Make some sort of sense to me to think about it, but tbh we already pay vat on tokens and also on machinery (robots), home appliances etc. Also its something payed on possession and not use, i can use it for different things other than make profit out of it. I think it's not the correct way or at least, not the correct time to do something like that, we must see how it evolves before doing something so bold. Also taxing is a difficult matter because is different in countries, region, etc.
Already AI leading companies have problems finding how to back the enormous debt they are in. I don't think it's useful to somebody to act in a preemptive way on this topic.
First we must have the problems, only then we must solve them, like in the past five hundred years at least.
What should be done today is to make sure everyone benefit, at every latitude, and also to use it to improve security and productivity in every field. Make sure that all this debt and bubble-like aura around AI will deflate smoothly without major crisis.
The unemployment will be self-regulated after that. Like usual. Somebody will suffer in the meantime, hopefully a very small portion, okay, but nothing our society (and not only the american one) have already overcome in the past. History is beautiful only in retrospect.
But we, and our respective countries, can make this transition less painful than before, because we are much more free and connected and with better education and tools (also AI). So if something become a problem we have better instrument to adapt society to it and to protest, together, if the price to pay in term of standards of living its too high for a significant amount of people.
I think the important things to achieve in the short term are:
* access to AI must be a universal right in some degree and at a reasonable price, like access to internet is today
* country and community should push to decentralize power over models, data-centers location and invest in open models and open RD
The one thing about this article correct is human reserve, we must have human quota in every business or activity.
But to slow down it's not a good idea for me in general and in particular using taxes.
This would imply public corporations would have to disclose their quarterly token consumption. I wonder the impact of having access to that number publicly among competitors, would it encourage being the greatest spender of showing off how efficient you are with less tokens?
Cutting token consumption could be the new layoff-driven stock bump?
That transparency alone could change the market dynamics a lot.
Just hope that as an individual you wouldn't be taxed for this though...
Or maybe it's not just a craze and we'll soon look back and think "I wonder what we might've done differently if we'd been clear-eyed about what was happening."
Show me a person who truly revised their priors to put forward a novel solution to AI that isn't just "do the thing that I like and that benefits me in some way". The discourse around all this is a mix of quasi-religious beliefs (pro and con) and blatant opportunism. So what could we be doing differently, really? What's the big, good idea here?
Can you describe a method of slowing frontier AI development that you yourself would not classify as monopoly-seeking by US AI companies or power-seeking by governments and technocrats?
Limiting the development of powerful AI requires restricting free enterprise in some way, backed by the threat of force; hence, there will always be a cynical framing available.
> Do the ideas put forward by Gates strike you as worth pursuing? What are the trade-offs - how can they go wrong?
I agree. However, why aren’t we applying the same standard to AI technologies themselves? That doesn’t strike me as worth pursuing, the article discusses all the ways that is and will go wrong
I think it will be seen similar to the crypto craze to be honest. People on this very site, with an actual reputation, were telling us non stop how we were losers who were going to be left behind and stay poor. They never had to apologize for being obnoxious when that didn’t happen
Comparing AI to crypto makes zero sense. We get tremendous use from AI right now, while 2 decades after invention of crypto there are still basically zero legitimate use cases. I repeat - the utility of AI is not some hypothetical but something that happens right now at very large scale (so please don't reply with something in line "well, crypto was supposed to have lots of uses also").
I spent about $200 on openai tokens today to solve a seriously hard distributed systems problem that would have taken me probably three weeks to understand, let alone fix.
Crypto only ever allowed me to buy drugs over the internet. I'm not really buying this argument.
If you’re one of the people wondering why there’s an enormous backlash to AI, here’s your answer: “this thing we’re building is going to he fucking terrible and ruin your life. Please give us all your money to build it.”
It’s exactly the same set of questions and challenges as with nuclear power
Nuclear power is one of the most amazing magical things on the planet but nuclear weapons present a massive challenge to social coordination. So maybe we should sit as a society and try to figure out how to manage that instead of just freak out and burn all of our own faces off because we can’t handle complex conversations.
> It’s exactly the same set of questions and challenges as with nuclear power
I'm not sure "provide cheap and plentiful power that may be risky if not managed correctly" is really on the same level as "replace all human labor we possibly can"
I agree with you that ideally we would coordinate and cooperate better, but the sad problem (the prisoner's dilemma, really) is that some (many? most? all?) people will not coordinate or cooperate no matter what, especially if they have an opportunity to get ahead by not cooperating. As long as people like that exist we have to find ways to work with them or work around them
It doesn't help that people with the most wealth seem like they mostly have that sort of personality
Most people are reasonable. Most people will not screw you over if you just give them an inch. Living in society would be hell if this was not the case.[1] Living day-to-day is just fine in my experience. Because most people are not assholes.
But a lot of society is run by asshole sociopaths which (who) unfortunately lack some pretty fundamental wiring. And that creates some far-away problems, global even, that don’t really have anything to do with normal people that you meet daily. But you can notice how when these concrete problems are pointed out that someone will want to retreat back to generalities. Human nature. That’s IMO from a concerted effort to spread this talking point that asshole-caused problems are really human-caused problems. Because the asshole sociopaths (AHS) in charge are not AHS. They are not just human. And any human would do what they are doing. And in fact by going to war with X we are just stopping them from doing Y to us—
It is a coordination problem for normal, reasonable people. There’s too many key AHS personell disrupting the information flow. So those reasonable human beings have to make more of an effort.
[1] Some poorer societies (as have been told to me) have problems with vile opportunism of the sort where you help some apparently in need person get home and you get gutted and robbed by a gang of thugs. So I asked, okay so you can’t trust anyone then? No, they said. There’s a small and real chance that that could happen. But not to the point that you have to shun strangers. You can still choose to help.
No? Nuclear never had those massive valuations. Did not hyped itself this ridiculous way. Was not forced down on everyone. And the vision of nuclear power was not "make humans obsolete".
We did not had nuclearmaxxing. We did not had doomtrolling from nuclear companies.
While nuclear power is not intrinsically anti human you can't really separate the public perception of nuclear power and that of nuclear weapons. Anti nuclear sentiment is a lot easier to understand when you realize everyone who grew up under the Cold War had went through nuclear attack drills.
None of that is comparable to us having ai buttom on whatsapp, camera, email, jira, vacuum cleaner, everywhere. To workers being threatened with firing if they dont burn enough tokens.
Indeed, comparing AI button in some app to putting deadly substances in basically everything is really dumb. They are not comparable at all and one is much worse than the other.
Wasteful in terms of money and debt? Looking at current ai debt, no.
Nuckear arms race was something that could cause harm. This is intended to cause harm by design. And it is already clear it will cause harm to many people regardless of whether ai companies succeed or not.
have you just never read or seen anything about the 50s? nuclear power was absurdly hyped and there were in fact efforts to put it in damn near everything, there were plans for nuclear powered cars planes hell there were even some children's toys with radioactive components.
Sandbags against a tsunami. When has "pump the brakes on progress" ever worked out?
What we need is deep thinking on the implications and make the default way forward one where we want to go. In other words, we build a channel for the flood waters, not a wall. Creating a role for "somebody to deal with AI issues" is one of those hand-wavvy upper management moves that looks defensible but simply passes the buck to someone who structurally lacks the agency to actual move the needle.
> When has "pump the brakes on progress" ever worked out?
the only thing i can think of was when that Chinese scientist produced gene edited human babies. He was pretty much excommunicated from the scientific community and, eventually, spent some time in jail.
AI, on the other hand, has a huge profit potential and way too far out of the bag to go back in. To quote the movie Terminator, "it will never stop".
Yes. We can decide not to do bad things. Globally. Occasionally, it takes some time to figure out if something is bad. Especially when there's money to be made. But when we do, we can and have stopped doing them.
"A lot of people are skeptical of AI because it makes mistakes. But what's going to happen when it stops making mistakes??? It'll change the world!"
Honestly I suspect he's probably just cynically talking his book. All of this "AI is scary good / it'll be even scarier gooder soon" is no less advertising than slapping a cigarette logo on a racing car.
AI are improving their capacity at blazingly fast rate. Is noticing that AI psychosis? I'd say that believing that we are at the peak of LLM capabilities is some kind of psychosis due to large dose copium inhalation.
I'm sorry. I absolutely do not trust Bill Gates. His history of ruthless, cutthroat tactics in the 90s is not easily forgotten and when he proposes things like international organizations to audit everyone's AI[1] I can't help but question his motives.
I also don't understand what gives him the moral standing to propose a solution at all. He's not the fucking pope but a ruthless tech businessman with vested financial interest in the technology involved. He should stick to his lane.
[1]
> But even a country that gets its own house in order will still be exposed to risks that cross borders. This is why an international organization will need to be built in parallel.
> It will be unlike any other institution we have ever created, though it can follow the model of some existing systems. There’s an inspections regime for nuclear weapons, regulations for international aviation, and agreements that protect the ozone layer. A new global organization for AI will need elements of all three and more.
Interesting that you chose the pope as a positive control. I'd say that taken the whole history of papacy into question, there is no particular reason to trust the pope either.
> ruthless tech businessman with vested financial interest in the technology involved
Isn't that past-tense? What's his financial interest these days? He still owns a small amount of MSFT, but he doesn't exactly have control over the corporation any more. I imagine Nadella always answers his calls, but I'd bet most CEOs will make time to take a call from Bill.
IDK. Just dismissing him because of the monopoly thing in the 90s is probably premature. Yeah he was very driven to succeed back then.
Considering the amount of books he reads and that he's super smart (irrespective of what others negatively post about him), he's probably got a very good understanding of the current landscape and worth listening too.
I've tried to type this up a couple times, each time backing out because at the end of the day, I really don't give a rats fart about anything Gates has to say. I just wish he would disappear.
I mean, I'm as opensource as they come: not used windows since '98, despise microsoft and mostly everything he's done in business. But lets be fair, he moved the needle forwards, and after he retired he's done some fairly good shit around malaria and so on. I don't really begrudge him, if it wasn't him, it would have been some other wanker. We are where we are, better to just let go of the hate.
It's a fairly concise description of the challenges ahead, and hopefully some people read it and think a bit. He didn't have to write it, I'm glad he did.
If AI is really that powerful, shouldn't someone like Gates be able to end hunger and homelessness with a one-time capital expenditure? Somehow "give people jobs" is the primary problem he wants to solve, not ending poverty. It borderline seems like he's most worried about maintaining the working poor as a valuable societal institution.
Gates is worth about $100B, Elon Musk about $850B. Let's call it $1T altogether. That would be at most about $115 per human on Earth (~8.5B humans). Not sure that is sufficient to "end hunger and homelessness."
Your calculation only makes sense if every one of those 8.5B humans is suffering from hunger and homelessness. Somewhere around 1 in 1000 people are homeless (https://worldmetrics.org/global-homelessness-statistics/), and 1 in 12 or so has experienced "hunger" some time in the past year (https://www.fao.org/hunger/en/). Obviously the numbers depend on how you define "homelessness" and "hunger", but in any case this multiplies the $ that could be used to help individuals by those factors. Between $1000 (hunger) and $115,000 (homelessness) may not sound like a lot, but when your yearly income is around $1000, it can mean a lot.
He is planning to leverage AI with significant capital expenditures:
> The Gates Foundation has 19 years left of the 20 years in which it will spend its remaining $200 billion. AI will help it achieve its ambitious goals by both accelerating the discovery of vaccines and medicines for HIV, TB, malaria, and malnutrition and helping the healthcare workforce and patients know how to use those tools. The foundation’s goals include cutting the number of children who die every year in half again, as was done from 2000 to 2024. All of our work, not just health but also agriculture and education, will take full advantage of AI.
---
I think "maintaining the working poor" is a strange misread of "give people jobs". His point is that people get a ton of meaning out of their work in our current capitalist system, and regardless of whether or not we change our system in the long term, if we don't do something about jobs in the short term, there will be a civil crisis:
> In a capitalist society, employment is the way most people get the money they need to pay for the basics of life as well as being a key source of dignity and social connection.
> When a community has high unemployment, the ripple effects can be pervasive. Research suggests that in some parts of the United States, factory closures contribute to a rise in deaths from opioid overdoses. Now imagine similar pressures on both white-collar and blue-collar workers nationwide.
> Waiting until people are already displaced or underemployed will be too late. AI is a structural challenge to the way our economy is organized, and it requires thinking and action now.
I still think there is a contradiction he glossed over. In a capitalist society, jobs have a dual role that is sometimes paradox: For workers and employees, jobs are a source of income, skill, status and participation; for the (capitalist) state, a job is often a justification why the worker deserves to exist. But for the employer, jobs are a cost center and expenditure that should be minimized - that was why this tech was developed in the first place.
The state papers over that contradiction with the idea of perpetual economic growth, where the economic activity simply grows so fast that more jobs are created than are lost by efficiency improvements.
Gates is somehow expecting that "we" (whoever that is) agree just like that to keep certain jobs from being replaced by AI, even if they could. But that would go against all the forced and incentives that make up the dynamic of the current system.
Yes, I agree robot taxes and "human reserved" jobs would be difficult to maintain long term, as it would require sustained coordination on the part of governments against market forces (the physics term "metastable" is apt here).
But the alternative (to refuse to regulate) involves mass disruptions across every level of society as human labor is replaced. A worst case scenario could involve wars or waves of violent revolution + suppression akin to the 1930s.
Ending poverty is a purely mechanical problem, with very straightforward solutions if you have robots that can do any mechanical task. You don't need to give anyone a job, you just need to give them food and shelter. (Ideally also medicine, but it sounds like that part at least he's thinking about, although again, perhaps not for anyone who doesn't already have it.)
Curing diseases and making people feel fulfilled are great goals, but they're not the primary ones in my mind. They are in Gates' mind because those are the biggest challenges he has personally experienced in his life.
Making the structural changes to our economy should be trivial if AI works as well as he's supposing it will. We won't need an economy at all, not to solve the problems caused by unemployment. (And that's the thing, is unemployment is not the actual problem, it is a cause of problems, but AGI would give us easy alternatives to the fundamentally unattainable goal of full employment.)
Sound very much like a Resource Based Economy. Back when discussion of the future and futurology was small and hopeful, that was one the well sought after goals. Now cynicism has destroyed that, and we're left to deal with the ashes.
They probably didn't read it. The Parent comment suggests Bill is saying AI is wildly powerful, more precisely Parent comment is implying Bill has ulterior motives because AI isn't that powerful yet as he fear-mongers.
The actual text Bill wrote is more about highlighting upcoming problems and convincing governments to work together on potential harms coming soon.
One is the obvious one, to produce things. AI can somewhat help with that, although less than you might think because it's largely producing knowledge output rather than physical things. If AI lets one knowledge worker do the work of 10, the effect of that is to free up 9 knowledge workers to take up manual labor, it's not to make manual laborers 10x more productive.
The second purpose is to distract people from attempting to work on their own interests and instead corral them into working for others'. You see examples of this in the form of acquihires (where they obviously don't need you to do the work, they already have a competing product, but they do need to take your product off the market), and in bullshit jobs, and in think tanks and consulting firms and legal machinery. It's a form of social control and a means to maintain social order, not a requirement for actual productivity. When you don't give people a job to do, they start working for their own interests, and their own interests frequently conflict with others. And we've seen that over and over again: the Great Depression nearly led to a fascist coup in the U.S. [1] and did result in fascism in Germany, Italy, and Spain, and large communist movements in France and China. The COVID pandemic and subsequent lockdowns led to widespread protests and civil unrest in the U.S. and other countries.
Popular discourse concentrates almost exclusively on the first purpose, because the second one is icky to talk about. In reality most people are individually better off being independently wealthy and not working, but society can only absorb so many free agents with their own ideas and wills to power. (This is also why retirement and stay-at-home parenting are broadly institutionalized: it's okay for people to not work when they no longer have the energy to work, it's only dangerous when people are not working and yet have the energy to do things other than working). But the second reason actually has more explanatory power. If you assume that the true reason for employment is to take people out of the competitive will-to-power market, lots and lots of irrational moves suddenly make sense, like high-paid executives, bullshit jobs, bureaucratic incompetence, projects where billions are invested and then the whole thing is canceled, competing internal divisions, people not getting fired for doing nothing, etc. The whole point is for them to do nothing.
There's a Chesterton's Fence [2] issue at work here. Many people, particularly young people, grow up with the first explanation for work and take it at face value, and then a new technology comes around that saves all this work and they're like "Of course this is a good idea! I want to be part of it!" But if the actual purpose of work is to distract people from pursuing their own interests, this is incredibly dangerous. The actual consequence is going to be widespread unrest and quite possibly war.
>It borderline seems like he's most worried about maintaining the working poor as a valuable societal institution.
This is the foundation of the economy. AN engine needs energy to keep going and workers are that energy. AI serves to automate a lot of that energy... but this leads to a side effect.
The side effect is that the economy is a self feedback loop, as workers not only run the engine but they also create the fuel that they themselves run on. If workers aren't working and AI has replaced them then what where are companies getting money from if workers can no longer buy things from the company? This is the fundamental loop of how an economy works... money gets paid to workers, workers spend the money and the loop continues and the economy self improves... If workers are replaced the economy no longer works.
You're addressing a flaw and an imperfection in the existing infrastructure (poverty),The article is talking about the very engine that maintains society itself which in theory can completely break if AI is introduced.
I don't know if it will ACTUALLY break, the circumstances surrounding all of this are very very complex that no one can fully predict the outcome but the theory behind it is legit and much more significant then a poverty issue.
>end hunger and homelessness with a one-time capital expenditure?
A significant percentage of the world's hungry live under warlords, militias and dictatorships that would simply kill anyone trying to peacefully give their subjects food.
Unless you're proposing AI military interventionism, you cannot fix this.
Homelessness on the other hand is mostly a mental health and drug addiction problem. Fixing it, in many cases, would require violating the consent/wishes of the homeless person.
The world's biggest problems, unfortunately, cannot be solved by throwing money at them. A lot of rich individuals, NGOs, and nation states have tried.
Look, either AI is going to magically solve all our problems or it isn't. If you're in the AI maximalist camp, then you better have an explanation for how AI is going to fix warlords or it's hard for me to take your singularity seriously.
> Homelessness on the other hand is mostly a mental health and drug addiction problem
No it's not. It's not a coincidence that homelessness only skyrocketed at the same time housing price took off to the point even professionals are spending half of their salary on rent. And if you look at the statistics you'll see addiction is a widespread issue but homelessness is not — WV and NM has some of the worst overdose deaths and the least homelessness.
Please for god's sake let's address the elephant in the room for a second. It's the housing prices.
I volunteered with homeless youth in LA, and it can be even more practical than that. When you're homeless, you definitionally don't have somewhere secure to sleep. When you're sleeping, you're vulnerable, to theft or violence (either by other homeless people, or the cops). So you don't sleep much, and sleep lightly. That leaves you exhausted. All the time. You need some energy to work whatever hours you can find at your woefully inadequate part time jobs, so you start using stimulants. That works for a while, but now you can't get any sleep, and the extreme stress of your perilous existence is only exasperated by stimulant abuse. At some point, depressants are the only real option to get some rest. And, as a lovely side effect, they make things feel okay. There's a reason you don't find a lot of otherwise-fine heroin addicts, unlike, say, coke or meth. Mostly, the feeling of opiates is just contentment. The addictive potential is closely tied to how miserable your unmedicated life is. Which, you know, for someone sleeping in the backseat of a 30 year old car in sketchy part of town or worse, is pretty fucking miserable.
And that's how someone can go from scared 14yo runaway to bartering favors for drugs in like, a year. Hell, I'd go even further and say that all the assholes blaming homelessness on mental health issues have the causality backwards, too. Living under constant stress would break anybody's psyche. Not to mention the fact that for most people I was helping, their lives before they were homeless weren't exactly rosy, either.
> It's not a coincidence that homelessness only skyrocketed at the same time housing price took off to the point even professionals are spending half of their salary on rent.
But this is also highly correlated with a skyrocketing opioid epidemic. How are you disambiguating those variables? I suspect you've decided what you believe is the cause and you're working backwards from that instead of earnestly interrogating the causes.
It's likely these are both factors, and there are more contributing to homelessness. One of the largest contributors was de-institutionalisation. Turning out mentally unwell people into the streets was clearly a terrible social strategy.
> Prevalence rates were similar across rural and nonrural counties. Higher risk of homelessness was observed among young parents; black, Hispanic, and lesbian, gay, bisexual, or transgender (LGBT) youth; and those who did not complete high school.
Youth homelessness rates were similar in cheap and expensive areas. However, factors such as race, sexual orientation, and educational attainment appeared correlated.
I'm strictly talking about developed countries where we have rule of law and it's trivial to house and feed people if you have free housing and free food. I'm not suggesting AI can solve warlords, I'm not suggesting anyone should even try. But you don't need to solve drug addiction or mental health to solve homelessness if you can build apartments for $1/sq.ft. and provide food for pennies per day. You don't need to put people into make-work jobs to justify their existence.
Housing problem is actually NIMBY problem. People pretend that they want housing problem fixed, but at the same time they don't want to build any houses. Look at Austin TX vs other cities.
You are a lazy idiot. It takes 5 paragraphs of Wikipedia
to understand that everything you are saying is complete nonsense, you absolute illiterate moron
What if he's right, and AI+Robots can do everything?
We still need to provide for all the humans. We will need to come up with a new system of distributing the world's resources that doesn't depend on "jobs" and allows the world's population access to the resources they need to live.
That's going to be very difficult, in the "we have to have a war about this" sense, traditionally.
All his good things (possibly help medicine! agriculture! government!) involve a belief in ever-increasing superintelligence and so much "hope" and optimimusm and may not ever exist.
All his negative things (increase of fraud, skill atrophy, cyberattacks, removing employment, consolidation of capital and wealth) exist currently now with AI as it is.
How long will negatives outweigh the positives? And will those positives ever actually come to fruition? How anyone (who is not a sociopath and expects to be one of those pyramid tops) can see a bright future is beyond my pessimism and sense of reality as it is now.
640k ought be enough for this AI. I'm not sure takes like these are helpful. It's widely acknowledged that this is a technological shift, like electricity, like industrialization, etc. It could be (good or bad) but most likely (good AND bad). Does anyone really believe, sincerely, that AI will be the great equalizer? It's a hilarious proposition. They are working their hardest to make sure that its not imo.
"In my second one, which I started full time in 2008, I am giving back the wealth I made at Microsoft with the goal of making the world a healthier, better educated, and more equitable place."
Like inventing waterless toilets when people just need access to clean water ?
The sad story is that we don't need innovation, we need better politics.
We should not be listening to a man who drugged his wife with Epstein supplied antibiotics after getting an STD (among many, many other bad things he's done).
Does anyone else feel incredibly nostalgic for the time when Bill Gates and Warren Buffet were the richest people on earth? I remember feeling pretty good about the fact that they were trying to do so much good, and about the giving pledge. Things feel so different now.
There was a lot of PR, but the money Gates and Buffet gave away, the foundations they created, the attention they drummed up for various charities and causes is certainly not to be scoffed at.
I'm sure they could have done more with less fanfare, but you seem to be setting the bar at "fixing the world's problems", which seems unfair. They have undoubtedly changed more lives than I'll ever manage to.
>you seem to be setting the bar at "fixing the world's problems"
No, this is the bar the general public sets for billionaires. They, ignorant of the scale of the problems, believe billionaires could afford to solve the world's problems. That is a widespread belief.
>The transition to the AI era will be one of the most turbulent times in human history. Right now, we are not preparing adequately for that transition. If the world takes the right steps, AI will be a force for good and leave everyone better off.
It will be a force for good. But like all evolutionary steps the path to good will be paved with darkness.
The very existence of humanity and life itself rests on the foundation of natural selection. Humanity will be better, it will get better but before we get there... there will be a culling.
It's an unfortunate truth. We all might be culled.
It goes back to the industrial revolution and every revolution before and after. There was a point where they thought a majority of people would be working as switchboard operators and then telephone switching came along. Horses to cars, etc.
Jobs transition, new space is created, people are caught in tbe crossfire of supply and demand curves. Incan agree the bullwhip of this one is going to be aggressive.
I think it is a mistake to bet on AI as a driver of meaningful growth in a commercial capacity in the current state. When the data fed into an AI model is based on real data you can get predictability. When the AI inputs start coming from the AI outputs the wobble is introduced which results inevitably into delirium.
I think the world at large is "learning on the job" what data scientists and others in ai have known for a long, several decades long, time have already known.
> When the data fed into an AI model is based on real data you can get predictability. When the AI inputs start coming from the AI outputs the wobble is introduced which results inevitably into delirium.
This is asserted without evidence. There is absolutely no proof that AI requires humans in the loop to function or improve itself.
Compare it with Deepmind's go-playing program, AlphaGo, which mainly involved training a neural network on a large database of high-level human games. It defeated one of the top-ranked players in the world, Lee Sedol, but arguably it was drawing from human experience just like you described.
But it didn't stop there. After that, Deepmind developed AlphaGo Zero, a version that was trained exclusively through self-play, with no human feedback in the loop. That's what you would call "AI inputs coming from AI outputs" but it didn't have the result of “resulting in delirium”: instead, it became orders of magnitudes stronger than the original version (which it defeated in a 100 to 0 competition after 3 days of training).
This shows that AI can improve itself without having access to any human knowledge, and indeed transcend human performance by orders of magnitudes. There is absolutely nothing to suggest that general AI cannot improve itself the same way.
People who claim otherwise are engaging in wishful thinking; they just assert their conclusion, but have no rational arguments to back it up.
> You have no evidence that self improvement can work at generalized tasks.
I never made that claim. I just said it's way too early to rule it out: there is no logical reason why AIs will (always) need to have a human in the loop, and we don't have enough experience with LLMs to know what their true limits are.
I referenced AlphaGo not because the game of Go is exactly like every other task AI might perform in the future, but because the evolution from AlphaGo (which was trained on human games) to AlphaGo Zero (which was not) shows that at least in certain domains, it's not only possible to take the human out of the loop, it can actually make AI perform better.
I'm not claiming this will definitely be possible in every other domain, but people who state it definitely won't be, are jumping the gun.
That paper has had a pretty turbulent reception and looks pretty conclusively wrong at this point.
It used an incorrect theoretical framing that assumed that data was being replaced rather than accumulated as a result of more training (see https://arxiv.org/abs/2404.01413 which explores this). This is incorrect because this simply isn't how real-world datasets are created via synthetic data generation (which generally accumulate more data over time rather than replace their data). As a result most of the theoretical results were invalid.
Empirical evidence has also cast a considerable amount of doubt on the paper. For example, Microsoft Phi-4 was an empirical test in specifically what happens if the majority of your training data is synthetic rather than human and it turns out that Phi-4 did significantly better than previous models which relied primarily on human data.
There's some nuance to all of this in how exactly you do this, but the original claims of the paper are looking really shaky at this point.
The key word in that paper is “indiscriminate”, as in:
> We find that _indiscriminate_ use of model-generated content in training causes irreversible defects in the resulting models
If you view AI training as lossy compression of their training data, then lossily compressing the same data repeatedly will result in data degredation; this is well known from other domains (try repeatedly compressing a JPEG image, for example).
That means it's extremely important that there is some content curation in the loop. But there is no reason to believe this content curation must be done by humans, or that it must exclude all AI-generated content by default.
For example, the recent LLM-generated disproof of the Jacobian conjecture would probably be beneficial to include in the training data, despite being the result of an LLM.
Both programs had programmed in to their routines the rules of go (and for Alpha Zero the rules of chess and shogi). The programs then followed an optimization program. Absent the extremely specific "rules of this game" set by a human none of the programs work.
Amusingly, all three games are in the game theory sense trivial because they provably have a correct solution (probably a draw for chess), but the search space is so large that it is unlikely we will ever collectively invest the resources to discover the answer.
You are referring to gans which have complete understanding of the rules and an api interface to leverage interacting with the world.
While I think there could be an argument made here you are mixing apples and rocks if you believe ChatGPT is using gans.
I hate to be that person but you are showing that you are unable to differentiate between linear regressions, clustering, gradient descent, and generative adversarial networks. You are over your ski-tips and don't see itc
That seems to be the crux here. You think it will be, I (and a lot of other people) aren't sure it will. If new space for jobs are created, I am certain we'll be fine long term.
What do you think will happen if it isn't (I know you have strong evidence for and reasons to believe it will be, but just indulge the hypothetical)?
I can indulge the hypothetical and I appreciate your openness to engage in both sides.
I think as you stated, human nature and history is where I am looking at this from. People have a tendency to be unable to sit and do nothing for extended periods of time. I believe we are purpose driven but purpose is pretty loosely defined here.
Indulging the hypothetical (and arguably realistic in some circumstances) that AI replaces a tier of repeatable tasks I believe it would work from the bottom up. Tasks that are extremely repeatable would probably be gone 1:1 and it would be a sliding ratio as complexity increased where having human intervention as a backup becomes justifiable.
This would create a squeeze on the labor force which would bring up the very real and fair question of "How is it that I am unable to survive daily life when 100 - 200 years ago none of this automation / technology existed."
I think this could drive a sort of re-examining of what exists RE: a very inefficient and flawed system with a large amount of people possessing a lot of anxiety and time on their hands. It would be like the Eye of Sauron made up of those who were caught in the crossfire.
On the other side of the haves and have-nots I can imagine there being a very Harrison Bergeron style mindset of trying to popularize pacification until the next great innovation just solves itself.
On the extreme end, I just cannot imagine euthanasia, genocide, or population control schemes. It feels like we have come so far. How could it be possible we regress that far? I could imagine myself not wanting to live through any part of that. This seems like a very extreme and existential point to reach. I am trying to keep my thoughts pinned closer to biological and historical drivers.
I think we are hormone driven chemistry sets. I believe we respond to stimuli. I find it hard to believe anyone at any level could shut off that drive at such a massive scale and accurately predict the other side.
Yeah definitely not fascist and I believe that is a control tactic not necessarily an equitability thing.
I don't think we have any evidence that anything will be done fairly. In fact the idea of Capitalism in a "fair system" has been objectively broken as incumbents apply pressure on free markets to suppress competition.
I think all forms have proved they can be broken where they no longer are "playing by the rules".
I was entertaining the hypothetical. I think the biggest unknown I feel which the article gently touches on is the unknown of how deeply this is trusted to make decisions that could be life-changing.
Imagine the danger of an AI model built to scan and understand a state or national level of policy that would be beneficial to society. Not only does that create a vector of manipulation to government that could be blindly trusted but it opens a cyber attack vector from other nation states.
I could see someone coming from a place of good without realizing that these functions are somewhat opaque without ceilings/floors which can increase deviation to elicit hallucinations.
Whether it is intentionally misused or a road to hell paved with good intentions it does feel treacherous. I do think there are many many democratizing aspects it offers but how do we stop mixing the baby and the bath water when we can't tell the difference.
Well clearly. But the above comment was about jobs that didn't exist before AI that AI will lead to the creation of.
Tiktok arguably led to more datacenters being created, AI is not unique in this regard and it's silly to imply that datacenter jobs meaningfully contribute to employment on the same scale as the loss of jobs AI could potentially lead to.
This only works until general-purpose robotics gets good enough!
There are certain jobs where the human connection fundamentally commands a premium (e.g. celebrities, sports players, the relational economy), but it's hard to imagine a world where this sector employs the majority of people. It would be a precarious existence as these jobs would be the first to go in a crisis such as a war, etc.
I don't think it "works" at all. I believe the only parties that get a net benefit out of all this are basically the ownership class. They'll get some rental fees out of it. And they'll get a more direct lever to silently manipulate public sentiment. And the working class gets poorer, and more hooked on the dopamine machine.
This is potentially much more upending since it would put hundreds of millions of people out of work and concentrate all wealth at literally a handful of individuals. We could lead this towards a beneficial scenario with proper taxation, but the individuals who stand to gain will do anything in their power to prevent this.
326 comments
[ 0.24 ms ] story [ 26.0 ms ] threadYou had me here.
Already AI leading companies have problems finding how to back the enormous debt they are in. I don't think it's useful to somebody to act in a preemptive way on this topic.
First we must have the problems, only then we must solve them, like in the past five hundred years at least.
What should be done today is to make sure everyone benefit, at every latitude, and also to use it to improve security and productivity in every field. Make sure that all this debt and bubble-like aura around AI will deflate smoothly without major crisis.
The unemployment will be self-regulated after that. Like usual. Somebody will suffer in the meantime, hopefully a very small portion, okay, but nothing our society (and not only the american one) have already overcome in the past. History is beautiful only in retrospect.
But we, and our respective countries, can make this transition less painful than before, because we are much more free and connected and with better education and tools (also AI). So if something become a problem we have better instrument to adapt society to it and to protest, together, if the price to pay in term of standards of living its too high for a significant amount of people.
I think the important things to achieve in the short term are: * access to AI must be a universal right in some degree and at a reasonable price, like access to internet is today * country and community should push to decentralize power over models, data-centers location and invest in open models and open RD
The one thing about this article correct is human reserve, we must have human quota in every business or activity.
But to slow down it's not a good idea for me in general and in particular using taxes.
That transparency alone could change the market dynamics a lot.
Just hope that as an individual you wouldn't be taxed for this though...
Maybe it will be seen similarly to the dancing crazes of the Middle Ages.
Can you describe a method of slowing frontier AI development that you yourself would not classify as monopoly-seeking by US AI companies or power-seeking by governments and technocrats?
Limiting the development of powerful AI requires restricting free enterprise in some way, backed by the threat of force; hence, there will always be a cynical framing available.
I agree. However, why aren’t we applying the same standard to AI technologies themselves? That doesn’t strike me as worth pursuing, the article discusses all the ways that is and will go wrong
Crypto only ever allowed me to buy drugs over the internet. I'm not really buying this argument.
[dupe] Discussion: https://news.ycombinator.com/item?id=49448137
Nuclear power is one of the most amazing magical things on the planet but nuclear weapons present a massive challenge to social coordination. So maybe we should sit as a society and try to figure out how to manage that instead of just freak out and burn all of our own faces off because we can’t handle complex conversations.
Why is that so hard to grasp?
I'm not sure "provide cheap and plentiful power that may be risky if not managed correctly" is really on the same level as "replace all human labor we possibly can"
I agree with you that ideally we would coordinate and cooperate better, but the sad problem (the prisoner's dilemma, really) is that some (many? most? all?) people will not coordinate or cooperate no matter what, especially if they have an opportunity to get ahead by not cooperating. As long as people like that exist we have to find ways to work with them or work around them
It doesn't help that people with the most wealth seem like they mostly have that sort of personality
But a lot of society is run by asshole sociopaths which (who) unfortunately lack some pretty fundamental wiring. And that creates some far-away problems, global even, that don’t really have anything to do with normal people that you meet daily. But you can notice how when these concrete problems are pointed out that someone will want to retreat back to generalities. Human nature. That’s IMO from a concerted effort to spread this talking point that asshole-caused problems are really human-caused problems. Because the asshole sociopaths (AHS) in charge are not AHS. They are not just human. And any human would do what they are doing. And in fact by going to war with X we are just stopping them from doing Y to us—
It is a coordination problem for normal, reasonable people. There’s too many key AHS personell disrupting the information flow. So those reasonable human beings have to make more of an effort.
[1] Some poorer societies (as have been told to me) have problems with vile opportunism of the sort where you help some apparently in need person get home and you get gutted and robbed by a gang of thugs. So I asked, okay so you can’t trust anyone then? No, they said. There’s a small and real chance that that could happen. But not to the point that you have to shun strangers. You can still choose to help.
We did not had nuclearmaxxing. We did not had doomtrolling from nuclear companies.
https://en.wikipedia.org/wiki/Radioactive_quackery
While nuclear power is not intrinsically anti human you can't really separate the public perception of nuclear power and that of nuclear weapons. Anti nuclear sentiment is a lot easier to understand when you realize everyone who grew up under the Cold War had went through nuclear attack drills.
You took the goalpost, moved it to another country and now argue that ball did not moved with it.
In fact you could probably make a strong case that the US and USSR did NuclearMaxxx
It’s called the nuclear arms race/cold war and was the defining characteristic of society post WWII until 9/11
Nuckear arms race was something that could cause harm. This is intended to cause harm by design. And it is already clear it will cause harm to many people regardless of whether ai companies succeed or not.
https://en.wikipedia.org/wiki/Atomic_Age
An AI ministry, “human-reserved” jobs, and taxes on AI tokens.
What we need is deep thinking on the implications and make the default way forward one where we want to go. In other words, we build a channel for the flood waters, not a wall. Creating a role for "somebody to deal with AI issues" is one of those hand-wavvy upper management moves that looks defensible but simply passes the buck to someone who structurally lacks the agency to actual move the needle.
Globally, it is unrealistic.
the only thing i can think of was when that Chinese scientist produced gene edited human babies. He was pretty much excommunicated from the scientific community and, eventually, spent some time in jail.
AI, on the other hand, has a huge profit potential and way too far out of the bag to go back in. To quote the movie Terminator, "it will never stop".
https://en.wikipedia.org/wiki/He_Jiankui_affair
"A lot of people are skeptical of AI because it makes mistakes. But what's going to happen when it stops making mistakes??? It'll change the world!"
Honestly I suspect he's probably just cynically talking his book. All of this "AI is scary good / it'll be even scarier gooder soon" is no less advertising than slapping a cigarette logo on a racing car.
I also don't understand what gives him the moral standing to propose a solution at all. He's not the fucking pope but a ruthless tech businessman with vested financial interest in the technology involved. He should stick to his lane.
[1] > But even a country that gets its own house in order will still be exposed to risks that cross borders. This is why an international organization will need to be built in parallel.
> It will be unlike any other institution we have ever created, though it can follow the model of some existing systems. There’s an inspections regime for nuclear weapons, regulations for international aviation, and agreements that protect the ozone layer. A new global organization for AI will need elements of all three and more.
No, but he did link to the Pope in this very essay.
Pope Leo XIV, "On Safeguarding the Human Person in the Time of Artificial Intelligence":
https://www.vatican.va/content/leo-xiv/en/encyclicals/docume...
Isn't that past-tense? What's his financial interest these days? He still owns a small amount of MSFT, but he doesn't exactly have control over the corporation any more. I imagine Nadella always answers his calls, but I'd bet most CEOs will make time to take a call from Bill.
Considering the amount of books he reads and that he's super smart (irrespective of what others negatively post about him), he's probably got a very good understanding of the current landscape and worth listening too.
We must tax lines of code. Like property tax, but for code. Every year, companies (and individuals) must declare all repos. Have them appraised.
This part was a good point, so I couldn't tell if the second part is satire or not.
It's a fairly concise description of the challenges ahead, and hopefully some people read it and think a bit. He didn't have to write it, I'm glad he did.
> The Gates Foundation has 19 years left of the 20 years in which it will spend its remaining $200 billion. AI will help it achieve its ambitious goals by both accelerating the discovery of vaccines and medicines for HIV, TB, malaria, and malnutrition and helping the healthcare workforce and patients know how to use those tools. The foundation’s goals include cutting the number of children who die every year in half again, as was done from 2000 to 2024. All of our work, not just health but also agriculture and education, will take full advantage of AI.
---
I think "maintaining the working poor" is a strange misread of "give people jobs". His point is that people get a ton of meaning out of their work in our current capitalist system, and regardless of whether or not we change our system in the long term, if we don't do something about jobs in the short term, there will be a civil crisis:
> In a capitalist society, employment is the way most people get the money they need to pay for the basics of life as well as being a key source of dignity and social connection.
> When a community has high unemployment, the ripple effects can be pervasive. Research suggests that in some parts of the United States, factory closures contribute to a rise in deaths from opioid overdoses. Now imagine similar pressures on both white-collar and blue-collar workers nationwide.
> Waiting until people are already displaced or underemployed will be too late. AI is a structural challenge to the way our economy is organized, and it requires thinking and action now.
The state papers over that contradiction with the idea of perpetual economic growth, where the economic activity simply grows so fast that more jobs are created than are lost by efficiency improvements.
Gates is somehow expecting that "we" (whoever that is) agree just like that to keep certain jobs from being replaced by AI, even if they could. But that would go against all the forced and incentives that make up the dynamic of the current system.
But the alternative (to refuse to regulate) involves mass disruptions across every level of society as human labor is replaced. A worst case scenario could involve wars or waves of violent revolution + suppression akin to the 1930s.
Curing diseases and making people feel fulfilled are great goals, but they're not the primary ones in my mind. They are in Gates' mind because those are the biggest challenges he has personally experienced in his life.
Making the structural changes to our economy should be trivial if AI works as well as he's supposing it will. We won't need an economy at all, not to solve the problems caused by unemployment. (And that's the thing, is unemployment is not the actual problem, it is a cause of problems, but AGI would give us easy alternatives to the fundamentally unattainable goal of full employment.)
I wonder how the hell did you arrive at this conclusion from this text.
The actual text Bill wrote is more about highlighting upcoming problems and convincing governments to work together on potential harms coming soon.
One is the obvious one, to produce things. AI can somewhat help with that, although less than you might think because it's largely producing knowledge output rather than physical things. If AI lets one knowledge worker do the work of 10, the effect of that is to free up 9 knowledge workers to take up manual labor, it's not to make manual laborers 10x more productive.
The second purpose is to distract people from attempting to work on their own interests and instead corral them into working for others'. You see examples of this in the form of acquihires (where they obviously don't need you to do the work, they already have a competing product, but they do need to take your product off the market), and in bullshit jobs, and in think tanks and consulting firms and legal machinery. It's a form of social control and a means to maintain social order, not a requirement for actual productivity. When you don't give people a job to do, they start working for their own interests, and their own interests frequently conflict with others. And we've seen that over and over again: the Great Depression nearly led to a fascist coup in the U.S. [1] and did result in fascism in Germany, Italy, and Spain, and large communist movements in France and China. The COVID pandemic and subsequent lockdowns led to widespread protests and civil unrest in the U.S. and other countries.
Popular discourse concentrates almost exclusively on the first purpose, because the second one is icky to talk about. In reality most people are individually better off being independently wealthy and not working, but society can only absorb so many free agents with their own ideas and wills to power. (This is also why retirement and stay-at-home parenting are broadly institutionalized: it's okay for people to not work when they no longer have the energy to work, it's only dangerous when people are not working and yet have the energy to do things other than working). But the second reason actually has more explanatory power. If you assume that the true reason for employment is to take people out of the competitive will-to-power market, lots and lots of irrational moves suddenly make sense, like high-paid executives, bullshit jobs, bureaucratic incompetence, projects where billions are invested and then the whole thing is canceled, competing internal divisions, people not getting fired for doing nothing, etc. The whole point is for them to do nothing.
There's a Chesterton's Fence [2] issue at work here. Many people, particularly young people, grow up with the first explanation for work and take it at face value, and then a new technology comes around that saves all this work and they're like "Of course this is a good idea! I want to be part of it!" But if the actual purpose of work is to distract people from pursuing their own interests, this is incredibly dangerous. The actual consequence is going to be widespread unrest and quite possibly war.
[1] https://en.wikipedia.org/wiki/Business_Plot
[2] https://fs.blog/chestertons-fence/
This is the foundation of the economy. AN engine needs energy to keep going and workers are that energy. AI serves to automate a lot of that energy... but this leads to a side effect.
The side effect is that the economy is a self feedback loop, as workers not only run the engine but they also create the fuel that they themselves run on. If workers aren't working and AI has replaced them then what where are companies getting money from if workers can no longer buy things from the company? This is the fundamental loop of how an economy works... money gets paid to workers, workers spend the money and the loop continues and the economy self improves... If workers are replaced the economy no longer works.
You're addressing a flaw and an imperfection in the existing infrastructure (poverty),The article is talking about the very engine that maintains society itself which in theory can completely break if AI is introduced.
I don't know if it will ACTUALLY break, the circumstances surrounding all of this are very very complex that no one can fully predict the outcome but the theory behind it is legit and much more significant then a poverty issue.
A significant percentage of the world's hungry live under warlords, militias and dictatorships that would simply kill anyone trying to peacefully give their subjects food.
Unless you're proposing AI military interventionism, you cannot fix this.
Homelessness on the other hand is mostly a mental health and drug addiction problem. Fixing it, in many cases, would require violating the consent/wishes of the homeless person.
The world's biggest problems, unfortunately, cannot be solved by throwing money at them. A lot of rich individuals, NGOs, and nation states have tried.
No it's not. It's not a coincidence that homelessness only skyrocketed at the same time housing price took off to the point even professionals are spending half of their salary on rent. And if you look at the statistics you'll see addiction is a widespread issue but homelessness is not — WV and NM has some of the worst overdose deaths and the least homelessness.
Please for god's sake let's address the elephant in the room for a second. It's the housing prices.
And that's how someone can go from scared 14yo runaway to bartering favors for drugs in like, a year. Hell, I'd go even further and say that all the assholes blaming homelessness on mental health issues have the causality backwards, too. Living under constant stress would break anybody's psyche. Not to mention the fact that for most people I was helping, their lives before they were homeless weren't exactly rosy, either.
But this is also highly correlated with a skyrocketing opioid epidemic. How are you disambiguating those variables? I suspect you've decided what you believe is the cause and you're working backwards from that instead of earnestly interrogating the causes.
It's likely these are both factors, and there are more contributing to homelessness. One of the largest contributors was de-institutionalisation. Turning out mentally unwell people into the streets was clearly a terrible social strategy.
We have research on this, so there's no need to retread. [For example.](https://pubmed.ncbi.nlm.nih.gov/29153445/)
> Prevalence rates were similar across rural and nonrural counties. Higher risk of homelessness was observed among young parents; black, Hispanic, and lesbian, gay, bisexual, or transgender (LGBT) youth; and those who did not complete high school.
Youth homelessness rates were similar in cheap and expensive areas. However, factors such as race, sexual orientation, and educational attainment appeared correlated.
While you may never experience it, some people can’t come up with thousands of dollars every 30 days to give someone else
https://en.wikipedia.org/wiki/Homelessness_in_the_United_Sta...
Spoiled little bitch, you know nothing
What if he's right, and AI+Robots can do everything?
We still need to provide for all the humans. We will need to come up with a new system of distributing the world's resources that doesn't depend on "jobs" and allows the world's population access to the resources they need to live.
That's going to be very difficult, in the "we have to have a war about this" sense, traditionally.
All his negative things (increase of fraud, skill atrophy, cyberattacks, removing employment, consolidation of capital and wealth) exist currently now with AI as it is.
How long will negatives outweigh the positives? And will those positives ever actually come to fruition? How anyone (who is not a sociopath and expects to be one of those pyramid tops) can see a bright future is beyond my pessimism and sense of reality as it is now.
Like inventing waterless toilets when people just need access to clean water ?
The sad story is that we don't need innovation, we need better politics.
https://www.btimesonline.com/articles/176578/20260201/epstei...
It was mostly PR. Gates and Buffet were (and are) rich, but their money simply does not come close to the scale needed to fix the world's problems.
Gates is worth $110 billion. Global GDP is $126 trillion.
I'm sure they could have done more with less fanfare, but you seem to be setting the bar at "fixing the world's problems", which seems unfair. They have undoubtedly changed more lives than I'll ever manage to.
No, this is the bar the general public sets for billionaires. They, ignorant of the scale of the problems, believe billionaires could afford to solve the world's problems. That is a widespread belief.
was that before or after he caught STDs?
[1] https://youtu.be/HEMbp6Epfz8?is=D6YFfNuYQo3o0VTi
It will be a force for good. But like all evolutionary steps the path to good will be paved with darkness.
The very existence of humanity and life itself rests on the foundation of natural selection. Humanity will be better, it will get better but before we get there... there will be a culling.
It's an unfortunate truth. We all might be culled.
It goes back to the industrial revolution and every revolution before and after. There was a point where they thought a majority of people would be working as switchboard operators and then telephone switching came along. Horses to cars, etc.
Jobs transition, new space is created, people are caught in tbe crossfire of supply and demand curves. Incan agree the bullwhip of this one is going to be aggressive.
I think it is a mistake to bet on AI as a driver of meaningful growth in a commercial capacity in the current state. When the data fed into an AI model is based on real data you can get predictability. When the AI inputs start coming from the AI outputs the wobble is introduced which results inevitably into delirium.
I think the world at large is "learning on the job" what data scientists and others in ai have known for a long, several decades long, time have already known.
This is asserted without evidence. There is absolutely no proof that AI requires humans in the loop to function or improve itself.
Compare it with Deepmind's go-playing program, AlphaGo, which mainly involved training a neural network on a large database of high-level human games. It defeated one of the top-ranked players in the world, Lee Sedol, but arguably it was drawing from human experience just like you described.
But it didn't stop there. After that, Deepmind developed AlphaGo Zero, a version that was trained exclusively through self-play, with no human feedback in the loop. That's what you would call "AI inputs coming from AI outputs" but it didn't have the result of “resulting in delirium”: instead, it became orders of magnitudes stronger than the original version (which it defeated in a 100 to 0 competition after 3 days of training).
This shows that AI can improve itself without having access to any human knowledge, and indeed transcend human performance by orders of magnitudes. There is absolutely nothing to suggest that general AI cannot improve itself the same way.
People who claim otherwise are engaging in wishful thinking; they just assert their conclusion, but have no rational arguments to back it up.
I never made that claim. I just said it's way too early to rule it out: there is no logical reason why AIs will (always) need to have a human in the loop, and we don't have enough experience with LLMs to know what their true limits are.
I referenced AlphaGo not because the game of Go is exactly like every other task AI might perform in the future, but because the evolution from AlphaGo (which was trained on human games) to AlphaGo Zero (which was not) shows that at least in certain domains, it's not only possible to take the human out of the loop, it can actually make AI perform better.
I'm not claiming this will definitely be possible in every other domain, but people who state it definitely won't be, are jumping the gun.
>This is asserted without evidence. There is absolutely no proof that AI requires humans in the loop to function or improve itself.
Nope, it's actually fairly well established.
https://pmc.ncbi.nlm.nih.gov/articles/PMC11269175/
It used an incorrect theoretical framing that assumed that data was being replaced rather than accumulated as a result of more training (see https://arxiv.org/abs/2404.01413 which explores this). This is incorrect because this simply isn't how real-world datasets are created via synthetic data generation (which generally accumulate more data over time rather than replace their data). As a result most of the theoretical results were invalid.
Empirical evidence has also cast a considerable amount of doubt on the paper. For example, Microsoft Phi-4 was an empirical test in specifically what happens if the majority of your training data is synthetic rather than human and it turns out that Phi-4 did significantly better than previous models which relied primarily on human data.
There's some nuance to all of this in how exactly you do this, but the original claims of the paper are looking really shaky at this point.
> We unknowingly gave particularly high rewards for metaphors with creatures
It was the human feedback that caused the bias, not a change in training data.
> We find that _indiscriminate_ use of model-generated content in training causes irreversible defects in the resulting models
If you view AI training as lossy compression of their training data, then lossily compressing the same data repeatedly will result in data degredation; this is well known from other domains (try repeatedly compressing a JPEG image, for example).
That means it's extremely important that there is some content curation in the loop. But there is no reason to believe this content curation must be done by humans, or that it must exclude all AI-generated content by default.
For example, the recent LLM-generated disproof of the Jacobian conjecture would probably be beneficial to include in the training data, despite being the result of an LLM.
Amusingly, all three games are in the game theory sense trivial because they provably have a correct solution (probably a draw for chess), but the search space is so large that it is unlikely we will ever collectively invest the resources to discover the answer.
RLVR for LLMs works for coding and math domains which can be verified by code.
It doesn't work for teaching dextrous manipulation for robotics, for example. Learning in robotics depends on human data.
While I think there could be an argument made here you are mixing apples and rocks if you believe ChatGPT is using gans.
I hate to be that person but you are showing that you are unable to differentiate between linear regressions, clustering, gradient descent, and generative adversarial networks. You are over your ski-tips and don't see itc
That seems to be the crux here. You think it will be, I (and a lot of other people) aren't sure it will. If new space for jobs are created, I am certain we'll be fine long term.
What do you think will happen if it isn't (I know you have strong evidence for and reasons to believe it will be, but just indulge the hypothetical)?
On one hand. We’ve never had a technological revolution that didn’t yield more jobs
On the other hand, Terminator
a) How many technological revolutions do you count?
b) Of these, which directly created jobs vs in which did the job numbers eventually recover?
Before electronic computers, rooms full of workers, often women, did the math by hand. That job was decimated.
Before ATMs, people went to bank tellers.
Before mechanized looms, there were weavers. Look up where the word “Luddite” comes from.
Before industrial manufacturing, shoes were made by cobblers.
Elevators used to have human operators.
There’s never been a technological revolution that didn’t yield more jobs than it replaced
Since such a thing has never happened, I doubt anyone, even scholars of history and anthropology, know.
Capitalism probably won't work in a world of ultra abundance though.
I think as you stated, human nature and history is where I am looking at this from. People have a tendency to be unable to sit and do nothing for extended periods of time. I believe we are purpose driven but purpose is pretty loosely defined here.
Indulging the hypothetical (and arguably realistic in some circumstances) that AI replaces a tier of repeatable tasks I believe it would work from the bottom up. Tasks that are extremely repeatable would probably be gone 1:1 and it would be a sliding ratio as complexity increased where having human intervention as a backup becomes justifiable.
This would create a squeeze on the labor force which would bring up the very real and fair question of "How is it that I am unable to survive daily life when 100 - 200 years ago none of this automation / technology existed."
I think this could drive a sort of re-examining of what exists RE: a very inefficient and flawed system with a large amount of people possessing a lot of anxiety and time on their hands. It would be like the Eye of Sauron made up of those who were caught in the crossfire.
On the other side of the haves and have-nots I can imagine there being a very Harrison Bergeron style mindset of trying to popularize pacification until the next great innovation just solves itself.
On the extreme end, I just cannot imagine euthanasia, genocide, or population control schemes. It feels like we have come so far. How could it be possible we regress that far? I could imagine myself not wanting to live through any part of that. This seems like a very extreme and existential point to reach. I am trying to keep my thoughts pinned closer to biological and historical drivers.
I think we are hormone driven chemistry sets. I believe we respond to stimuli. I find it hard to believe anyone at any level could shut off that drive at such a massive scale and accurately predict the other side.
What historical evidence do we have to believe that these newfound capabilities are going to be distributed equitably?
I don't think we have any evidence that anything will be done fairly. In fact the idea of Capitalism in a "fair system" has been objectively broken as incumbents apply pressure on free markets to suppress competition.
I think all forms have proved they can be broken where they no longer are "playing by the rules".
I was entertaining the hypothetical. I think the biggest unknown I feel which the article gently touches on is the unknown of how deeply this is trusted to make decisions that could be life-changing.
Imagine the danger of an AI model built to scan and understand a state or national level of policy that would be beneficial to society. Not only does that create a vector of manipulation to government that could be blindly trusted but it opens a cyber attack vector from other nation states.
I could see someone coming from a place of good without realizing that these functions are somewhat opaque without ceilings/floors which can increase deviation to elicit hallucinations.
Whether it is intentionally misused or a road to hell paved with good intentions it does feel treacherous. I do think there are many many democratizing aspects it offers but how do we stop mixing the baby and the bath water when we can't tell the difference.
Tiktok arguably led to more datacenters being created, AI is not unique in this regard and it's silly to imply that datacenter jobs meaningfully contribute to employment on the same scale as the loss of jobs AI could potentially lead to.
There are certain jobs where the human connection fundamentally commands a premium (e.g. celebrities, sports players, the relational economy), but it's hard to imagine a world where this sector employs the majority of people. It would be a precarious existence as these jobs would be the first to go in a crisis such as a war, etc.
I think there is value in someone of Gates' stature stating the obvious, plainly.