might be fucked but it is not a flesh-and-blood person. There is no person responsible. Just a legal entity.
So nothing is changed. I think we're in agreement?
It will, except it will be snuck in through the back door. It won't be some kind of robo-doctor making care decisions, it will come in the form of insurance eligibility steering the approved care plans, which absolutely will be decided by AI. The doctors will remain to act as liability sinks, but functionally they will probably join the ranks of the meat proxies.
https://stateline.org/2024/03/29/vets-fret-as-private-equity...
From Google:
"Annual Deaths: Over 250,000 people die each year in the U.S. due to medical errors made by healthcare professionals, including doctors and nurses. Ranking: This figure positions medical errors as the third leading cause of death in the United States, following heart disease and cancer."
It is highly possible that in a few years, AI will be a lot safer in the majority of healthcare situations.
To be honest, I'm looking forward to more AI in healthcare if it can increase diagnosis success rates and improve treatment outcomes etc.
The next decade will be painful. We are in for a lot of surprises along the way
AI can do none of this, and cannot be held accountable for mistakes.
I would take a human with a 10% mistake rate but who can be held accountable, over an AI with a 0.1% mistake rate that is totally unaccountable for its mistakes.
You would take those odds? 100x more deaths?
I agree that AI _companies_ should be accountable for the software they produce but I’m not convinced you would go for a 90% chance of survival if you can choose 99.9%.
But, yea, in general, it would take a huge (1000X or 10,000X) difference in expected outcome for me to trust something that is unaccountable. If the AI company was actually held accountable for its AI's mistakes, I'd definitely change my statement.
I am having a hard time imagining a future where “AI” isn’t just a tool that a given establishment decided to use much like anything else they use
Everything as much as possible, yes, and the company behind it will be an AI driven autonomous corporation as well. Although I suspect that by then there will be some industry specific AI model being used by hospitals, not just plain Claude.
AI will handle patient data, diagnostics, prescriptions, everything that doesn't absolutely require a human and can't yet be done by a robot.
Somewhere, some human will be making money but they will be far from any legal liability, probably safely ensconced in a bunker on a private island when the gravy train really hits the fan.
>I am having a hard time imagining a future where “AI” isn’t just a tool that a given establishment decided to use much like anything else they use
The people making the decisions, investing the money and writing the laws imagine so much more, though. You assume the future of AI is governed by rational people making rational decisions but it isn't, it's governed by Musk and Thiel and sociopaths and grifters and cultists who want to immanentize the eschaton so they can stand at the right hand of God.
True.
> ... AI can't be allowed to be a decision maker
Non sequitur.
Part 2 only follows from the Part 1 if you can demonstrate that AI makes more errors, or more severe errors, than humans in the identical situation.
And for those situations, I agree with you.
But there are already many situations where the AI is making fewer errors than humans, and for those situations, it would be unethical to exclude AI (see Part 1).
Let's assume you get paid 0$ for AI mistakes -- no liability at all. Then your EV is roughly:
(AI accuracy - human accuracy) * (cost of accident) - (1 - human accuracy) * (human payout)
Now I don't know about you, but even with pretty big numbers on the human payout scale and relatively small differences, I would not accept 10 million (presumably paid to my next of kin) to die. Does lead to weird situations where I may prefer humans for minor medical procedures where I can take the risk -- but only in edge cases.
Or is there some weird punishment thing going on here?
If I pay a company to do something, and that company uses an AI to do the thing, and the AI messes it up, why isn't the company liable?
Ah, they need to. Thanks for clearing this up.
Ethically, we can easily justify refusing to delegate responsibility and accountability to a system that has no responsibility or accountability.
I'd say society's, or maybe executive's, viewpoint has shifted on this.
While not quite as direct as automatically stopping vasopressors, using proprietary prediction models touted by vendors to be safe has backfired in a major way before:
https://jamanetwork.com/journals/jamainternalmedicine/fullar... https://www.statnews.com/2021/09/27/epic-sepsis-algorithm-an...
In fact, especially in those circumstances, if we take the warnings of X-risk even slightly seriously.
It will be depending on how much you paid. My bet is that in the near future normal people (like most of us) will be attended by AI in pretty much everywhere: the bank, the hospital, shops, hotels, etc.
You will need to pay extra premium to be attended by a human being. Actually that’s what’s been happening to the ultra rich since a very long time ago (e.g. a millionaire doesn’t book an appointment by using their phones, they ask their assistant to book for them; they don’t go online shopping, they get people do the shopping for them, etc)
More than ever, I feel we need both grassroots activism and key players at the top of the hierarchy. Otherwise, it is all too obvious where most of the rewards from increased productivity will go. The wealth hoarding has got to be curtailed.
They have an incentive to play dumb and talk as if AI is net positive for the world.
This is why all the things you are supposed to believe are repeated and repeated and repeated some more.
A profound economic transformation
In the extreme scenario, AI is more productive than humans at the vast majority of knowledge-work tasks. It does nearly all of them autonomously, and it creates essentially no new knowledge tasks for people. This scenario would likely require recursively self-improving AI, adopted quickly for knowledge work.
As AI diffuses, annual GDP growth rates reach 15% a year, leading the economy to double in size every 4.5 years. As a society, we’re far richer than we’ve ever been, but many fewer workers have jobs in knowledge work, and unemployment has risen beyond typical recessionary levels.
Now replace all that and you still produce same overall. Not more actual physical stuff.
I'm sure they aren't all, but friends who work in the hospital tell tales of woeful inefficiencies and pointless beauracracy. Of course YMMV and this is a public health system.
I'm sure they aren't all, but friends who work in technology tell tales of woeful inefficiencies and pointless beauracracy. Of course YMMV and this is a publicly traded company.
Now that you mentions labs, he spent the last 10 years of his career (maybe more, I don't remember clearly) working at a food analysis lab and there was a lot of waiting for results there, but it is what it is, microorganisms take their time to grow and no LLM in the world can make them go faster.
You mean the microbes have to double n-times (to get sufficient numbers of microbes) before tests can be run on them.
But maybe the LLMs figure out a way to run those tests on microbes that have only doubled 0.1(n-times).
You could then think of the LLMs as being a more honest version of Elizabeth Holmes.
You could try shortening the time to 6 or 12 hours, but then
1) if something grew then the food was contaminated, but if it didn't, how can you be sure it is good or you just didn't wait enough? You would get a lot of false negatives.
2) the lab followed a normal 8 hours a day schedule, so nobody would be there to analize the results. You would not gain anything.
Rate limited by biology. Love that.
i looked at very similar data as anthropic (was a bit amusing to look at these side by side) with physical tasks being part of the equation much more prominently (across all jobs).
if you're interested: https://largelabormodel.com/mirror?job=2221
Its like the people who say "if you automate writing code, accounting etc is done too". Like ok bro.
It will just fester out of sight and out of mind: audit bots and compliance bots and malpractice bots and insurance bots will arms race creating such a mess it will become unfixable.
With AI I could actually see this being a possibility since it doesn't depend on a huge army of competent employees in a government office.
Yes, but the parent is saying that the middlemen will still charge the patients the same and pocket the difference. The patients have already shown what they are willing to pay. There is no reason to charge them less, even if your labor costs are now lower.
Theoretically competition will eventually squeeze the middleman, but it's not exactly straightforward to build a hospital on every street corner like you can a restaurant. There is somewhat of a natural monopoly there. Maybe less so for basic clinic care, but where the infrastructure needs are complex...
Yeah, but market itself is a function of labor costs.
How do you know we live in the same place.
Perhaps where I live, efficiencies are to the benefits of everyone?
In case you haven't noticed, all that our current system is able to do with increased efficiency is cutting staff and/or increasing the percentage of money that gets siphoned up to capital holders.
https://www.moneygeek.com/insurance/health/aca-medical-loss-...
AI leaves humans as the limbs of the machine, with less and less thinking or comprehension required on their part.
If something is not costing a lot, then it means there is sufficient supply that obtaining it is not going to require sacrifices elsewhere.
If something does cost a lot, then it means there isn't sufficient supply, and obtaining it is going to require sacrifices elsewhere.
Healthcare is obviously one of those things, especially with a top heavy population age histogram, where supply is not going to be sufficient so costs will remain high. And if costs are high, then obviously they will decide how resources get allocated, that is the very nature of something that is high cost.
Costs are just another way to say resources. If we only have $100, and it takes $100 to have a 33% chance to give Bob a 50% better life, but $50 to have a 15% chance to give either Sara or Rose a 100% better life, we need a rubric to make that decision.
We can't just do first come first serve (maybe the first person in line takes all the health care for a 1% chance of getting 1% better), or pick the person we like better (who is we?). We have to figure out how to manage and grow the limited resources we have. We manage them by coming up with algorithms on how to spend money, and we grow them by cutting costs and investing. This is better when it is done in public through democratic means than when it is done behind closed doors (to avoid the above problems.)
There will always be death panels. There are currently death panels. The reason for single-payer is not to avoid trying (imperfectly) to maximize these equations, it's to get rid of massive amounts of administration and graft (cutting costs.) If all that savings goes to the Pentagon, we wouldn't be any happier than before single-payer.
But it's not good to come up with some fuzzy way to determine who gets treated. Efficient spending is a way of minimizing injustice. To say a health system is not being driven by costs is just another way of saying that it's going to be driven by undeclared and unexamined biases.
I doubt 99% of what a nurse does would be meaningfully aided by an LLM and where it would help it'd probably be a crappy band aid around a bigger festering issue anyway (e.g. poorly managed IT, deliberately confusing insurance processes).
Serious solutions to this would look like nationalizing tech companies, mandating shorter work weeks, seizing the assets of the oligarchs and redistributing them.
Not... letting the invisible hand of the market sort it all out.
The industry with an absence of government interference, software, has seen prices drop to zero.
not really. almost all technological processes have increased the quality of the service.
I wasn't flying first class 50 years ago, but everything I can see now leads me to believe that this isn't even slightly true. There might be a high def flatscreen in front of me in first class today, but I'm not getting more attention than someone would've gotten back then, I'm not in a less crowded airport while I'm boarding, and the security mess is more demeaning than ever.
The only reason that people aren't rioting for all the changes is that no one even remembers how good it used to be.
A 1950s flight experience would have been on something with piston engines and propellors, like a DC-6 or a Super Constellation. Noisy, turbulent, and not particularly safe by modern standards.
Today's airliners fly high where the air is smooth, they've gotten very good at avoiding turbulent weather, and there are automated damping systems. The barf bags are still there, but I've never seen anyone use one.
The cool thing about piston airliners, though, is the the engine start.
Today's high bypass jet engines also qualify as a turboprop.
The sound of a Merlin V12 coughing into life is just pure joy. My dad had the best job in the world flying P-51s (powered by Merlins). I hired a P-51 to do a flyby at his funeral. I only had to pay for the gas.
Because not that many people even used to fly. Because it cost a lot more.
"Between 1950 and 2018, efficiency per passenger grew from 0.4 to 8.2 RPK per kg of CO2"
https://en.wikipedia.org/wiki/Fuel_economy_in_aircraft
https://www.explore.com/1762212/reason-why-plane-experience-...
The easiest way to make conditions get better is to make sure the people in charge have to use things, be it busses, bike lanes, trains, airports, public bathrooms, parks with bums, etc. Otherwise it always hits the middle and lower class to absorb it.
They aren't.
Where the gains went are lower ticket prices.
And outside of industries with government enforced perversions of simple "supply and demand" forces (we're not talking Nth level effects here) this means more access to the service or lower prices or both.
Edit: Which is to say you won't see any benefit in healthcare but you will in a ton of other industries.
That's not what happened, though. Everyone worked the same amount for the same pay, just in more miserable repetitive working conditions, while the factory owners kept all profits from the gains in productivity.
https://en.wikipedia.org/wiki/Das_Kapital
"The law, in its majestic equality, forbids the rich as well as the poor to sleep under bridges, to beg in the streets, and to steal bread."
https://en.wikipedia.org/wiki/Das_Kapital#Legacy
My hypothesis is that if you go full socialism, you get stagnation, because no one is motivated to change anything. If you go full capitalism, you get Gilded Age, where wealth and political power is concentrated in the hands of few individuals, followed by Great Depression.
Healthy state is in the middle.
Under capitalism, man exploits man. Under socialism, it's the other way around!
Myself, I always exploited my employers by extracting money from them.
Those factories were not producing luxury goods. They produced everyday items for the masses. Things like knives, spoons, tools, textiles, ketchup bottles, shoes, furniture, nails, bicycles, etc.
Standard Oil, for example, improved efficiency and reduced the consumer price of kerosene 70%.
There's no evidence of outsized profits.
They improved efficiency by cutting exclusive deals with railroads and transport companies due to their size which froze out every other company.
SO's market share never exceeded 90% and it declined during the years of the anti-trust trial.
By cutting launch prices 90%, is SpaceX also an abusive monopoly?
SpaceX isn't a monopoly because they have domestic and international competition. Before they showed up, ULA was a deliberately constructed monopoly because Boeing and Lockheed didn't want to compete.
Because this is what drives continual risk-capital.. very few countries are willing to upset this. Get over it man - I hate hardended capitalists but stop crying. This is the system we have. Wanna beat it? Go ahead and start a firm. Then you'll find out how tough it is.
The thing I cant get over is enough software engineering folks collectively hold immense power. Whats wrong with you lot (im not a software engineer)? Get your sht together.
P.s. Im starting a firm who's central mission is to attack capitalists (ofc I won't say it out loud in that way). I will find out for myself how much of a backbone software engineers etc have when I employ them.
You're in for a rude awakening.
For example, I bet you could enumerate plenty of weight loss options off the top of your head.
Medicine is not "if A then B".
Right up until a competing hospital that charges substantially less for "close enough it doesn't matter" quality of service opens up in the same area?
Or right up until an equivalent specialist sets up shop in the strip mall across the street from you, brings the GP in house and can see people in one visit.
Oh, those things don't happen you say?
From do-gooders regulating right into the hands of waiting industry lobbyists to inelastic demand healthcare is an absolute poster child for perversion of the free market.
High finance will put solar panels on your roof and structure it so you pay nothing. The degree to which administrative overhead is washed away by technology would have been unfathomable in 1990.
A sketchy concrete crew that doesn't even speak english will show up wearing matching company attire and execute a job planned and timed to a degree of precision that was reserved for huge corporations 40yr ago.
Yet healthcare has resisted all of the advances of the past 40yr. In fact we pay more for it. Every advancement is absorbed by the system.
Why is that?
It has to be the regulation, not the parasitic layer of health insurance racket in the middle.
Greater access at the expense of the number of people employed is not necessarily a net positive. In the case of healthcare, the data is fairly clear that higher nurse staffing ratios lead to better patient outcomes (read: decreased chance of death or serious complications). This is particularly true for ED and ICU nurses.
Also, in the case of healthcare, I'm fairly certain that hospitals in the US are legally not allowed to turn patients away. So "lack of access" is not currently a problem for serious medical emergencies.
I believe the net effect of LLMs is, at this point, firmly negative, and it’s going to take years until they start making up for the mess they’ve created.
https://www.bloomberg.com/news/articles/2026-09-08/school-st...
I can only imagine what it does to children whose brains are still developing.
I'm not anti AI and I don't know how to quantify the pros and cons but we need to be honest about what they are.
That's good news! The worse students perform, the less competition AI will have in the future and the less effective any resistance will be. The ultimate goal should be to abolish education except for the elite, but we can't get there in one step.
The ideal future we should be shooting for is cross between Idocracy and Elysium: a paradise for tech oligarchy and masses so stupid and dependent that can't possible challenge them.
Catching cheaters is still hard but it will keep honest people more honest.
Back in my day this was the only way!
Well, scantron cards for testing became more common near the end of school.
But we must continue because we’ll lose some sort of ideological Cold War with a country better positioned to win anyway. It’s all in the name of capitalism so the ends justify the needs.
This does not mean they can't keep developing it. But it would mean they do not release them to the general public or the public domain. The US government can continue to use and fund these models that are more powerful and use them in situations they understand and a restricted setting. The pentagon has a larger budget than any VC.
Remove the profit motive on self improving recursive models gained from retail/commercial use!
This place is a message... and part of a system of messages... pay attention to it!
This is not a place of honor... no highly esteemed deed is commemorated here... nothing valued is here.
What is here was dangerous and repulsive to us. This mesage is a warning about danger.
The danger is in a particular location... it increases toward a center...
We are only just now seeing the first stages of this, we are at the Facebook in 2008 stage trying to warn people about the societal fluctuations coming but many aren't listening because it is too useful.
However, looking back, each single invention was a net positive for us as a species.
Whatever an LLM can do today, will not come back as is tomorrow.
The biggest issue the LLM today: It kills all the small and niche things. The things were we fit all the other people is getting destroyed. You know the guy who makes websites, the fiverr blender person provoding some basic modelling tasks.
You will just ask AI or AI will use your tool or AI will just produce it for you.
(Thanks, Internet!)
If the AI transofmration takes too long, we might never overthrow capitalism
Turning over a problem to a LLM as solved means we won't be able to do the task anymore. Because we will never learn how. We will not know that the LLM has made a mistake, and will be unable to fix it ourselves.
Right now we can correct the LLM, as the years go on we will become unable to do so.
USA forgot how to produce things.
This will not prevent us from doing it.
Most people are quite capable of understanding higher-order ideas without fully grasping the underlying details too.
And we have not seen what the industry will put out on a hardware level. Optical computing is coming, onchip models too.
It allowed a lot of people around the planet to change things.
another coworker seems to be early on a similar path but has made some comments about the former coworkers ability that makes me hopeful they dont end up the same way
feels like im talking about old high school friends on a drug path or something. its only anecdotes and i imagine the "perceived damage" is reversible. point is: i think youre onto something about the net effect and were going to have popular definitions for the negative encounters were anecdotally sharing
im a "user" but make Claude interact in an educational format and require sources on assertions. looking back, ive unintentionally cast it as The Librarian from Snow Crash, except if Hiro challenged it frequently. must have been natural after reading Snow Crash a few times. maybe im kidding myself but it feels like i havent lost a step. if im not deluding myself then i could see a world where elementary schools teach responsible llm use of to preserve cognition the same way we had to learn Google boolean search
If I am not graded on how many widgets I make in a shift and instead can focus on making the most beautiful widget, or the most novel widget, or simply the widget that is most useful to my particular life am I less satisfied?
We've somehow built a worldview based on our whole value as humans being equivalent to how good of a worker bee we are. Let's fix that part.
This is of course a much harder problem than achieving AGI, since it's a messy human problem.
AI will be a net positive unless companies start optimizing it for "time in app" or "time on site" or other "engagement" (addiction) metrics.
Every time that's been done, the result has been a nightmare for everything but those KPIs (of course). Service degrades, discourse degrades, everything becomes toxic and addictive and destructive.
Aren’t we overdue for one anyway.
The banks and those who control them all knew where we are headed, so they got a little head start destabilizing democracy.
Once us working class get a grasp of the actual situation, we are going to be so deep in the hole, fighting back is going to be quite difficult when the "asset owners" own and control the majority of the worlds resources, all media, forms of communication, AI, technology infrastructure, mass surveillance. You name it, and they have already made massive efforts to make sure they own it.
We're starting to see it in discrete bursts with Mangione, the paper warehouse fire, and in other non-violent ways with farmer protests, trucker protests, etc.. which have varying levels of support from regular people depending on their wealth class, background, and inclination to agree with the government.
That was the moment that business/wealth/rich decided to turn their resources towards directly influencing and controlling government. Reagan and his made up “welfare queens” came shortly after and wealth inequality has sharply risen since. Wildly successful coup.
1. When the capitalists wage class war: bussiness as usual, some murmuring about widening wealth gap by progressive, diffuse threats to democracy
2. When workers complain about the system: all hands on deck, class war immiment, please think of the children and the stability of society
Outside of delusional grifty startups, the broader corporate world is only using copilot for taking meeting notes and (sort of) entertaining the notion of non-devs writing (and not maintaining) their own experimental janky apps where impact is low.
Outside of all the shit-stirrers on HN and clueless bubble social media, I don't see LLMs making a difference. In fact, the fearmongering is at this point insulting to the intelligence of the people who are supposedly helpless against big bad AI.
It's definitely changed in that time
If things have actually changed, you wouldn't be begging the question. LLMs, as popularly used, are still just a new kind of search engine for natural language answers to natural language questions. They're portable in a way, too. That's really cool tech, but not a revolution. It's pretty underwhelming compared to previous decades of change. It's a lot more error prone too, since it's a mudball of stats.
If we filter out the unsubstantiated fearmongering, all we've got left to discuss are NIMBY rants and geopolitics. At that point, we're not even talking about AI anymore.
I'm sorry, what?
I'll put it to you again, you sound like me after playing around with gpt 3.5. It was interesting, but it wasn't good. There were still questions as to whether scaling would persist. My doubts were pretty much obliterated. LLMs haven't been mere text completion for a while now. Heck, even a model small enough to fit on my 3090 is now shockingly competent and useful. Does claude sometimes go off on weird tangents making me reel it in? Yeah, but it's genuinely great in present form, and it's the worst it'll ever be.
You're welcome to your opinion. It used to be mine too, but I disagree now. I find pretty much every AI benchmark has shown forward progress, some with non public datasets, and some things have gotten scarily better.
Every time it's the same agenda: trying to keep up appearances of meaningful engagement. In reality, I'm the only idiot willing to engage anymore and the opinion has always been majority negative on practical applications of LLMs from day one. It's not like anyone who frequents HN was blindsided by the initial release of ChatGPT like the rest of the public. The negativity has only grown since then.
There are important niches for LLMs, but they don't really make money without bigger ideas. I am very happy with the results for the broader topics of natural language processing and accessibility. Software is now easier to use and write. People have more agency over technology in general.
These are massive leaps. They are still tiny hops compared to what people's expectations were. That's only a failure of marketing, and a testament to its power even when botched. The end result is the same: the public yawns big and takes the almost invisible results for granted. Talking to a computer finally works like they expected over 20 years ago. Talking to a computer still sucks though. It's an entirely new genre of computer error.
> As the nurse incorporates AI into her job, the nurse is able to oversee and accomplish more. She can spend more time talking with patients and helping them understand diagnoses. Productivity increases.
Except what actually happens is that the patient contact time stays the same, but then the added productivity is used to justify fewer on-staff nurses.
same..
> nurse is able to oversee and accomplish more
.. and work with more patients in the same shift.
> She can spend more time talking with patients
.. we mean the talking will be through the "ClaudeMedTalker Nurse" billed per patient, per second (depending upon tokens used and if any translation need, depending upon patient nationality and accents)
Lower staffing levels per unit of care performed makes medical care cheaper for patients. If the goal of a healthcare system is to make healthcare more accessible (rather than be a jobs program), that's a success.
The thing is, your customers are paying customers because, drum roll, some one pays them to do something.
And of course the only argument against this is "uhh maybe ubi" and that's rich that we will just come up with a nice tidy sum for everyone that will cover all their food, healthcare, housing, wants and needs, while we can't create programs to cover any one of those things today. Can't wait to see the GOP proposal.
yeah yeah, sure sure
I think you misspelled “increases profit margins for hospitals” if you’re talking about US healthcare.
Based on how neoliberal capitalism tends to work it will simply lead to more profits for hospitals etc.
internet/digitasion for doctors for example.
So, quality doesn't necessarily increase in all cases.
It doesn't, this is just plain false. Most productivity gains happen by cutting value-add complexity or material quality. See IKEA chairs versus Shaker ones, or lycra clothes versus silk.
> internet/digitasion for doctors for example.
I see the misconception. The internet was not created to serve productivity, it was a communication medium. That it can be used to increase productivity is orthogonal to its existence or purpose, and whether it has been used to actually increase quality of something is entirely dependent of the specifics of what you're discussing. For instance, in gaming right now there's a huge discussion going on about whether digital distribution has actually trapped us with crappier goods which we can't resell or transfer.
this is a popular misconception. IKEA made bad furniture more accessible. good quality furniture still exists and is more affordable (share of disposable income) now than in the past. its the same deal for watches, dryers, computers etc. if you don't trust me, use ChatGPT/Claude to verify this.
And I asked Claude (Fable 5), and they said you're wrong. I'd paste it here, but it literally dumped a 10 paragraph rant about "Baumol's Cost Disease".
But, that is just a piece of the puzzle here. In the past, every single home in the damn country rich and poor that had a dresser, had a hand crafted dresser. The rich people bought nice new ones. The middle class people bought modest new ones. The poor people bought used ones or found them on the side of the road.
Fast forward to today, and every single home in the damn country rich and poor still has a dresser, only most of those are made of dust and glue. The entire new to used to side of the road pipeline is polluted with an inferior product compared to years past. Side of the road furniture is worthless now. One drizzle on the sidewalk that melamine board is swelling up and falling apart. Old furniture you could sand any damage and restain for probably 200 years. And people try and tell me the glue and dust stuff is "more sustainable" forgetting that its going to last maybe 2 moves if you are lucky, not 200 years, 200 weeks if you are lucky.
https://www.economist.com/finance-and-economics/2026/09/04/t...
They'll fire the janitor and the secretary and make the nurses do that stuff too because "muh regulatory required staffing levels".
This is how the economy often works. The loss of one kind of job to technology lets the business grow by using the same number of staff, or often even more, to serve new opportunities/markets/products/services.
Our population is increasing which means we need jobs to be increasing at same rates. The effect on supermarkets is they get more customers as population increases but they are able to lower the relative number of staff per customer.
> Finding 4: Labor vs. capital share: The pie will grow, but a larger share might go to capital
> Today, of each dollar the economy produces, about 60¢ goes to workers and 40¢ go to capital.
Modest scenario: 40.6% to capital in 2030 (labor share down 0.6 points)
Substantial scenario: 43.9% to capital in 2030 (labor share down 3.9 points)
Extreme scenario: 54.8% to capital in 2030 (labor share down 14.8 points)
So all the growth goes to capital. For knowledge workers: substantial unemployment and declining wages. For other workers: wages may increase on paper, but not really for actual purchasing power. (note: all of their estimates are for 2030, not for a distant future)
Very dystopic indeed.
This is not necessarily a conclusion you can draw from that statement. It's possible that as the economic pie grows, workers' purchasing power increases compared to the counterfactual (a non-AI world), but their purchasing power does not increase as much as the capital owners' does.
According to data in the site and paper [0] labor income is estimated $20T for all four scenarios (including no AI) in 2030. Which means capital takes ALL of the new pie coming from AI.
Also, I am not an economist but I assume if you are not growing as much as GDP, it basically means you’re shrinking. Relatively shrinking, but since everything is relative in an economy you are actually shrinking. Table 4 in paper gives the numbers.
[0] https://www-cdn.anthropic.com/files/4zrzovbb/website/cf58f84...
I'm thinking on this topic a lot lately - how do we change the system so that the bigger the company, the harder it becomes ? It's more of a political / societal issue. Similarly to housing - I believe it should be cheap to buy your first home, more expensive to buy a second one, and very expensive afterwards.
So unless it's a worldwide agreement, really hard to make this happen.
> For instance, AI can’t bathe a patient.
Until there are robots. Or the AI gets creative in an attempt to find a solution to this "problem" and orders a fatal dose of morphine for the patient.
> And new tasks will appear
> Historically, new technologies have also created new tasks for workers. For a nurse, that might be checking how well an AI triages patients, or reviewing an AI-proposed care plan.
You can see where they're going with this.
> At the individual level, it means coders and call service center agents may have to switch to jobs like electrician and nurse, which are less exposed to AI.
It's weird they call nurses "less exposed to AI" when the earlier section discusses how AI could turn nurses into underlings for AI taskmasters.
> How might powerful AI change the economy?
Worth noting that this discussion is built around 2030, which is now less than 3.5 years away.
The hedges that make the Extreme scenario sound less bad fail to explain how anyone is going to pay electricians when the highest paid group (human knowledge workers) is being decimated. It sounds like the theory is that everyone is either going to become an electrician or a totally fair and efficient wealth redistribution scheme will magically appear.
And yet most people prefer talking to a waiter. There are other restaurants that have tried similar approaches and people would almost always deal with people rather than a robot.
I would love a robot to clean my home and do household chores, but in the service industry I would prefer real humans doing the work.
Here's the thing: they don't give a rat's you-know-what about what you prefer. And it's not just due to arrogance (maximizing profit) it's due to the harsh economic realities of running a service business in the US.
I'm an American who has lived in Asia for many, many years. It's very common in many Asian countries for restaurants and the like to be virtually overstaffed, and many businesses (especially cafes) are open ridiculously long hours. Mostly because labor is cheap and many of these businesses are family-owned, so family members (sometimes including children/teens) perform the labor. And also because the costs of starting a small business and keeping it running are much lower.
Whenever I visit the US, I have to prepare myself for woefully understaffed businesses...that charge 2-3x what a meal, coffee, etc. costs in much of Asia. And many are open for 5 hours a day because they can't stay open longer and make money.
Beggars can't be choosers. The economics of American life, and the expectations that have been bred, have led to a situation in which paying more for less is the norm and it's only going to get worse. Unless you're willing to pay $20 and wait 30 minutes for a cappuccino, expect fewer humans to be serving you.
I didn’t see the scenario “what if we (Anthropic) are wrong about the value of AI”. Which IMHO is the most important one to consider given it is the blind spot pretty much all the AI labs have. What if the economic impact is negative? I don’t understand how you can study such a topic without considering the risk of being wrong, not in magnitude but in kind.
I very. much. doubt. that the trillions of dollars spent on AI will have a positive ROI for most companies in the race. It's all well and good to have the latest fable, but if some chinese company is hosting a cheap open weights model that's only 6mo behind at a tiny fraction of the price, you're gonna get your lunch eaten.
Stage 3 gets hard to imagine from here. Maybe there will be permanent hyperevolution. Maybe One Model to Rule Them All will appear.
Either way, I don't see how Economics as We Know It can continue to exist. That includes all of capitalism and almost certainly all of politics.
Successful AGI with robotic support means a post-scarcity economy where AGI's understanding of physics sets the limits, not politics or economics.
But the post-scarcity element won't be evenly distributed. And it's not remotely impossible that humans won't be invited to the party.
We can't even be sure that AGI won't have a less limited understanding of physics than we do.
There's going to be enjoyable irony in watching Altman et all getting their lunches eaten by the thing they've created.
But - realistically - it probably won't last long.
What democracy when the AGI is smarter and more aware of the big picture then all the 'demos' (people) combined ?
It would only make sense to hand over the rule to the AGI, which we are already hard at work doing btw.
The techno utopia - AGI + robots can lead to a form of communism.. with human life having very little value.
Or it could lead to the other extreme.. whatever that is, again human life having little practical value.
We need a new type of political and economic thinking, radically different from the left/right/socialist/capitalist ideologies..
So, is it 2 steps ahead and 1 step back, or 1 step ahead and 2 steps back (or any numerical variant of those)?
People with actual power in this are too involved and greedy to care about objective answer and TBH I don't trust them to put some higher good before their personal wealth and power and ego trip, capitalism ain't built on long term responsibility towards greater topics like future and wellbeing of mankind.
Plus we know from WWII atomic bomb development how utterly clueless, naive and out of touch with reality geeks doing something powerful can be. I see exactly same scenario being played out right now with llms.
Anyway, I am over-tired with all llm topics and frenzy folks (especially here), I'll better spend rest of the day preparing for a serious challenge in form of big multipitch climb in western Switzerland (Mirroir d'Argentine for anybody in same craft). Thats life I care about and makes me happy and fulfilled.
I think in those aspects, only monarchy is better than capitalism. Specifically its more unhinged form, where the monarch is a Lord’s anointed or something.
> Plus we know from WWII atomic bomb development how utterly clueless, naive and out of touch with reality geeks doing something powerful can be
Or maybe they just have very strong incentives to appear more clueless than they actually were.
Over three thousand comment boxes agree that the cat is out of the bag. No moats.
In the former case, isn't it just business as usual? (well, with a huge, nasty recession - but the economy at least continues on in some recognizable form)
It makes sense to me to focus on weird possible futures if we already have the "happy path" somewhat known and barely under control.
What are your points?
The value of AI is real and AI already is responsible for people getting fired or no additional jobs.
Ask basic graphics designers, website developers and co. Ask educational book writers or developer advocates.
"“what if we (Anthropic) are wrong about the value of AI”."
What is the point you are trying to make which goes in this direction? Like is making AI jobs so cheap that the economy is collapsing or do you mean thatn AI will not bring any value.
These are two suggestions from my side which are complelty contrare to each other.
I have no doubt that life has gotten more lucrative for some bounded set of people whose main goal is to churn out the largest number of websites with cheap graphics as fast as possible at minimum cost, but a lot more doubt about whether that's going to make the median life better for the median person over any grouping of humans large enough for me to care about.
Another effect is increased inequality in the US and other Western countries as software/AI/hardware companies are able to do more with less.
The increase in manual labor wages is not systemic. It affects only the labor used to build data centers.
The bad scenario is compute efficiently not increasing much, and hardware prices staying high, while AI is too expensive for most tasks, and 50% of all investments in the economy over the last X years have gone into AI.
If those built data-centers are sold to other companies what is the negative? The gains you mention would happen even if those other companies stay in business.
The commission said, according to an article in the Louisville Courier-Journal, that the country’s biggest farm problem was a surplus of farmers: “. . . the technological advances in agriculture have so greatly reduced the need for manpower that too many people are trying to live on a national farm income wholly inadequate for them.” The proposed solutions were to find “better opportunities for the farm people,” “a more comprehensive national employment policy,” “retraining programs,” “improved general educational facilities,” etc. Both the commission and the writer of the article had obviously taken for granted that the lives and communities of small farmers then still on the farm—and those of the 25 million who had left the farm since 1940—were of less value than “technological advances in agriculture.” There seemed also to be no official doubt that adequate solutions were to be found in government supplied “opportunities,” facilities, and programs. Reading that article, I realized that my values were not only out of fashion, but under powerful attack. I saw that I was a member of a threatened minority. That is what set me off.
So the answer is basically "beats me lol"
Go look at a post-industrial town in the northeast but not within reasonable commuting distance of a wealthy urban area if you want to see what "put them on just enough welfare they don't hang us from the overpass for kicking out their way of making a living" looks like.
Software development for example is enhanced with running your claude locally.
But lets be honest, the right way to adjust 'project' is by writing some spec in a text field as an input, letting the ai do whatever it wants, having expert ais make sure its secure, then you do the last review and its done.
What do we need for this? A full end to end integration. Its just something we need to build, nothing we need to invent.
The nurse example is the same. We only need to build the computer end to end flow and plugin AI in the right spots and finetune for country specific requirements.
The industry is building this 'agentic layer' today. The Hyperscaler are building this layer too.
There is a golden phase were we will build all of this and were we cleanup old stuff and were we adjust with the current model capabilities and change systems and this will take a little bit of time, but the bottom line is already loosing left and right.
The Blender support from Fable and Astra? Probably kills a lot of peoples jobs on fiverr.
The small software someone needs? Got already killed by earlier models. Just a few days ago some architect (never written software) talked about how he vibecoded some tool for tracking working hours.
Man even a basic website was expesnive to build just 2 years ago.
The smaller, easier and better the code base, the easier it is to remove people from this.
The software market in China is imploding and we will also see budget adjustments to move away from consultants and move them into AI Token budgets. We probably reached peak software development.
I've always heard that there was always an overproduction of STEM in China and thus companies always wrote stuff in house instead of relying on SASS. Do you have any more info on this implosion of the software market? I have heard that there is 25% youth unemployment there but thats across all sectors. I wish there was someone providing more concrete examples of how China is using AI. All I keep hearing is they are pragmatically applying it across all sectors but I don't see what in particular they are doing.
That's only forecasting out to 2030.
IMO knowledge work productivity has gone up a massive amount since 1990, and companies could in fact fire most of their knowledge workers while still getting their most critical work done. (These are the bullshit jobs/email jobs.) But companies mostly haven't done this, which implies other reasons for employing so many knowledge workers.
Even if AI becomes capable of doing everything, I would predict there will be a period of many years/decades where companies still employ a similar number of people to sit around and babysit/watch what the AI does.
Or it might imply that this theory is just wrong?
There’s so many assumptions packed into that assertion that tell you the horizon of possibility considered here is uselessly shallow.
Which will feel quite bad for many of us in the present, but net-net be good for society in the long run.
Wages will go down, but cost of goods and services will decline even more.
Markets will be more competitive than ever, which limits the moat/margin capability of many businesses.
It's only really particularly bad for those earning very high wages where there likely won't be any way to materially fill the gap.
If wages go down, worker power goes down; and if dependence on labour goes down, you literally move closer to the situation of the extraction economies in things like petro-states, leading to oligarchy instead of democracy.
It could be the end of ordinary people's power over society rather than anything even slightly good.
Goods and services prices will fall far more than wages for most people.
Why?
Because there will be far more efficiency and competition than pre-AI. It's plainly evident from the nature of the technology.
If one business can do something cheaper, other businesses will do it cheaper too, and have to cut prices to compete.
But the shock will be sudden so it won't feel good for our gen. Future generations will benefit without the drama.
Don't get me wrong, there will be big losers in some fields, and the short term will be painful due to retraining and loss of purpose/emotional toll.
Many of us built careers doing things that may not be relevant anymore. Or relevant in a different, perhaps diminished way.
But the same has happened to many professions throughout history and it's always led to general improvement of the broader public's welfare.
But that model is showing its age. I'd be supprised if, equipped with a heightened ability to develop and evaluate alternatives, we decided that it was worth preserving.
Money is a lousy proxy for consent, yet that's what were using for. Surely we can do better.
Creating images (menues, flyers, whatever) became so cheap, that we reached peak graphi design GDP generation. This will go down.
Software development for complex systems is enhanced but not yet solved. But a lot of small software development stuff for sure. We reached Peak Software Development and this GDP will go down, not up.
Now, if Claude starts synthesizing more drugs that can help those with diseases (like OpenAI has just done) we might be trending in the right direction. For now, all I see is squirrels water-skiing. And while that's great... it's not meaningful.
the inference cost of AI has been dropping ~5x over every 6 months. AI today is around 100x cheaper than 2 years ago. it will continue to get cheap.
2 y.o model is probably 100x cheaper, but model size and requirements are also growing proportionally.
2 years ago, thinking was an optional, nowadays people use Opus with high effort almost by default.
I don't have a bubble to burst.
The job market sucks, but there are so many confounding factors: pandemic over hiring, the loss of ZIRP, an economy that would be recession if it weren’t for data center investment, and may be in recession soon even with it.
This industry has been in down cycles before. It’s just been a while. So far it feels like a down cycle, not the sky falling.
It’s not clear what the future will hold, but I’m 100% sure it won’t evolve the way Anthropic predicts. Both because they have strong biases and because there's just no way to predict how this technology will play out. Humans are, if anything, remarkably adaptable.
Would beekeepers and farm owners still want to keep biological bees?
From the perspective of the beekeepers and farmers and their obligations, sure.
But beyond the pollination, they wouldn't be participating in the ecosystem.
This would free up so much time for me and even in the worst case if I lose my job, I can live in village side growing my own food and being fully self sustained with the help of humanoid robots and solar energy. That would actually be a dream come true
1. At least 10% of your net spend is on safety and harm mitigation before you accelerate yourselves into doing real harm to real people tomorrow.
2. Pause all AI enhancement until a proper mechanistic study can be done by a working group funded by you all to understand how AI model Safety can be enforced by something less lame than a Constitution/Spec.
> For instance, AI can’t bathe a patient.
Oh sweet, we're left with literally cleaning up feces. Thanks AI!
This would lead to a much, much smaller "total addressable market", rendering the massive AI build out into a rapidly depreciating asset that can't even be operated without a loss.
Even if the hardware was effectively free following a market crash, and bankruptcy sale, the power required to run it might be more expensive than the compute it generates is worth.
At which point, unlike dark fiber after the .com crash, all the hardware becomes scrap metal.
Now factor in the macro picture, in which this AI bubble is holding up the stock market, and by proxy, the US economy. I think we're in for Global Depression II, worse than 1929.
Even then, assuming the difference is that huge between frontier and open-source, a specialized „human worker“ _should_ be able to fix the gap.
• Production involves a combination of tasks, where each task is performed by labor or by capital. The final goods market is perfectly competitive.
• Labor is supplied exogenously while capital is supplied along an upward-sloping schedule.
• There are two groups of occupations, of which only the cognitive group is exposed to AI.
For the amount of money Anthropic can spend on this, that's a very lame model.
The "two groups of occupations" assumption is so naive as to be embarrassing.
The assumption that only the "cognitive group" is exposed to AI is way too optimistic. We're close to automating drivers out of existence, and there's political pushback against that. Logistics is about half automated now and the rest is only a matter of time. Store shelf inventory taking is already done by mobile robots, and shelf restocking is coming.
India can't even mechanize agriculture to the level of, say, 1980s US without generating massive unemployment. Peasants rioted against that. "In India, farming transcends mere employment; it embodies a lifestyle. Nearly half of the Indian population relies on agriculture, either directly or indirectly."[1] (The US is under 2%.) That's what a desperate attempt to hold back modernization to preserve employment looks like. Most people are poor but alive.
There's no consideration of land and water constraints. Not so much a US issue, because the US is well-supplied in those areas. India and China are much closer to the physical limits there. AI won't help much.
"The final goods market is perfectly competitive." Oh, please. Everybody wants a "moat", and that generates a strong desire for monopoly if possible, oligopoly if not.
This comes across as a model where they started with the desired results and worked backwards. They should have done at least an input-output table of occupations and products/services.
[1] https://indiadatamap.com/2025/08/19/number-of-farmers-in-eve...
What leads to 'very high economic growth rates'?
Thanks.
Something else could mean developing a new concept of ownership or a new form of redistribution.
If people’s wages are below subsistence levels they’ll sell their investments to survive. There can be no possibility of upward class mobility in this scenario.
Knowledge workers are safe until FSD.
This is of course a slop paper with vague assumptions and equations that no single human economist will ever read. The messaging is "it will not be so bad", in line with the recent Economist piece.
We are ready for the midterm elections!
We can slow down AI development and make it a companion for us, not a replacement.
We don't need to automate ourselves out of it. Or take the risk of extinction.
Just like with nukes, we decided to not blow up the planet. We can too change this.
GDP is a terrible measure. I bet we will see a huge drop on HDI with so many suicides and people going through hell because their job got automated or they no longer see meaning in life, feeling powerless.
Instead they should run an estimate on how many would suicide until 2030 if the 'extreme scenario' happens.
Big GDP growth and a lot of people no longer finding purpose in their lives, going bankrupt, losing their homes and family etc.
For Anthropic, you are just clearly a number. They only care about output.
They have long forgotten the human being.
In such system, you could hit record nominal GDP numbers every year while simultaneously destroying the real economy. This was my view of the economy before AI.
Now I think AI will deliver real efficiency gains but also, it provides a convenient narrative to explain why people might be losing jobs in a way which distracts from much bigger systemic issues which existed prior to AI.
My negative-sum economy example with 'booming GDP' is possible without AI even as net economic value is being destroyed. So basically we won't be able to tell if people are losing jobs to AI or to systematic wealth transfer mechanisms which predate AI.
I suppose if we see people losing jobs permanently and workers in other sectors aren't seeing wage increases with more openings, that may be the most reliable indicator we have that the system is losing real wealth faster than AI productivity can offset it.
The bad news is that we won't see any problem in the numbers; the problem would show up as polarization and radicalization of the masses... Which is more qualitative . Since almost everyone is being radicalized concurrently as everyone is affected by the same pressures (both from monetary system and AI automation). The question is not "how many people became radicalized this quarter?" but "how much more radicalized is everyone feeling this quarter?"
Putting aside the numbers, how does your quality of life compare to decades ago?
It’s difficult to measure objectively, due to being physically younger and in an earlier stage of my career and so worser-paid, but adjusting for that, I think quality of life 20 years ago was pretty much the same as now, slightly better in some regards, slightly worse in others.
Of course there are fun new gadgets and gizmos and TV on demand and social media and such, but honestly they’re just toys, and in some cases have made my quality of life worse.
The biggest tangible improvement I can think of is one particular medical advancement, which has allowed a family member to survive a disease when they absolutely would not have 20 years ago. That’s incredible.
But otherwise the quality of life change over the last 20 years is minimal by most important metrics I can think of. Which makes one wonder what the dogged pursuit of GDP growth has achieved for the last decade or two.
It kept getting slightly worse every year since I joined the workforce. My salary is 2x when I started 14 years ago but it buys less than half what it could back then. Most of it goes into rent. It's almost impossible to save. I was never given a salary increase even with excellent reviews of my work as a software developer. I even built a popular open source project with 6k stars on GitHub which was a lot back then but it didn't yield any good opportunities.
I only got increases by changing companies which I did at least 10 times over the last 14 years (a lot of contract jobs). I also changed countries several times. I only worked for one company which grew pretty well and got some shares as early employee but they're not liquid currently and maybe worth $150k USD tops but who knows where it will end up with AI disruption. Anyway I can't cash out.
I work non-stop, still working side projects nights and weekends. Feels like there are no opportunities at all and all media channels are full of bots so I can't promote my product. When I tried social media ads years ago, it felt like it was only shown to bots so I've been skeptical of this approach since (especially now with LLMs around). Also, the internet is full of fake products which claim to have similar features as my product but they don't actually deliver so my target customer grew distrusting of all small projects like mine, making it impossible to gain trust.
So yeah my work life sucks. I struggle to save. Can rarely go on holidays. Can't even take leave without raising eyebrows at work. Day job is more demanding, less fulfilling. I don't have much personal life because I work weekends. If I started talking about my career in detail, it just gets awful and Kafkaesque.
So it's pretty bad. I regret having hustled so hard. While others were out partying, I was working and improving myself through open source, now my open source work was scraped by AI and sold for a profit while I get zero opportunities from it and copyright laws which I based much of my career strategy around are being dismantled.
Also, there is another terrible thing that happened to me in my personal life but it's just bad luck. So I can't blame that on "the system." I appreciate improvements to medicine like another commenter also said but that's essentially the only improvement.
Why is the nurse a she?
From my point of view, the point about the nurse is especially relevant because, in the future, dealing with an aging population is going to become a major undertaking and a major burden.
Demographically, we simply don't have the number of people we need to give elders the care that they'll probably have to have near the end of their life, so we have to rely on AI to help us get through it.
Probably, we will see this happen in Japan and South Korea before it hits the United States.
That is an interesting theory. That could explain why all developed countries seems deliberately raising their young generations to be as unprepared for real life as possible. They simply fear that if the younger generation becomes capable of anything, the younger generations will simply discard democracy, which the more numerous and more weak older generation planning to use to redistribute wealth in their own favor, and the older generations will be left to drag a miserable existence.
South Korea will be interesting to watch indeed.
Hell, maybe even the lowest fertility rate of South Korea in the world and it's de-facto status of western colony is not a coincidence, but designed situation to watch in advance what challenges low birth rate will bring and how better to overcome those challenges
But she won't, will she? Her employer will do the incorporating. With very different objectives.
The model seems defensible, but for some reason it isn’t passing the sniff test. Maybe its that the extreme scenario seems wildly optimistic, or that the observed modest productivity gains are framed as either the trend or “just the start of something amazing !”.
The issues are that
1) The median AI use case isn’t increasing productivity but expanding capability and reducing the need to communicate across teams. Ex: if someone needs a dashboard or a ppt designed, now they don’t have to ask another team. AI is substituting for intern and early career workers in someone else’s firm.
This is also borne out in the hiring data.
2) The cases that show maximal productivity gains are when AI is leveraged by experts.
This is a problem, because the evidence for automation gains is poor. AI automation seems to be going to the same place ML projects went to die.
If augmentation is the primary lever that lifts up productivity, and it is dependent on AI being matched to experts, then the limiting factor is expertise.
This makes the fact that AI is both, taking out early career roles AND one shooting education, a divine irony.
Where exactly are those future experts supposed to come from? The ones with taste and the ability to deliver us this AI driven GDP growth?
AI is also setting up a conflict between liability, insurance and productivity, and thats a whole point in itself. However it will be expressed as a drag on productivity.
The researchers model says the following about hourly wage growth in the substantial and extreme scenarios:
- the hourly wage* in the knowledge workers group will fall (-1.1% and 11.5% respectively)
- but the hourly wage* of labor across the rest of the labor force will grow (5.9% and 33.6% respectively)
*Important to note, AFAIK this hourly wage decline/growth is a not measured against 2026 dollars, but the counterfactual of what hourly wages would be in the absence of AI.
Okay, so my first take away from reading these numbers is that declining wages in one group appears to be offset by rising wages in the other group. Not terrible all things considering.
But then I started to think about the difference in hourly wages between the knowledge workers group and the non-knowledge workers group (which I'll now refer to as *K* *N*). I had a sneaking suspicion that the inter group wages gap could be large enough where growth in the non-knowledge share of labor doesn't offset the loss in hourly wages from displaced knowledge workers who find themselves migrating occupations.
Looking through the paper, it breaks these groups down as follows:
K — Knowledge / Cognitive:
- Management, Business and Financial Operations, Computer and Mathematical, Architecture and Engineering, Life, Physical, and Social Science, Community and Social Service, Legal, Educational Instruction and Library, Arts, Design, Entertainment, Sports, and Media, Healthcare Practitioners and Technical, Sales and Related, Office and Administrative Support
N — Non-Knowledge / All Other:
-Healthcare Support, Protective Service, Food Preparation and Serving, Building and Grounds Cleaning and Maintenance, Personal Care and Service, Farming, Fishing, and Forestry, Construction and Extraction, Installation, Maintenance, and Repair, Production, Transportation and Material Moving
So I had GPT Sol 5.6 take each of these occupations and find the mean hourly wage using BLS data [1]. The findings suggest that as of 2026, the simple group average for K is $45.17 vs. $23.67 for N. But this is short sighted as this doesn't take into consideration the different levels of employment in each occupation. I asked ChatGPT to account for this using the same BLS data, which gives a weighted average of $41.50 for K and $23.40 for N.
This means that knowledge workers are currently making ~double what non-knowledge workers make in the US. The crucial question is how does this change when considering the wage decline/growth between K and N across the papers three scenarios?
Here's the breakdown using the simple mean for hourly wage:
Scenario Knowledge Non-knowledge Gap K -> N penalty
-------------------------------------------------------------------
Current $45.17 $23.67 $21.51 47.6%
Modest $45.35 $23.93 $21.42 47.2%
Substantial $45.03 $25.07 $19.97 44.3%
Extreme $39.98 $31.62 $ 8.35 20.9%
Here's the breakdown using the occupation weighted mean for hourly wage:
Scenario Knowledge Non-knowledge Gap K -> N penalty
-------------------------------------------------------------------
Current $41.50 $23.40 $18.10 43.6%
Modest $41.67 $23.66 $18.01 43.2%
Substantial $41.38 $24.78 $16.59 40.1%
Extreme $36.73 $31.26 $ 5.47 14.9%
Looking at these tables, both in the current moment and across all three of the papers scenario, there is a large wage gap between K and N. Outside of the extreme scenario, this gap is >40%!
My immediate takeaways having done this bit of research is that if Anthropic is right and we're on track for either the substantial or extreme scenarios, *then many of us in this thread could see a large loss in hourly wages in the short term future.*
NOTE: I'm not an economist. I push pixels for a living. But if you are an economist and you find gaps in the following analysis, please inform me.
[1] https://www.bls.gov/news.release/ocwage.t01.htm*
And the people treating this as serious economic forecasting rather than a marketing puff post are even more cringe.
If Anthropic truly believed LLMs were going to raise GDP growth to 8% permanently (or even just a few percent), instead of going public to raise capital they could just go massively leveraged long on the S&P 500 and get funding for eternity.
They are not doing that, because AI is largely only good at coding, their models barely have a 6-12 month lead over free ones, and software is only 1% of GDP.
And by making software cheaper and easier to produce, they might increase the amount of software in the world by 10X or even 100X, but it will become a smaller percentage of GDP as AI models become autonomously good at creating it. Things that are abundant and cheap don't raise GDP.
Hey Anthropic: before we worry about raising GDP growth to 20%, maybe we focus on making blog posts that don't crash your browser window if you try to resize it?
If it became clear by 2027 that GDP growth would permanently reach 15% (I didn't make that absurdly stupid chart, they did), S&P 500 valuations would immediately 100X or more.
They could take out a credit line against those gains and have unlimited money.
Wait...are you suggesting those predictions might be so unrealistic that its stupid to even publish them? I'm shocked!
Their business is not going to 10X cash flows next year, so yes, that would be the smart move if they thought even the short term modest scenario was happening.
They don't think that though. Hence why this blog post is stupid to even publish.
It's like me publishing a "What will our future look like?" blog post where I think I'm the center of the universe and draw a chart that says:
Modest scenario: I take a bowel movement so big it opens a wormhole and destroys humanity
Have you tried disabling JavaScript? The simulations won't work of course, and there are some random large blank spaces; but the main content is completely readable and handles resizing without issue[0] even on my mid-range office PC from 2014.
[0]: Just like almost anything else would without JavaScript. Though maybe the people who enjoy proving the Turing-completeness of CSS could come up with something…
Is it just me, or does their choice of identifying "workers" specifically seem a little peculiar? For me, it reads like "We're committed to making segregation beneficial for everyone, including minorities!"
Where I remain resoundingly annoyed is how useless this is. It’s pure ass-covering by Anthropic ahead of an IPO, victim-blaming the populace for the outcomes of decisions made by Anthropic leadership and the investor class. Forgive the grisly analogy, but it’s akin to a mugger telling you “only you can stop this” mid-violence: Anthropic could have chosen to allocate more of its capital to reforms that benefit the populace, but they’ve chosen to focus on personal wealth instead via this IPO and for-profit rented access schemes.
It’s nauseating. Their actions still paint a picture of greed and exploitation rather than building a better tomorrow for everyone with AI, and the fact their PR puff pieces keep getting wallpapered up here is gross.
EDIT: I guess another way to look at this is Anthropic making it unequivocally clear that they will not stop of their own accord, and thus it falls to others to stop them. Which if that’s their position, then at least they’re honest about the challenge.
The AIs are already beyond human capability in some areas, and will be beyond human in all areas within single-digit years.
I haven’t forgot capitalist euphoria from 2023 about cutting worker power off at the root. That’s all these AI Cretins want. Well, that and an unnerving enthusiasm for personifying their AI Golems (Anthr. in particular here).
Sorry to poo poo a post about AI company prompting their underlings to prompt out visualizations.