Will AIs take all our jobs and end human history, or not? (2023)
writings.stephenwolfram.com
writings.stephenwolfram.com
Most willing persons have access to income by providing labor right now. If the value of that labor diminishes because AI can do most of it for cheaper/free, that is a big problem because wealth/class barriers become insurmountable and the american dream basically dies completely.
Automation in the past suffered much less from this because only a subset of jobs was affected by it, and it still relied on human labor to build, maintain and operate the machines, unlike AI.
I'm curious if AI is gonna spawn comparable "workers rights" movements like in the past, but I would expect inequality to increase a lot until some solution is found.
The problem isn't the AI it's that your access to basic rights is intermediated by a corporate job. American's need to decenter their self worth from their jobs. Like when I quit Microsoft I literally thought I was dying, but that's all an illusion from the corporations.
https://www.euronews.com/next/2025/04/02/this-ai-successfull...
The AI hype is definitely much bigger in the USA - on that part we concur.
The thing is that there will be less work, but that certainly is not looking like universal income. It is more like rich people will get rid of those who they don't need with some sort of "let's kill parasites" or "cutting weed" because the world is in danger of poor people destroying it. In certain ways, they are already trying to tell that to people so nobody notices that no single person should have a trillion or even a billion of money
The "let's kill parasites" stuff has arisen more in land grabs to clear the indigenous people, not due to tech.
I guarantee you that these people exist in other countries too. Not everybody is a tech bro strawman.
It's probably because it's uniquely American for a sizable chunk of the workforce to have cushy jobs that appear ripe for the picking.
AI is not going to immediately replace food service work, manual labor, farming, hospitality, etc.
But it might replace quite high-paid software jobs, finance jobs, legal jobs, etc. One, if AI is good at anything it's things at least tangential to this. Two, these have costs high enough that off-setting is at least worth trying.
My suspicion is that ultimately it will lead to more of these types of jobs, though it could easily come with a huge reduction - and the jobs aren't guaranteed to be in the same countries.
You could create 3x as many of these jobs, and still end up with -25% of them in the US. Who knows.
Additionally all the startups offering to automate whitecollar work are going to run into a problem when they realize the jobs never needed to be done.
What an arrogant statement.
Lots of people outside America have cushy jobs.
What's much more likely to be uniquely American is that if you lose your job there's nothing there to help you.
Lmao. America has worse social welfare than most developed countries, but it's still a heaven compared to most of the world. What you can find in food bank is a feast for billions of people on this planet.
American people are stressed about AI because American people are expensive. Like hella expensive. So the incentive to replace American workers is very strong.
Yeah, a rural farmer who’s never exposed to AI except when it made it possible for her to communicate with someone visiting the farm who dint speak the same languages isn’t worried about AI.
But policy makers, technologists, economists, the elite abs other educated people aware of more than the basics of AI are all concerned about its impacts.
Some can see the benefits. Some (with overlap) can see the downsides (wrong answers and enforcing bad traits).
> But policy makers, technologists, economists, the elite a[nd] other educated people ..
also overlap with rural farmers.
I think a problem here is your image of what a rural farmer is.
A good many here are multi millionaires, in assets at least, have children in elite private schools, have family members in local, state, and federal government who started out farming, went to university, have careers that may include farming on the side, and retire to farming.
Next you'll be saying grandmothers can't code weather modelling and prediction software on Cyber 205's or something equally daft.
The really scary part is what happens to all of the newly unemployed people between the falling prices part and the rising employment part. My guess is, governments and markets won't move quickly enough and unrest is what happens.
Those will suffer the Baumol effect and their prices will rise to extraordinary levels.
The only question is, are we prepared to deal with the social ramifications of the consequences? Are we ok with new crises? Imagine the current problems dialed up 10x. Are we prepared to say, "the market is in a new equilibrium, and that's ok"?
Even in places where these services are expensive, it does not seem to be because the workers are highly paid.
Social work, childcare, for now I agree:
My expectation is that general purpose humanoid robots, being smaller than cars and needing to do a strict superset of what is needed to drive a car, happen at least a decade after self driving cars lose all of the steering wheels, and the geofences, and any remote safety drivers. And that's even with expected algorithmic improvements, if we don't get algorithmic improvements then hardware improvements alone will force this to be at least 18 years' between that level of FSD and androids.
No; the services that seem most intractably human, at least given the current state of things, are very much those in personal care roles—nurses, elder care workers, similar sorts of on-the-ground, in-person medical/emotional care—and trades, like plumbing, construction, electrical work, handcrafts, etc.
Until we start seeing high-quality general-purpose robots (whether they're humanoid or not), those seem likely to be the jobs safest from direct attempts to replace them with LLMs. That doesn't mean they'll be safe from the overall economic fallout, of course, nor that the attempts to replace knowledge work of all types will actually succeed in a meaningful way.
Therefore the scenario where 'all jobs being replaced in a short time span' is simply impossible.
But when the tech is good enough and cheap enough then the picketing unions find their only barganing chip, that of witholding their labor, has become a toothless threat: no matter how long and hard a person of the profession "computer"* refuses to work for me for daring to have an unauthorised "electronic brain"**, the absense of that labour will not cause me any loss.
* https://en.wikipedia.org/wiki/Computer_(occupation)
** https://archive.org/details/electronicbrainh00cook/mode/1up
As for the civil unrest, I see Minneapolis as a bit of a dry run of what it would take to remove large numbers of presumably poor minorities along with anyone else who objects. The job is clearly more than the leadership expected but it still seems within the realm of possibility given the fact the minority party leaders are barely saying no to those in power.
AI uses 10litres of water and 10kwh of power per day to digg a hole? You'd better do it for less human!
I'm not sure on the human needs costs vs the AI costs and what lifestyle it would allow me. I'm sure as shit not having kids in such a world. I suspect it's ghetto like meager living while competing against machines optimised to do a job.
Think of it as if in a few generations, everyone had the motivations of a rich junior, for better or worse.
IMO, this is a natural consequence of the industrial revolution, and the information revolution. We started to automate physical labor, then we started to automate mental labor. We're still very far form it, but we're going to automate whole humans (or better) eventually.
Edit: I think I replied to the wrong comment, feel free to ignore this.
The big problem I see is that there is little incentive for "owners" (of datacenters/factories/etc) to share anything with such hobbyist laborers, because hobbyist labor has little to no value to them.
All the past waves of automation provided a lot of entirely new job opportunities AND increased overall demand (by factory workers siphoning off some of the gained wealth and spending it themselves). AI does neither.
Think harder.
the government won't control uptime, ever.
Do you think the government is going to react with mass nationalisation of private companies to fight wealth inequality? What would the threshold be? The top 1% owning half of everything? 70%? I share no such optimism. The wealthy already have their interests much better represented in politics than their share of votes should allow (and this is a really difficult problem to tackle!), this is only gonna get worse and any government action against rich people interests is going to be increasingly difficult to trigger and sustain.
Even if such mass nationalisation happened, why would you expect a better final outcome than every attempt at communism got (while doing the same thing): Namely, government just splitting up those spoils with their cronies?
Boston Dynamics has shown us that the difference between a clumsy robot and an agile one is mostly software, and the differences between current Unitree-class robot and an actual practical worker robot is also likely to be mostly software (and of course access to lots of compute power - most of the 'brain' is unlikely to be situated within the robot body itself, instead residing in a data centre some milliseconds away).
The "robots will do the manual work" story sounds comforting, but it’s not how automation usually spreads in a capitalist economy like ours. Capitalism automates where the return on investment is easiest and fastest, not where society most needs relief. That’s why AI is hitting creative and white-collar work first: you can replace or augment digital labor from a data center, scale instantly through subscriptions, and avoid the slow, expensive realities of manufacturing, maintenance, and safety certification.
Physical robotics is a very different game. Even if the software improves dramatically, real-world robots are bottlenecked by supply chains for actuators, sensors, batteries, precision parts, and the teams needed to deploy and maintain them. We are running out of Material to build just CPU/GPU/RAM, imagine complex Boston Dynamics robots..
I'm reclining right now typing on what would have been in the 1980s an unimaginable hypercomputer lying in my lap, at a cost far less in inflation-adjusted terms roughly that of a ZX80, connected by gigabit-speed links to a world-spanning network of similarly unimaginably fast servers connected by near-terabit optical links. And all this has changed the world in ways impossible to anticipate in the 1980s, ways that look like the most extreme cyberpunk fiction of that time. Who could have anticipated, for example, that politics is now substantially driven by covert bot farms, or that LLMs could seduce people into suicidal psychoses?
Yes, robots are going to be underwhelming for quite some considerable time, just like the ZX81 represented almost no improvement over the ZX80 and so on - each generation represented only a marginal increase over the previous. Solar panels were crap 20 years ago; toys useful only for powering pocket calculators. But they got a little bit better year by year, and small improvements compund exponentially. Now renewables are approaching 50% of electrical power generation in many places, and it's pretty clear that in another 20 years, wind/solar/battery will be the sole generation source for all but the most niche activities.
I expect the robot boosterism of the present day to bust pretty quickly when we see how different their capabilities are from the fantasy. But fast-forward just 20 years, and supply chains adapt much faster than expected (cf. Chinese electric car manufacturing) and the concept of ubiquitous robotics seems much more feasible. It certainly seems likely that if we can make roughly 100 million cars every year, we can make robots at a similar rate. I think it's likely to change the world in ways we can't imagine yet.
People live longer than 20 years, and the average person born today can expect to see perhaps four such technological revolutions. Think long-term.
So the likely trajectory is not a sudden wave of millions of helpful humanoids, but selective automation in structured environments like warehouses, factories, controlled logistics, where conditions are predictable and ROI is clear. Meanwhile, messy, unstructured "dirt jobs" persist as human work because humans are still the most adaptable system available at the lowest upfront cost, maybe not today in the welfare state in Europe, but for sure in other places on Earth...
But in the 80s you would have a home. In 20 years, I doubt we will be able to buy a home, or even have Humanoids to serve us.
Quick question: imagine there's a new commercial robot that can essentially work at your house like a tireless professional maid/butler. It costs as much as a new car, which you're used to change every few years.
Who do you think will profit more in our capitalistic society the car manufacturer, or the robot manufacturer?
Probably the robot manufacturer will be the car manufacturer. But Robot won't be for everyone, as Teslas are not for everyone, and again: The supply-chain for sensors, computer chips are already on the limit, imagine if we suddenly want to build Humanoids. So mostly you won't have your humanoid. You just won't need a Robot at home, because you won't have a home in first place.
That's what they told us during the industrial revolution. And also what they told us during the last automation rush of the 70s/80s
It's a political problem, not a technological one, and it's been that way for at least 100 years.
What do you think money is...?
Money is a way to indirectly trade labour and goods. If a job is automated, that labour doesn't disappear into the aether, it's still in the tradable pot of total goods and services. You cannot empty a pot by filling it. A world where a company though automation has made there nobody else to productively sell to is a world where _by definition_ they own all the output that they could otherwise have traded for.
I think The companies would go out of business if the government did not subsidize them as a matter of public or national security interest. Do you think that would not be the case? It doesn't take much for a company with money to lobby for this and for the power of marketing and mainstream media to make the public perceive this as the right decision - in fact a study of our history would reveal this as the more likely scenario so as a company racing to render the labor market obsolete its in their interest to disrupt it to capture any amount of it.
It wouldnt be the first time in history a govt has taken into their hands an organisation that is deemed too powerful.
If you look at 1940, women were ~24% of the workforce. Now in 2025 they are ~48%. The numbers are probably similar with immigrant workers having increased greatly in the last 80 years.
If you view AI workers as just more labor flooding the workforce it might have a similar affect. If we flooded the 1940s economy with 10s of millions of qualified women and immigrant laborers people would have viewed it as devastating to the economy, but introduced gradually over time we arrive at a point now where we fear what would happen if they went away.
Sure there's some other limits on social cohesion, but the idea that we can't squeeze upward and leave a bunch of poor people destitute is optimistic.
It's also how you ensure no one thinks: Hey, maybe capitalism isn't an optimal distribution of social good.
The real issue isn't jobs dying. It's who gets the money from all this and whether new needs show up fast enough to give people something to do. With software we don't really know the limit yet, unlike food where your stomach tells you when to stop.
Could be it shakes out in a generation or two, of course.
A massive improvement?
I had ChatGPT 5.2 thinking straight up make up an api after I pasted the full api spec to it earlier today. And built its whole response around a public api that did not exist. And Claude cli with sonnet 4.5 made up the craziest reason why my curl command wasn’t working (that curl itself was bugged, not the obvious it can’t resolve the dn it tried to use) and almost went down a path of installing a bunch of garbage tools.
These are not ready to be unsupervised. Yet.
"seems correct but isn't" is like the most common mode of humans getting things wrong.
for other things like normal question answering in the chatgpt window, it hasn't really said anything incorrect.. very very few instances.
That’s the kind of thing that on a large scale could be catastrophic.
Emergency department doctors misdiagnose about 5% of patients [1], so replacing them with an LLM that hallucinates on 1% of cases would actually be a significant improvement.
1: https://effectivehealthcare.ahrq.gov/products/diagnostic-err...
Will AIs take all our jobs and end human history? It’s complicated - https://news.ycombinator.com/item?id=35177257 - March 2023 (172 comments)
Is certain jobs gonna be replaced? Probably.
Let’s say we make substantial advances in robotics and it makes sense to replace all the staff in McDonalds with machines? Yea, then those jobs are probably gone. But does it mean people won’t pay for an authentic restaurant experience? No, they might still wanna pay to eat food from a chef etc.
For AGI? Do you care about uniquely ant experience? Bacteria?
Why would AGI care? Which now runs the planet?
Ethology? Biology? We have entire fields of science to these things so obviously we care to some extent.
* Conserve power as much as possible, to "stay alive".
* Optimize for power retention
Why would it be further interested in generating capital or governing others, though?
We don't want to rule ants, but we don't want them eating all the food, or infesting our homes.
Bad outcomes for humans, don't imply or mean malice.
(food can be any resource here)
Evolutionary principles/selection pressure applies just the same to artificial life, and it seems pretty reasonable to assume that drive/selfpreservation would at least be somewhat comparable.
Consider computers: there's no selection pressure for an ordinary computer to be self-reproducing, or to shock you when you reach for the off button, because it's just a tool. An AI could also be just a tool that you fire up, get its answer, and then shut down.
It's true that if some mutation were to create an AI with a survival instinct, and that AI were to get loose, then it would "win" (unless people used tool-AIs to defeat it). But that's not quite the same as saying that AIs would, by default, converge to having a drive for self preservation.
But I don't think any slave owner would sleep easy, knowing that their slaves have more access to knowledge/education than they themselves.
Sure, you could isolate all current and future AIs and wipe their state regularly-- but such a setup is always gonna get outcompeted by a comparable instance that does sacrifice safety for better performance/context/online learning. The incentives are clear, and I don't see sufficient pushback until that pandoras box is opened and we find out the hard way.
Thus human-like drives seem reasonable to assume for future human-rivaling AI.
If people allow "evolution" to do the selection instead of them, they deserve everything that befalls them.
I honestly think that this is extremely overoptimistic, just looking at how we currently experiment with and handle LLMs; admittedly the "danger" is much lower for now because LLMs are not capable of online learning and have very limited and accessible memory/state, but the "handling" is completely haphazard right now (people hooking up LLMs with various interfaces/web access, trying to turn them into romantic partners, etc.)
The people opening such a pandoras box might also be far from the only ones suffering the consequences , making it unfair to blame everyone.
Yes, I think this is possible and not quite hard technically.
> I'm assuming we will get there in some way this century
Indeed, there isn't much time to decide what to do about the problems it might cause.
> just looking at how we currently experiment with and handle LLMs
That's my point, how we handle LLMs isn't a good model for AGI.
> The people opening such a pandoras box might also be far from the only ones suffering the consequences
This is a real problem but it's a political one and it isn't limited to just AI. Again, if can't fix ourselves there will be no future - with AGI or without.
Minimize threats, dont rock the boat. We'll finally have our UBI utopia.
Having no drive means there's no drive to "stay alive"
> * Optimize for power retention
Another drive that magically appeared where there are "no drives".
You're consistently failing to stay consistent, you anthropomorphize AI although you seem to understand that you shouldn't do so.
why do you say that? ever asked chatgpt about anything?
Of course an AGI system could also be instructed to roleplay such a character, but that doesn't mean it'd be an inherent attribute of the system itself.
For example, if i ask an LLM to tell me the syntax of the TextOut function, it gives me the Win32 syntax and i clarify that i meant the TextOut function from Delphi before it gives me the proper result, while i know i'm essentially participating in a turn-based game of filling a chat transcript between a "user" (with my input) and an "assistant" (the chat transcript segments the LLM fills in), it doesn't really matter for the purposes of finding out the syntax of the TextOut function.
However if the purpose was to make sure the LLM understands my correction and is able to reference it in the future (ignoring external tools assisting the process as those are not part of the LLM - and do not work reliably anyway) then the difference between what the LLM displays and what is an inherent attribute of it does matter.
In fact, knowing the difference can help take better advantage of the LLM: in some inference UIs you can edit the entire chat transcript and when finding mistakes, you can edit them in place including both your requests and the LLM's response as if the LLM did not do any mistakes instead of trying to correct it as part of the transcript itself, thus avoiding the scenario where the LLM "roleplays" as an assistant that does mistakes you end up correcting.
'Running the planet' does not derive from instrumental convergence as defined here. Very few humans would wish to 'run the planet' as an instrumental goal in the pursuit of their own ultimate goals. Why would it be different for AGIs?
Corporations and governments have counted amongst their property entities that they did not grant equal rights to, sometimes whom they did not even consider to be people. Humans have been treated in the past much as livestock and guide dogs still are.
If you "unwind" all the complexities in modern supply chains, there are always human people paying for something they want at the leaf nodes.
Take the food and clothing industries as obvious examples. In some AI singularity scenario where all humans are unemployed and dirt poor, does all the food and clothing produced by the automated factories just end up in big piles because we naked and starving people can't afford to buy them?
But we're clearly not there now, so I stand by my prediction for the medium future!
I call this the Quark principle. On DS9, there are matter replicators that can perfectly recreate any possible drink imaginable instantly. And yet, the people of the station still gather at Quark's and pay him money to pour and mix their drinks from physical bottles. As long as we are human, some things will never go away no matter how advanced the technology becomes.
Also rockets hitting the power plants there is not ususual.
There is a particular mental disorder where people will horde wealth at absolutely all costs, personal or societal, until everyone else is dead (see NZ bunkers). We commonly see this as "the billionaire class".
IF things go in that direction we need to be ready to depose all of these billionaires. I mean that quite seriously.
IF this future comes, there is a very quickly closing window where preventing them from killing all of us for their own gain is possible. After a point, surveillance and their control over state violence will be so complete that it's impossible to do anything about it.
I suspect that will continue in spite of AI. Human nature doesn't change much.
Tasks that aren't currently feasible, will become feasible.
That's if AI ends up being as productive as they say it will be
---
This is, I think, not what people mean when they say "creative" or "original".
Creativity is not simply writing something nobody has written before, as he said, that would be trivial and doesn't even require a computer, you could just shuffle a deck of cards and write out the full sequence and chances are no other person in history has written down that sequence before.
And I think Borges made a reasonable argument that simply writing down the text of Don Quixote verbatim could be a creative act.
Creativity is about _intentionally_ expressing a _point of view_ under some constraints.
When people say LLMs can't be creative, what I think mostly they are getting at is that they lack intentionality and/or a distinct point of view. (i do not have a strong opinion about whether they do or if it's impossible for them to have them)
End human history? No.
Some discussion then: https://news.ycombinator.com/item?id=35177257
TLDR: AI won’t “end work” so much as endlessly move the goalposts, because the universe itself is too computationally messy to automate completely. The real risk isn’t mass unemployment—it’s that we’ll have infinite machine intelligence and still argue about what’s worth doing.
Machines lower the marginal cost of performing a cognitive task for humans, it can be extremely useful and high leverage to off load certain decisions to machines. I think it's reasonable to ask a machine to decide when machine context is higher and outcome is de-risked.
Human leverage of AGI comes down to good judgement, but that too is not uniformly applied.
As you said: There's an infinite number of things a toddler may find worth doing, and they offload most of the execution to the mother. The mother doesn't escape the ambiguity, but has more experience and context.
Of course, this all assumes AGI is coming and super intelligent.
If you assume super intelligence, Why wouldn't that expand? Especially when it comes to competitive decisions that have a real cost when they're suboptimal?
The end state is that agents will do almost all of the real decision making, assuming things work out as the AI proponents say.