But the biggest change wasn't what it did to farming, but enabling people and societies to start doing much more than just farming, as well as enabling some great social change as well by simply economically obsoleting slave labor. And trying to imagine all of the implications of this, as well as much society might look like, from the perspective of somebody living in an agrarian society would probably have been simply impossible.
I think people keep ignoring this possibility for things that LLMs will change. There's a vast amount of the 'cognitive economy' that LLMs stand to be able to automate. And I think that will open up a vacuum in society for people to build on top of what LLMs will do (and already are doing). I don't know what that means exactly, but that's because we still live in that 'agrarian society' and trying to imagine what things will look like after the 'Industrial Revolution' is probably just impossible.
If an AI can replace me on the mental aspects of work, and robotics are on their way to replacing humanity on the physical aspects of work... then what's left? When there was agrarian societies, there were writers, priests, bankers, merchants, and laborers before and after - I really don't think things were that unclear even at the time. Now that we have machines that are close to exceeding humans in every way, what good are humans?
But lets consider what that end state could look like when machines exceed humans in everyway, what good are humans?
One possibility which I don't particularly enjoy is humans will be good for status games, art, creativity, story telling, IRL experiences, everything that involves human to human interaction and connection with other humans. If AI can supply everything in abundance than like rare cards there is only so many humans in the planet at any particular time then human's value increases. I really do believe connections with other humans will have a premium. As a nerd who is borderline anti-social, non artistic, creative, etc this does not appeal at all to me and properly to many others who read HN as well. For the majority of other people with some adaption time I think they will be fine. They were fine adapting to hunters and gathers to agricultural society, they were fine adapting to agricultural society to industrial society, they were find adapting from industrial society to information society and they will will be fine adapting from information society to the social/creative economy. Some short term pain but overall most people will accept this reality quite readily. Those born into it will not even know what we talking about.
> Some short term pain but overall most people will accept this reality quite readily. Those born into it will not even know what we talking about.
How do you propose this brave new world of yours will work economically?
Once many of those jobs get automated, I bet there will be many more people working in health research, which hopefully should lead to better health outcomes for society as a whole.
That was of course a complete fantasy. Now let's see, who is in charge of the nukes these days...
At the same time, I agree with the original comment as well. I don't think this necessarily leads to some doomsday scenario. Whatever happens it'll likely be better for us and imo we will merge with the AIs at some point, so it won't be a question of us vs them.
I actually see it more likely the opposite - a subset of humans will retreat into a (vastly smaller) human-first world, and AI will be of a separate world concerned with material/scientific/research concerns and appeases the masses who are still around.
We can already see this today in the form of upper class parents (like myself) who vastly restrict their kids screen time and ban social media. We may see AI-free societies that form, but supported by AI. The jobs and roles people play within that society may deal with services that really need to be done by a human, being a buffer from AI to the rest of this society, or just straight up fake.
I personally think this is close to a doomsday scenario - but it doesn't end with a big bang, but more of slow quiet death.
The subset of people who will want to merge will be people who will not be content with having no control over their future. No ability to create new inventions, new discoveries, new ways of working. They will not be content with just having everything provided to them, playing the social games that will remain and think their environment is a prison.
At the risk of sounding new age, my answer to this is "emotional work".
What exactly that would mean in the equivalent of the post-agrarian society that LLMs might bring, I cannot know.
There had always been abolitionists who were against slavery simply because they thought it was immoral, not because they thought slaves weren’t needed anymore.
Then what is preventing the owners of AI companies from collecting and training on tons of examples of this new work, until AI are equally good at it as humans?
The issue is that unlike the technology that automated farming or the like, AI is a general technology. So not only could it theoretically automate the work humans are currently doing, they could also automate any future human work, even if there’s some degree of lag.
Then, I think a lot of cognitive production and post-production code review/proof validation/design approval will start to look like this.
In my view that's a very charitable reading and sadly I don't think it aligns with the historical record. When the cotton gin was invented, there was a hope it would lead to a reduction in slavery. But of course it increased the demand for slaves since more cotton could now be processed, making cotton much more profitable. Slavery ended in the United States because a war was fought, not because of automation.
There’s now a chance to do more with software.
I actually think art is safe. The machines don’t value it but we do. There’s something in that.
Or will it be the cumulative total of various advances?
I've equated Claude Code, or Codex, to the looms that made fine fabric more affordable during the Industrial Revolution; life-changing, but not society-changing. Neither the steam engine nor the automobile.
Perhaps I've answered my own question in that it's the LLM technology itself that equates to the steam engine, and it will power superfarm analogs that have yet to emerge. I'm still curious what you think they will be.
So an obvious example there would be software. It's certainly true (if we assume LLMs reach their 'potential') that software will be able to reach new heights, and with a far smaller headcount driving the development. So some people see this as economically catastrophic for software developers, or an economic boon for certain large software companies.
But I think that when software can be built at the drop of a hat, software itself will no eventually no longer really matter in economic terms. Yet things you can build on top of it will matter more than ever. Those things are difficult to see from here, but I expect they will be the giants of the economy of tomorrow.
Salt is another one. Used to be payment for Roman soldier, now you can just take it from a McDonalds if you want.
Seems like we can have an abundance of software.
I hope we start to rebuild in person connections again with this technology.
Software allows us to push computation (intelligence) into our environment.
I do think there's something there though: I've spent the last 2 days building an app I've always wanted to build for myself with my computer in the corner running Claude and Claude Remote. Prototypes land on my phone and I don't even look at it for more then a few minutes before doing something else.
It's software, actually useful software, which doesn't take 100s of hours to build.
So I didn't even really spend two days on it: I mostly didn't look at all. I'll spend more time setting the result up on my home server.
Maybe a few thousand people for personal services of... various kinds. But no one's going to need the rest.
It's the ultimate capitalist fantasy.
And of course it won't happen, because long before things get to that stage AI will have independent plans of its own.
(Which is just as well, because if things did get to that stage the emperors would all wage war on each other rather than living peacefully and productively.)
I don't think we can imagine a post-ASI culture because - by definition - we're not smart or inventive enough.
It's not just farmers -> superfarms. Although in fact that did happen, but largely as a footnote to developments elsewhere.
It's more to do with the fact that our visions of the future haven't changed for over a century. They've been implemented in unexpected ways, and there have been unexpected social and cultural changes. But you can easily see the outlines of modern technology as far back as the late 19th century.
With ASI, the outcome could easily be something that doesn't look and act like technology at all. It would be some unimaginable New Thing. Literally no one on Earth has any idea what that would be or whether there would be room for trad-humans in it.
It’s not the model I don’t trust, it’s myself. The model is wrong _all the time_ because - it’s easy to verify the code - it’s hard to verify that I knew what I was talking about when I prompted it.
So the idea that you can broadly speaking take the human out of the loop. I think suggests to me a level of consistency in the contextual environment that would probably never exist.
At some point it’s politics. The model can come up with a better answer than my boss, and then my boss can just ignore it. Taking the human out of the loop broadly speaking implies that we all agree on what we’re trying to optimize.
This is, basically, 100% of the thing. We will never get to the level of automation some folks think for this exact reason.
But suppose some future holy grail AI can do much more than that.
Suppose it could find a cure for cancer, fix the climate, build fusion plants, Dyson spheres and so on.
But nobody can understand anymore how any of it works. We just ask and then trust the AI to deliver (as it always has).
Isn't it fun to imagine how life would look like in that scenario?
We would probably no longer care about code, engineering or even physics and mathematics among other things. We would probably mainly care about
The average person doesn't know how the medication they take works, the mechanics of climate and climate change, how the energy they consume is generated, etc.
But they work hard to try to understand it because the more they do the better the results for peoplea health. Also it goes without saying that if someone didn't understand many of the things that we do understand then things would be worse for us all.
The medical field as a whole isn’t generally interested in understanding how medication, only in empirical measuring and qualify the effects.
I don't think this is quite correct. I mean many practitioners of medicine will have the attitude of ... "if it works, it works". And that's perfectly reasonable.
But if you understand the mechanism of action of a drug (or other treatment), it (often) makes it easier to improve a drug.
So ... some sectors of the "medical field" understandably care only about empirical results. But other sectors would prefer to understand what's going on.
This is nonsense. Most, if not all professionals are interested in mechanism of action, but without Ms Frizzle, it is extremely difficult and expensive (time and money wise) to figure that out. So while the labs run the experiments with the very limited funding they have, we make do with using the second best thing we have, which are statistics.
That doesn't mean we understand why the medication works.
When you already know that system well, those effects are often just a matter of simple inference.
Just like here: most people are actually perfectly capable to foresee the detrimental effects of abandoning understanding.
Living in a fantasy world of "magic" makes you dependent upon your caretakers, who provide the ingredients.
Please define proof of "really understands why a medication works"
It reminds me of fynemans why do magnets work. Yeah sure does anyone really understand anything? Its metaphors all the way down
It’s not like magnets, some things really are gaps.
Now it is primarily used to treat neuropathic pain, and the mechanism for that is not well understood. The GABA receptor is not involved. This effect is just a happy accident, and nobody really understands why it works.
oh wait
They ask their employees to build stuff, and have no understanding whatsoever of how any of it works.
In their view, it is modern institutions (public and private) which, as supra-individual entities, have long since become autonomous systems. The fact that the individual office-holders are human beings, meanwhile, is of little significance.
Hannah Arendt, in her theory of totalitarianism, attributed the effectiveness of both Nazi and Stalinist policies of extermination to the largly moral indifference of bureaucracy as a system.
In this sense, the task of controlling AI is a variation on the problem of harnessing a complex society consisting mainly of autonomous subsystems. This is a problem which has increasingly challenged humanity already for quite a long time. And it has been very difficult so far, even without AI ...
We don’t generally have that insurance with LLMs/AI, yet?
It reminds me of how some religious people say that science is effectively no different to religion because we all take expert opinions on faith. But the difference is that there is a well-defined pathway to understanding, if you wish to do so.
(Don’t bother to argue this not true unless you disagree with the essence of the argument.)
Anyway, post-hoc explicability isn’t a counter-argument to the assertion that almost everyone takes almost all technology as magic, from medicine to computers.
I’m still trying to understand your argument. Are you saying that after the fact we understand AlphaGo move 37? But somehow we are never going to understand an LLM’s decision afterwards? Seems like a disconnected take to me.
The problem with "ignorance" in Western countries (particularly the US right now) is that it's very common for people who don't know to believe they know and form ignorant opinions that they often want to be applied society-wide in some way. You can see this with everything from climate change to vaccines.
In much of the world, even in middle income countries, people are comparatively poor and, in my experience living abroad in such countries for many years, much less concerned with "understanding" and forming opinions about everything under the sun. It doesn't mean they don't value education and are opposed to development/progress, but it does mean that they don't question whether the vaccine they're taking is the product of a conspiracy, think too deeply about why the river is flooding more often, etc.
They just deal with life the best they can and are more focused on supporting their families, enjoying what they can, etc.
Culture and religion play into this. The way secular and Judeo-Christian people look at the world is very different than, say, Buddhists, Muslims, Fulani tribespeople, and so on.
can’t use a computer (they’ve had like 30 years now in first world developed countries)
many can’t even use their smart phone beyond calling, texting (many can’t type well), and doom scrolling (they get addicted to drugs, gambling, and other LCD activities)
many read at a 6th grade level. most can’t even calculate tip in their head.
meanwhile, the same smartphone can give them access to literally any information and knowledge the world in seconds. and now gemini can explain stuff since most ppl can barely read or think.
it’s sad out there.
but more importantly. it’s not my problem.
But it absolutely is! Those people can and do vote.
> Isn't it fun to imagine how life would look like in that scenario?
This is horrifying to me.
People succumb to defeatism and acquiesce to regressing to zoo animals, with AI as their caretakers.
They simply cannot help but to apply the economic gauge of short term profits to value the alternatives.
Even though, obviously, here long term human survival and living conditions are at stake, necessitating an entirely different set of considerations.
It's interesting to me that you only mentioned the people using 'AI' in the 'short-term' ways, and not the ones that use it to better themselves in the 'long-term' ways. You can spend your own time focusing on either usage, it's really up to you and your concerns. Either group's sense of value is what determines their behavior. Where they spend their time and thinking must be elsewhere, and you disagree with it. Who judges the quality of time spent? You do.
Is it more useful to think about self-improvement, and how to navigate the future in ways that might help you re-establish value of yourself, life, and others? Acquiring knowledge is a struggle, the author mentioned this. There is also Plato's Allegory of the Cave, which highlights some of that struggle, a resistance to change. And we're all limited by time, our genes, our station in life.
The only way to help anyone out of the cave, is to help them believe something different about themselves. To help them believe there is good reason to spend time going deeper into knowledge, or at the very least, allow others with the passion and station for it to do so.
If not for the very least reason that it keeps us 'in the loop' of some central idea behind intelligence (prediction?). Or because we feel it keeps us safer, as a fallback measure because we acknowledge we have to trust other's knowledge to exist.
I think there's a lot of sci-fi out there that already did. Maybe it's not utopian because a pure utopia would not be likely to have an interesting story, but on the other hand, most huge technological advancements end up having just as much potential to reinforce existing power imbalances in society rather than solve them. It's not obvious to me that if we got magic super AI that can solve every scientific problem in society that it gets used in pretty much the same way as anything else: making the people who control it a lot of money rather than sharing the power with everyone without charging them.
I can almost guarantee you the first time you show cancer symptoms, you won't care whether the cure came from an AI or human's understanding. But we haven't seen that, so we can't make the judgement call.
For some people, fun is doing physics and mathematics. So they are going to keep doing that.
For even more people fun is TikTok, Snap, Instagram -> sounds like a collapse of a civilization to me if you increase the ratio even more towards dancing kids sharing their content non-stop with no added value to the society
This sounds like a boring existence. I take your meaning, but want to point out that not everyone learns about things because of practical utility, some of us find it incredibly satisfying to learn how things work just for the sake of learning.
There's also things I don't know and don't have the time to learn which are very helpful to have AI do for me: web interfaces are really useful and I look forward to them now working exactly how I want. I'm not ever going to regret not spending more time trying to figure out how to center divs or which framework I should use because they're all deprecated.
But point taken.
It’s terrifying to me to think we’d let AI make things for us we never understand. Like livestock not knowing how auto-feeders dispense their daily food were built and appeared, they just gladly eat until…
Look at the financially desolate subcultures with no option for advancement or dignified life.
That is the goal and that is how it will lool like, if the tech CEO managed to gain the power they want.
… about what?
This is such an incredibly naive and absurd vision; we've already proven that humans are very often very bad at implementing other humans' good ideas. There's nothing that AI is likely to bring that will improve this discernment.
I think you have a fundamental misunderstanding here, and it's not really explained because I think it seems self-evident from within the field. In short: writing code is a means to an end; doing mathematics research is not, but is the end in itself.
The human involvement is crucial because the entire purpose of mathematics research is to increase human understanding of mathematics. It is pursued because it is interesting, not because it is economically useful. In this sense it's a lot closer to the humanities.
A black box oracle that just tells you whether statements are true or false is not the goal of mathematics and would not be particularly interesting to the field (except insofar as it could be harnessed to improve human understanding).
Coding is totally different from this, where it is essentially always done as a means to an end. Likewise with many other fields, like pharmaceutical research or materials science or what have you, that are oriented around solving problems for some practical purpose. Pure math isn't really like that for the most part.
> A black box oracle that just tells you whether statements are true or false is not the goal of mathematics and would not be particularly interesting to the field
This is a crude distortion. The recent breakthroughs have come with proofs, reasoning and verification, and there is no proposal that I'm aware of that would do away with these foundations. There's also some rather ugly solipsism in the idea of keeping what interests the field as a limit. Mathematics has broader relevance to humanity than merely to please and support mathematicians, and if other fields can make practical use of profound well-proven future math, mathematicians will have a hard time making a case that their comprehension must come first.
> This is a crude distortion. The recent breakthroughs have come with proofs, reasoning and verification, and there is no proposal that I'm aware of that would do away with these foundations.
I believe that we are still at the point where these proofs serve as verifiable certificates of correctness, so that it's not a "trust me bro" situation, but where humans mostly still don't find them understandable, so that they are still just a highly reliable black box.
Of course, math research is cheap and most academics don’t rely upon grants, their salary covers most of their expenses. But here too, the mathematics professor spends a substantial amount of their time teaching future engineers/quants/other applied mathematicians, who need to understand math for instrumental purposes, not as an end in and of itself. Without the tuitions of these students, I can’t imagine universities maintaining the size of their math departments, let alone expanding them as Dr. Sahai advocates for.
So who or what funds the community of pure mathematics going forward?
If the goal is still eventually the applications elsewhere, we're back to what happens if the AI is simply better at this.
You can probably make an argument that human understanding is better as humans are better at finding new patterns or fundamental new ways of thinking and also applying them to new applications.
However, what if AI becomes better at humans for that as well?
No reason you couldn't have an AI be optimised for advancing basic research and understanding and a second AI to take these results and optimise for finding new applications for these discoveries.
The "if" is the problem. If it happens, then of course, let AI do it. For the moment AI is still bad at those type of tasks [1], so the discussion shouldn't focus on highly conjectural situations. We can't destroy the scientific ecosystem based on vague speculations.
[1] There are real reasons: it is not obvious how to optimize an LLM for doing basic science or other ill defined tasks. On the contrary, optimizing for writing a proof that passes the Lean test or code that passes the tests is a different story.
Thanks for providing a (much needed!) correction.
I’m not sure what to think about an analysis written by someone who didn’t catch THAT.
It's a guest post by Amit Sahai, FWIW.
If someone creates a new programming language/ framework or new better way to do async or whatever, no one will use it because it is not in the training data and it wont take off because everyone is using LLMs. It will be like using the same Lego pieces over and over.
You can do a lot of cool stuff with the same lego pieces.
Considering all music is subjectively influenced by the culture in which it's born (see the difference between Asian traditions of music, European traditions of music, African traditions, and traditions of the Americas) not even all of those have a given structure that is present today like the typical 4/4 and have polyrhythmic and multitonal structures by design. The fact that everything on the radio has converged towards 4/4 165bpm major chord progressions is evidence of that cultural phenomenon.
IMO for the moment the greatest value from these AI tools is that we can start an audit and hopefully proceed on a saner foundation, after we use the tools and think about it.
This is different than too many AI generated proofs or panic reactions from the academic system with its stupid incentives.
Thought experiment: How effective will 2026 LLMs be for humans in 2526?
It's not game over just because 500 years are missing from the training data. The important question is how well can 2526 humans make culture and knowledge navigable to LLMs via tool calls.
Today's LLMs might need for example sub agents to translate to 2526 English, sub agents to read 2526 docs.
It's _really not clear_ whether 2026 LLMs will be useless. To believe that reflects an enormous misunderstanding.
Be more specific about the "new data". If everyone is using LLMs for work (generating code), especially the juniors who won't get the chance to learn from first principles, LLMs will be training on the data they generated. How will new code enter the system at large enough quantity that it can be used for training?
> It's _really not clear_ whether 2026 LLMs will be useless. To believe that reflects an enormous misunderstanding.
They won't be useless, they will just be frozen knowing only whats in their training data. No new programming languages will emerge, in 2526 they'll still be using Rust and javascript, same exact code from 2022 which dominates the training data.
If we get a new programming language not in the training dataset, we could give an LLM a decent compiler with compile errors, and some sample code and it would be able to write code in the new language without training.
Of course some are subjective and that's where progress is harder, like "Is this website pretty?". But for tasks that can be objectively measured, LLMs will go beyond human level, just like with Chess and Go.
That's why RL is so important when training LLMs.
Chess/Go continues to progress because it is primarily a human vs human activity, people will always be learning to play chess and chess will continue to develop.
AIs are not continuing to get better at chess/go because humans continue to play at levels far below themselves who discover new techniques. They get better because they play against other AIs and discover new techniques that have a higher win rate that way.
I would bet that even if humans stopped playing chess/go and people were still willing to run these AI models against each other they would continue to get better.
Two things can be true AI drastically contribute to the advancement of chess and humans playing against each other also contribute (even if slowly) to the advancement of chess as it has always been since the invention of the game. The point is that because chess is primarily a human vs human game humans will always have the knowledge of chess, unlike with programmers who are giving it up to prompting, and programming being much more complex than chess (checkmate and win) will be stuck in 2022 because of the training data.
Computer Chess progress has nothing to do with human vs human activity. AlphaGo Zero used no human game data at all.
How much of that data can lead to innovation? Can you predict all innovation map it out on paper.
> Computer Chess progress has nothing to do with human vs human activity.
The point is that humans will always be learning chess because it primarily a human vs human activity they will be contributing games to the chess database, unlike with programmers who are stopping to code and only prompting, generating code stuck in 2022.
> AlphaGo Zero used no human game data at all.
Sure, but that instance of AlphaGo is still dependent on its training, its intelligence, so it is a question of is that the best and only way to win a game of Go. Just a few weeks ago, a Go Grandmaster found a way to beat one of the strongest Go AIs.
So a specific instance of an LLM might be the smartest based on what we know and need today but that is not the limit of how far we can go, this is why it is important for humans to always have an intimate connection with the code, math, science, chess etc for progress to continue.
If this were true then it would be impossible for these models to ever exceed the top human level as there would exist no training data that allows them to exceed the top human level.
However, despite there being no training data on ability to beat the top humans these models have achieved it.
> this is why it is important for humans to always have an intimate connection with the code, math, science, chess etc for progress to continue.
This is just you wanting to remain relevant rather than actually based on evidence.
Of course AI exceeds humans at chess, I never denied that. I am saying because chess is primarily a human vs human game, humans will always be learning and playing chess, their games will add to the chess knowledge base, AI also adds to this knowledge base. But programming is not primarily a human vs human activity so there is a risk programmers will forget how to code and all software will be stuck in 2022 because of the training data, this stifles innovation.
I guess I'm contesting that idea you are putting forward that the data from the games these humans are playing, which are at a vastly lower level that the top AIs are meaningfully important for helping the AIs to improve at the top level.
Would more people learning their times tables be helpful for top mathematicians in their fields to get better at the frontier of maths? Probably not right. Same applies here.
Would AIs advance at the same rate for the top level of chess in a world where humans completely stopped playing chess vs the world we have today. I would say they would as the human level data is of limited value to the frontier which is dominated by AI and AI game data, you are claiming that it does.
> But programming is not primarily a human vs human activity so there is a risk programmers will forget how to code and all software will be stuck in 2022 because of the training data, this stifles innovation.
Does it? Or will AI be able to run its own experiments and find better/more efficient abstractions that propagate because they are better and this will find its way into training data for future AI.
I never said human games are meaningfully important for training AI. Human games are still important for the advancement of chess, maybe Magnus Carlson can learn from games between two Super AIs but most humans still learn from games by humans, Grandmasters are continuously developing the opening, middle-game and end-game systems, adding to the chess knowledge base. Every serious chess player still reviews and study games by prominent Grandmasters, every serious chess player documents their own games, writing down every move. All rated games are recorded and added to the chess database that every player can review and study.
>Does it? Or will AI be able to run its own experiments and find better/more efficient abstractions...
Only if it is in the training data.
Are they? Why?
For a human vs human game sure but at the very top level? No of course not because it's all done by AI.
> Only if it is in the training data.
This is trivially not true, as how has AI managed to become better than humans if the knowledge of how to do so never existed in the training data.
We are well past AI can't do X unless X is in the training data. If your claim were true then AI could never surpass human expertise in any field because by definition all the available training data will at best be at the current human frontier and not beyond.
Glad that you finally agree,this is what I was saying the whole time.
> This is trivially not true, as how has AI managed to become better than humans if the knowledge of how to do so never existed in the training data.
AI can do more work, faster, AI it only needs sufficient compute and data. But that does not mean it is more intelligent than humans, it still uses the same code, algorithms, frameworks, protocols etc etc that are in the training data, sourced from human open source code on the web.
If you are a normal person research (e.g. https://arxiv.org/html/2606.22721v1 but there are a lot more, not necessarily on coding) has shown that you indeed are being less careful. It most likely also works better simply because more resources are being poured in.
For what it's worth, ten thousand terawatt fusion plants probably approaches the level at which the sheer intensity of energy production would cause significant disruption to the climate (it's roughly 5% of the Earth's entire solar input). Every energy source becomes dirty past a certain point. It would be wiser to learn how to build a utopia within a limited energy budget than find a way to produce enough of it to cook the damn planet, but who am I kidding, we're going to build a million of these things.
When we're talking about creating powerplants equal to roughly 5% of the insolation of Earth, I think we're sci-fi enough to discuss orbital datacenters or Mars datacenters or Jupiter fusion candle datacenters.
Using an entire Mars only nets us about a 2x multiplier for our energy expenditure budget. Then we need four planets to double it again. Exponential growth is a bitch.
Thermodynamics and the tyranny of exponential growth are going to win this battle every time, regardless of the unobtanium technology you try to invent.
> Using an entire Mars only nets us about a 2x multiplier for our energy expenditure budget.
We don't need to care about cooking Mars.
Greenhouse gases and Earth’s internal nuclear decay engine make the situation worse, but even without them this would boil the oceans.
You aren't really trying in good faith to think this through are you? This idea is over half a century old. Not getting it by now is willful.
The material alone is in a quantity beyond what we can reasonably manufacture.
and the material needs to be perfect. All design we have today have cascade failure mode -- any material failure translates to a total catastrophic failure.
and geostationary does not really meant Geostationary. There are lots of jiggling everywhere. It wear down over time. and let's hope nothing resonance
and we need some maintenance / decommission plan. How can we decommission this when it fail or need upgrade?
Then in the 80s, you presses 2 buttons and there you had it in your classroom without thinking twice if the electricity arrived correctly at the transistors.
How crazy will the world be once our [current gen] ANN are like that!
What an amazing thought.
Nothing is stopping LLMs to be more deterministic/correct over time.
Also you yourself is nondeterministic :)
All people are. That is how automation appeared to begin with - to provide deterministic behavior.
So I think the problem is to determine which problems under what instructions we can safely give to a model application to solve and how we test the output for safety and functionality. This would create more usable and safe, albeit a bit more boring, AI-based applications alin to a calculator or general computer. Whether this is posswith current model architecture is another thing.
Non-determinism is not an essential property of LLMs. It's an optimization that we've added intentionally.
Have you ever tried to achieve consistently deterministic output from an LLM? I have, and it's not easy.
That means output differs between machines and architectures. Running inference on CPU vs GPU also affects output. Even running the same prompt twice in a row on the same machine can lead to different outputs because a prompt that was partially stored in the kv cache will result in different output than an uncached prompt.
LLM output is very much not deterministic!
If you ran an LLM with infinite precision and guaranteed order of execution, it would be deterministic.
(I think determinism is overrated. Being deterministic does not make LLMs more reliable or correct.)
At the end of the day, an LLM is just a very big mathematical function. That is, by definition, deterministic. A particular implementation might give up on determinism for the sake of higher efficiency, but it you want a deterministic LLM, it can absolutely be done.
Put another way: if you could have a virtualization layer that guarantees deterministic floating point operations then a LLM set to 0.0 temp would produce deterministic output.
AI (IQ of Y, non deterministic) can write deterministic code.
Y is going to keep increasing, while X will not.
Will it keep up with Y? Probably not, unless people are willing to accept pretty radical interventions to their biology. But it almost certainly is not static
The increases still happen globally but mostly driven by developing countries.
How do you know?
Memory bits flip randomly. It's not a super rare thing either. You and me have experienced that many times without knowing. The only reason that computers feel deterministic is that we have error-correcting code to fix that. But in the most extreme cases, when multiple bits flip together, once "deterministic" program can generate unexpected output.
So why do you trust computers? Because statistically the case is just very unlikely. Therefore if AI is statistically unlikely to make mistakes there is no reason to not trust them.
With statistical models - such as LLM’s - there is no logic as such, but statistical assumptions based on given data. The output can ge very good or very bad, but you are fool to trust it blindly. Therefore you need a deterministic way to verify, whether meat- or software-based.
Imagine AI crushing quantum mechanics like Einstein pwned classical physics.
- Eric Hoffer
- Tech Bro
were gonna need a citation on this one.
There are so many parallels between what TV could have become and the progression of AI, a nearly free conveyance for culture, education, art and discourse. Yet, we allowed it spiral in a positive feedback loop, creating a cognitive gyre that now razes society.
At least the youngest now use "that's AI" as a pejorative, as in that is bullshit.
Because as complexity floor increases, it's "going to be correct" in hyper-specific, hyper-literal, insidious ways, with 10-50x more lines of code than necessary, and tens to hundreds of incredibly useless tests that give the illusion of quality, and cause cascading effects where seemingly irrelevant and orthogonal features that were once working end up breaking because of the agent's changes
I'm convinced this agency argument is correct [for the next N months]. But yeah, it's vibes. And you could probably create a reasonable proxy measure for this.
So I wouldn't call his argument unconvincing, I would call it unformalized. In order to walk this world you're gonna have to contend with some informal arguments that are powerful, correct, and should be convincing.
Reliable cheap fusion is the holy grail and used in moderation will fix most of our environmental and political problems, but it also forces humanity off this world. Maybe that’s not a bad thing, but there is no free lunch.
Since these resources are still extracted and allocated by humans, humans will need to be able to take apart what the AI produces, and if we want to scale this capability, we're going to need many more researchers.
Without bashing anyone in particular, a certain OS-vendor's desktop apps, have been 'good enough' to ship, but with p*ss-poor performance in many cases for the last decade or so. We crossed the 'good enough' Rubicon a few years back in terms of what end users receive as a finished app.
Hopefully LLMs will eventually bridge that last gap of efficiency when generating higher-level code that not only works, but is efficient. Maybe there's a future where they generate the final binary without even invoking a compiler.
But not the point of my comment. Computer programming has been a progression of physically wiring up valves, to soldering transistors, to punched cards, assembly, then higher level languages. Now we have natural language models.
The analogy being each that most people don't care about the assembly generated as the code works and it's really performant/efficient. Humans can still optimise assembly, but there's vanishingly small marginal gains for all but the most intensive/low-level tasks.
If LLMs produce things that work, and are indistinguishable from a careful human programmer (i.e. with some level of acceptable performance), people will simply stop looking at the high level code as the end result works, in the same way most people stopped looking at generated assembly after 8 bit computers (for example, as most games were written in raw assembly for... perforamance), as it was good enough.
> If instead I use an LLM to rewrite a feature of a codebase I can't be sure that it still functions like the original one.
Right now, with existing static analysis tooling, you can ask it to write a full suite of unit tests capturing existing behaviour without modifying the existing code with 100% code coverage, and start there. Plus fuzz tests as well. I actually have marginally more confidence in that than a human being doing it.
So extending that line of reasoning it's something like "I don't care about the internals long as the external effects pass my smell test" which is a quality/efficiency compromise.
Hopefully this million plus one mention shifts the right weights around the datacenters.
> Until very recently, I pored over every single line of code Claude generated with razor sharp scrutiny.
As LLMs generate better code in a higher level language (where better equals fewer defects, and does what you want), scrutiny of that code by humans will naturally drop. Human scrutiny will likely be replaced by something that doesn't exist yet, perhaps some sort of higher-order 'LLM linter', or Lean-esque language or tooling that somehow proves the LLM did the correct thing.
It's entirely possible in 2026, to further manually optimise compiler generated assembly, but vanishingly few people do that.
The point of my comment is that 'good enough' is almost here as demonstrated by the parent's comment.
> A fully deterministc compiler
Well, there's the rub. Humans and LLMs that asked to solve a problem at a higher level will rarely write the same code twice. Write the simplest regex, and you won't come up with this https://www.cs.princeton.edu/courses/archive/spr09/cos333/be...
The future is indeterminism.
Your “analogy” doesn’t hold up. The scrutiny applied to compilers are done by the compiler developers. Eventually if requirements don’t change the full test suite becomes the oracle. Not because of an attestation from a ghost in the machine but because of scrutiny done, let’s say over two years on a compiler that was reaching feature parity.
This obviously holds for compilers generating correct code since it is so well defined.
And this also holds for the efficiency of the generated code, since that is also obviously scrutinized by compiler developers.
Granted, the venerable LLM and the compiler do meet in a sort of functional intersection where all you can concievably care about is some thing that has a well-defined test for functionality or fitness. In the compiler’s case that’s the benchmark (good enough to not look at the assembly). But then one should go to that example directly and not to compilers in general.
I apply scrutiny to Common Lisp compilers, and have done this for more than 20 years. I'm not a compiler developer. I don't even look under the hood, at the code of the implementations.
Instead, I run massive random testing. Billions and billions of randomly generated functions, thrown at the compiler to either try to get it to crash or to generate code that produces incorrect results (detected by differential testing with different settings or transformations that should preserve what is being computed.) It's a remarkably effective way to surface compiler bugs.
As you described very well, as humans we are mostly interested in solutions, not problems. You don't have to understand how a car works to make the most of it. Increasingly, you don't have to review every line of code to feel confident it is correct. But there is inherent value in understanding the problem. The effort it takes provides a surface area for growth, perhaps the only one that is actually available to us.
The solution provider also holds the locus of control, and it is only balanced when there are other available solution providers. We certainly want some of those to be human.
Most of programming is reusing existing ideas in new shapes to solve new problems, but all the building blocks are there in the training set. Or new blocks can (easily) be derived from existing ones.
Math is different, it requires quite a bit of creativity, it's not just 'reuse all existing blocks'.
For the moment LLMs are good at discovering things that we overlooked in maths, or apply cleverly existing math blocks to make new results, but making a new theory that is really useful is out of reach for the moment in my opinion.
Curious how this ages.
Recursive self improvement, self-play and multi-agent RL could make useful new theories, eventually.
However, at the moment I consider that they stay in the 'convex hull' of their training set + a provided context, and I don't see that much research that made real improvements to the situation.
This is a guest post by Amit Sahai.
There's a lot of languages where that's nowhere near as simple as you make it sound. Floating point, decimals, etc.
Then there's other little quirks like rounding rules: https://en.wikipedia.org/wiki/Rounding
Basically, adding numbers together is exactly the sort of thing AIs still muck up spectacularly, precisely because they either fail to understand the context of the problem, or fail to ask about an assumption they make.
That you've had so many replies and no-one else has even mentioned this is in itself worrying.
Your own example proves your point is wrong.
Math problems and computer programs are two places where a model can get it's direction from the problem itself. Mathematics may well be larger than just problems.
Probably human accountability.
Hypothetically, if a system built by humans then helps humans build the next system that is then initially kicked off with "design something that may influence the lives of other humans" and we all write down that AI is really good so inductively we thought itd be really good at the next thing it builds, and then a critical error is introduced and does "insert tragedy that you personally care about" then your rage would cause you to act politically and want to ask who signed off on it. If the engineering costs outweigh the fine then yeah thats what's probably going to happen but from a human accountability standpoint thats going to suck for the unlucky ones
This isn’t a question of “what can an LLM normalize”.
That is essentially impossible, since if your pored over individual lines, your scrutiny cannot be razor sharp. There are few people who can pore over code with razor-sharp scrutiny (and different people are better at scrutinizing different aspects).
> Even back in March, you could tell it "give me a function that adds two numbers" and you could be 100% confident that it would write the correct function.
I am doubtful that this is the case. Even that supposedly-naive example is not as trivial as you might imagine, when you consider overflow, defined vs undefined behavior, and floating-point representation details. And you can't be confident like that about a human either.
And then when you stop checking it, the companies that run the service will tweak the model to benefit themselves in some way, possibly at your expense, and you will be none the wiser.
All the companies trying to get you to use AI are your adversaries. They can and will exploit your use of their systems for their own gain.
In the big scheme of things is it really that expensive to verify it if a lean proof is generated? The agent itself will likely have already verified such Lean code before calling it "done".
They will never make a logical error yet make terrible assumptions and poor long scale decisions.
Wake me up when an agent swarm can write gcc in a box sealed from the internet.
I think we are a long long looong way from AI designing 'terawatt fusion plants'.
I use "frontier" AI models daily at day_job. I can confidently say that anyone who is satisfied with the output of LLM code (enough to commit it straight off) is just an absolutely shit programmer. Sorry but I don't have any other way to put it.
The code is (with rare exceptions) atrocious on every level. It is only not atrocious if you take multiple iterations of "review and correct".
>If we find no flaws in the design whatsoever, will we be less careful about the second one? The third one? What about the ten thousandth one?
Like the saying goes, if my grandmother had wheels she would have been a bike. LLMs can't even produce quality maintainable code for a trivial web service or whatever. Why are we planning for what we will do when they can "design" 10,000 nuclear power plants without any flaw?
Do programmers use this website anymore? Me, myself, I am a DOGSHIT amateur programmer and even I can tell these things are terrible without constant revision and oversight.
> I just hope that there are more Terence Taos out there than people like me.
Just spare me. Being under external pressure to “ship code” is one thing, but being personally inclined one way or another (no external pressure) is another. And when you think being inclined like that is existentially risk (for human civ?) then, what? It’s just the way you are wired and hopes and prayers that collectively that doesn’t drive us off the cliff?
This aw shucks persona isn’t convincing. Same thing with AI Bros who are (1) making the most awesome tech that has ever existed, and (2) aw shucks hope it doesn’t kill us all in the end.