But perhaps whether or not his stance is correct, the students needed to hear this. They (we) have to believe human brains still have value and find a way out; for otherwise there'd be no point to try anymore.
Our value isn't predicated on our utility. The simple fact that we are sentient beings, capable of joy and suffering, gives us value. This is why we continue to support and care for the elderly and the disabled - we value them regardless of any practical utility we may derive from them.
If you go through life believing that your value depends on your practical utility, then things like AI are an incredibly scary existential threat. But denial is not a healthy way to cope with this threat. The solution is to recognize the value inherent in us as humans, and to demand public policies that reflect this fact.
"He said it was unreasonable to focus on "how much energy it takes to train an AI model, relative to how much it costs a human to do one inference query."
"It takes like 20 years of life and all of the food you eat during that time before you get smart," he said. "And not only that, it took the very widespread evolution of the 100 billion people that have ever lived and learned not to get eaten by predators and learned how to figure out science and whatever, to produce you."
https://www.theregister.com/software/2026/02/23/altman-you-t...
unless he plans to freeze the training data at this point and use that for another billion years, the cost of building AI will always be more than the cost of humanity.
Wow! Well said! so shouldn't we focus on ... fixing humanity first?
Altman is closer to a storybook evil CEO than he is to an average "AI cheerleader".
Or one I prefer, though unattributed: "If the only lens through which you can view life is value in currency, that which is priceless becomes worthless."
So the opposite of a Lisp programmer then!
People will understandably ask, what is the actual value of being capable of joy and suffering?
I frame it another way. There is value in affording all beings dignity, respect, and the opportunity to thrive. The question of our individual value as a being is undignified. People can be more or less valuable to a particular effort, but there should be no question about their worth as a person. It should not be a part of how we understand people and ourselves.
It is a healthy conclusion that your value doesn't depend on your practical utility, because that will come and go and is sometimes beyond your control. Your value isn't a question at all.
The lifeless dust and rock of the moon is an simpler value proposition to quantify than the messy intrinsic value of overlapping, ever-changing life here on Earth.
> there should be no question about their worth as a person
Dignity, respect, thriving, and even human worth don't exist without joy and at least a concept of suffering.
In the moral sense, sure.
But our modern day capitalist hellscape has made it extremely clear that if you aren't capable of providing value for shareholders, your life literally has no value. That's the reason the US government keeps cutting welfare programs, why union suppression exists.
The fact of the matter is that unless you are producing value for shareholders, you don't get to participate in society and are left to starve to death. No amount of flowery language is going to feed and house the unemployed. And we are running full speed into a situation with the explicit and overt goal of cresting as many unemployed people as possible while simultaneously ensuring that there are no resources or help offered to those unemployed people.
Flowery language will cover up the starving bodies in the streets the same way a can of febreeze will cover up a landfill. This is an enormous problem and if we don't fix it, people will die. Whether or not a human has intrinsic moral value by simply existing, we require money to survive in this society. A human life may be a mystical beautiful and valuable concept, but our society has determined that if you don't have money, you literally do not deserve to live.
That's what these students are so angry about. They're being pushed into a world that refuses to employ them and which delivers a death sentence for the crime of unemployment.
Our joy and suffering depends on our ability to fulfil our basic needs like shelter and food, and in current socioeconomic environment that ability intimately depends on the ability to create wealth, which in turn depends on our ability to provide value to others.
I can see both being undermined. What public policies reflecting this do you envision?
Yes it is. If human life was inherently valuable then the concept of poverty wouldn't exist because the entity that sees it as valuable would be willing to spend resources on maintaining it.
> The solution is to recognize the value inherent in us as humans, and to demand public policies that reflect this fact.
Most social programs keep expanding until they become unsustainably expensive. You can't just make a law "everyone gets free money" and expect this to have no negative consequences.
This only proves that injustice exists. Surprise: injustice still exists.
I'm hoping that you're still young and primarily motivated by survival, which can lure you into this cold world view. I think the reality is an inversion of that old "if you're not liberal at 20 you have no heart, but if you're not conservative by 30 you have no brain" chestnut.
Hopefully once you've made it past the raw basics of survival and the feelings of a dog-eat-dog world, you can look back and realize that compassionate people helped you over and over throughout your life, maybe without you even realizing it at the time. The next step is to realize that you can extend that same compassion to others.
This is literally how it went for me.
Likewise, most of the time you don't have social programs, somebody will introduce social programs. You can't just say "no social programs" and expect this to have no positive consequences... okay this is falling apart a bit, but the point is, what makes 'not expanding UBI' so much harder than 'not introducing UBI'? If you can convince people that introducing UBI will lead to expanding UBI and that that is bad, what's stopping you from just convincing them of the latter?
Do you expect the enlightenment project to be resumed and finished in a year? Or do you imply some religion-based worldview where the value is derived from God, flesh, nature?
No sir, grim times are coming. We can only hope that the current AI wave bites more than it can swallow.
It is very easy to get lost in between them, especially when listening to a good speaker who can flitter between those meanings at will.
What is worse is that those values interact. We indeed we continue to support and care for the elderly and the disabled, but only up to a point, and there is a reasonable discussion how exactly should countries divide their limited resources between vulnerable groups, including families with young kids. In that context, the future economic and societal value of a 5 y.o. vs. a 85 y.o. inevitably creeps up.
My value definitely depends on my practical utility. Everyone is capable of joy and suffering. A chicken is capable of joy and suffering.
Tell me this: in a hypothetical situation where you are going to end up on an uninhabited island and you can choose one person to go with you, who would you peek? An elderly disabled one or a person that has necessary skills to survive? Maybe utility matters?
Otherwise you get effects like;
* Just take drugs, feel meaningless "joy" because that's what you value,
* Don't do anything less "joyful" even though it's more meaningful.
I'll admit that knowledge isn't practical, and you can't always identify when you're creating it, and a lot of people don't think in these terms and there's a lot of intuition involved, along with societal mores about caring for people which help the growth of knowledge as general rules without getting all bean-counting about it. But I think it matters that hedonism is an incoherent motivation and that creating knowledge is a far clearer one (and hedonism tends to turn into creating knowledge, anyway, if you like meaning). Hedonism, utilitarianism, same difference.
> Otherwise you get effects like;
> * Just take drugs, feel meaningless "joy" because that's what you value,
> * Don't do anything less "joyful" even though it's more meaningful.
You could make the point that teaching, and thus furthering the collective knowledge of our species, may be somewhat objectively meaningful, because you impact the trajectory of humanity. But unless you draw joy from that specific fact alone, the joy from creating knowledge is just as selfish as taking drugs to attain a state of bliss (which, again, I don't oppose either.)
Also, I'd even challenge the notion that knowledge alone, at its face value, automatically equates to a benefit for humanity. Harari has made that point far more eloquently than I in Nexus.
But anyway I agree: motivations are arbitrary. Why you even got to do a thing? Just sit and be sessile and die. (This is not a personal attack, or recommended.)
I rely heavily on an assumption that we do all have more or less the same set of values - but this might be cultural, not biological: it's hard to get inside the head of, say, Aztecs, with whatever strange non-modern values they had.
I also make an assumption about knowledge being central among those values, although it's definitely not all that, and some people will say they don't even consider it. But I think they are doing anyway, if they live in the world as we know it.
Side comment: you've made "joy" separate from "bliss" and "meaning" separate from "knowledge", and then there's some undefined "benefit for humanity" that might not be any of those things, along with the apparent value of "impacting the trajectory of humanity" - is that good, just impacting it, in any non-specific way? lol terminology.
And I also made a distinction between knowledge and meaning, which you sort of seem to imply is a universally shared value, while I seriously doubt that is the case. There are many ways to derive meaning from existence that do not involve amassing knowledge - even just passively profiting off of the knowledge of others, but taking no curiosity in that at all.
And as you pointed out, I carefully phrased impacting the trajectory of humanity to avoid implying any moral judgement. People have many reasons for wanting to leave something behind that outlasts them, which may be good or bad or anything in between.
Why are you trying to avoid morality? That seems like a good way to never find out anything important, since importance is a moral judgement.
That’s just what I mean: I oppose your notion of a universally shared motivation of deriving meaning from creating knowledge. I don’t have an alternative to offer, because I believe no such objective motivation exists. Instead, it occurs to me you project your own belief system onto humanity (or at least your society) as a whole.
Yes, our world might be shaped most dramatically by those with a desire to create knowledge, but that still doesn’t support the generalisation that humans universally consider the creation of knowledge as a way to give their life meaning.
> Why are you trying to avoid morality?
I try to avoid bringing it into the question of what is and what isn’t a valid motivation for a continued existence, because that is one of the most fundamentally subjective aspects of being a sentient creature. Who am I to make a judgement?
(Which is itself a poor rendition of Nichomachean Ethics...this conversation you two are having is ancient)
The problem here is that joy-in-itself isn't anything. Say you're a huge hedonist, and you try to maximize your pleasure. Maybe you start with some notion involving a speedboat and cocaine. Then you might ask, how can you maximize your pleasure even more? That means you have to ask why you like things. You like things for reasons, and reasons have meaning, and meaning is knowledge. So maybe your next step is to add music or something. But in doing this your activity isn't just having pleasure, it's finding things out. The more you work at maximizing pleasure, the more you're finding things out, and the less of a cliche the things you enjoy are, and pleasure-in-itself becomes less real, because it never really meant anything. The alternate path is to stick closely to the cliches, ride around coked-up on your speedboat forever, and fail to really have a good time because mechanical behavior isn't genuinely enjoyable and trying to maximise pleasure is self-defeating.
Normal people don't self-reflect, realize they like joy, and then try to maximize this with speedboats and cocaine. If they realize they like joy and want to maximize it, they do things like spend more time with their families.
Tell that to my landlord.
Historically (in the USA), capitalism was paired with charity and supporting those around you (primarily for religious reasons).
One of the greatest downsides of the welfare system is that people don't give the money to others themselves (it's instead stripped from them and doled out without their input). They don't get to experience the good feelings that come from helping another person (only negative feelings about the government taking their money).
This removes the habits of practicing selflessness and it's positive feedback loop. As a result, we get all the downsides of capitalism with a trained selfish cohort who have no charitable feelings to counterbalance things.
Top of the line AI without educated, healthy, economically-stable children means no future for the country or our society.
We collectively spend decades and decades creating a sophisticated global capitalism, huge networks and infrastructures of trade and travel, just to find ourselves in some dark forest-esque race with everyone else anyway? Is this really consistent to you? What was the point of anything in the last, like 40 years to you if we just need to act like we are still in a cold war, except this time its a war with everyone?
It's a world prestige thing, and also a competitive edge, for better or worse.
But ok, even granting that framing, if the issue is China's placement on the spectrum of "liberal", what would it take for them to be the good enough guys here?
When I say China is a bad choice because it's not as liberal as ~the west, I do imagine a reader in China thinking the opposite. I don't think they're dumb and I don't think they've been duped; they have a coherent ideology that fits their values. I just don't want it to stomp out mine.
Maybe I'm wrong and you can solve for morality or at least the meta-morality of Liberalism/pluralism where you permit various moralities to coexist. Hopefully so. Maybe the value system in China is closer to mine than I imagine and it's just operating under different constraints. But I don't want to gamble on that when winning is within reach and is a guarantee given alignment to any human values is achieved at all.
The United States is falling behind in both.
Having great tools means more impressive solutions, not fewer blacksmiths.
“You’re right - I overstepped”
Is the new “You’re absolutely right”.
I don’t know if we can qualify something that actively goes against the explicit instructions you give it as “something great”. It just sounds like Dario is building snake oil and selling it too.
If I don't tell Claude about this behavior, it ignores the script output and lies about passing tests that validate if the config files were regenerated.
So I added to my prompt instructions to observe it, and if it sees that message, double check its work and then inform me and ask what to do before proceeding.
This has had the net result of Claude either running the script with the override flag from the get go (explicitly forbidden) or it seeing the message and convincing itself that the override is warranted and running it a second time with the override flag. It's never once stopped to ask me what to do like instructed.
They'd be out of company after a week
I really wish this were true.
Dario used to at least emphasize the potential positives of AI while being worried about the negatives, but unlike Hassabis/DeepMind he has done nothing to bring about the positive part and is now just accelerating the harmful part as fast as he can. Google is an AI company, bringing us things like AlphaFold, and Anthropic (also OpenAI) are just LLM companies.
Another similarity is the relative simplicity of the underlying structure of the system. You essentially have two hammers (one small one you swing with your hand and another big one that is planted on the ground), some material, and some heat. You build the rest.
Another similarity is the resistance to automation. A skilled blacksmith is a versatile worker. You can create assembly lines to automate any one thing they might produce. The end product will not have the same quality--it will not truly be wrought iron, each piece will not be unique, there will be nothing of the aesthetic taste of the artist in it, but if you're just some bean counter who doesn't care about those things you'll be able to sell it. But if you need the optionality to produce any of those things.. automation is not your friend. And some things just cannot be automated, at least not without extreme costs or very poor results--shoeing horses comes to mind.
I really like ‘Actual Intelligence’, that’s a clever one from Woz. People need to be reminded to use their brains, they’re a brilliant product of evolution (or your favourite god’s work).
My company actually did an internal study of agent usage for coding and found it only improved productivity by 10-20%, basically on the same level as good code templates or an autocomplete.
I'm guessing though that there are other improvements in code quality and feature velocity. I've noticed personally that AI is really good at catching smaller things that are easy to miss (e.g., if you ask it to rename fooTheBars it also updates all the relevant comments or enums that you might have missed).
That's still a pretty good outcome. 20% more output across a company is huge when you think about it. Definitely not going to change the world completely though.
> No AI company is addressing the elephant in the room that you need someone experienced constantly monitoring any agentic workflows. This means that the cost savings of agents are a myth.
I mean, it depends on the agentic workflow. Like for production code, definitely. For document and claim review, you probably need a targeted sample on a daily basis but you get massive gains.
AI gets better and better at operating self-supervised, and the amount of skill needed to supervise an AI in a useful fashion only ever goes up.
Of course, it will never be enough. The goalposts will move until we run out of them.
The reason we won't let AI run rampant is because it's anti-human. Even if AI can do everything, why would we allow it to? We don't have to do that, that's very much a choice.
We're at the point where we have humans advocating not only against other humans, but against themselves. If people are going to be out here arguing for self-destruction on the order of suicide, they might as well save us all the trouble and put a gun in their mouths. At least that way, humanity might have a few less pathetic losers who put computer programs before themselves.
Maybe that's harsh, but that's how I see it.
I have seen enough human incompetence that AI incompetence is beginning to look quite favorable. AI performance, at least, goes up generation to generation. If you've ever faced minimum wage workers, you know that they are, at best, consistently the same level of awful.
But if we move us faster than the goalposts, we can outrun them.
That does not matter when discussing its practicality; or whether they will cause drivers to lose jobs.
Sounds like you're talking about AGI, not AI. AI is here today.
You'll know were making progress towards AGI when LLMs start being called LLMs again, and something new starts being called AI.
You could legitimately call a thermostat "AI". Expert systems were previously called AI. Today it's Large Language Models. Tomorrow it'll be something else.
More to the point, there has always been a cottage industry in predicting an amazing future, just around the corner. 'AGI' is just the latest incarnation.
AI will be there, but it'll transform. When I say I don't use AI (i.e. LLMs, chat interfaces, agents and "autocomplete") for coding, research and whatnot, people label me as a luddite. The fact is I know how to use them. I test them from time to time. Occasionally these tools help. More often they hinder.
"Resistance is futile, hand your brain over!" is a hype filled dystopian fatalism noting that future is inevitable. It's inevitable. You can use this correctly, and we don't got back to our senses to understand how to use this correctly and efficiently.
We are just cooking our planet right now, with heat, poisoned water and slop.
Think about how much your own writing (and programming, if you were lucky enough to start early) evolved from, say, age 12 (when a lot of smart kids start to tackle 'real' books) to age 18 (when you supposedly have a good enough education for 50% of work in most countries) to age 25.
All of that evolution is a direct result of one thing: practice! But with a magic answer box available in everyone's pocket, it'll take truly Herculean effort from a learner to actually grind through the practice instead of just cheating for an answer. I really worry how much an LLM user will actually comprehend their own code or even prose; if you've scarcely written a line of code, how can you really understand what's going on in a debugger? If you haven't done the legwork of writing essays and constructing coherent arguments and comprehending grammar, how will you ever communicate effectively?
Maybe I'm just a dinosaur and these kids will sail a whole level of abstraction above my own understanding of writing and programming, much like how my own generation preferred Python to C, and how the previous generation evolved from assembly to C/BASIC/etc. But then I come back to those missing fundamentals, that empty mental model. It's not like my English or CS teachers had me grind through essays and implementing linked lists and Djikstra's Algorithm for pure busywork. They did it because practice is the only way to truly immerse a student in a practical subject. Maybe it'll work for programming, as long as LLMs get good enough that you can always ask them to fix low-level errors for you? But it seems unlikely to work in prose. And even those generational programming jumps I mentioned (assembly to C to Python) were lossy; most kids I went to school with would be absolutely useless writing C code, and even as a bit of a dinosaur I'm pretty awful at even debugging assembly.
Like you said: you still need to learn grammar and spelling. And I suspect a whole skill tree of other fundamentals!
It still gets things wrong, I can tell as I get through problems.
But it was either that or that dreary 'Cracking the Coding Interview' book. At least I'm learning fundamentals by asking question after question and making it track the concepts I had trouble with.
That's one use. Will most people use it to learn? Probably not. But most people are ... most people.
That's reasonable, but it doesn't mean that LLMs are close to being brains.
For a start, when humans think/talk, we often think ABOUT something - whatever is swirling about in our mind, or what we are currently seeing/feeling/etc. An LLM generating tokens/words is doing so only based on it's weights and the word sequence it is currently generating ... the human parallel would be more like a rapper spitting out words based on prior words, essentially on auto-pilot, or when we get triggered into spitting out stock phrases like "have a nice day".
If you want to compare an LLM to a human brain, it's basically equivalent to our language cortex if you ripped out all the external connections and ripped out all the feedback paths that make it capable of learning.
Of course there is a lot more to our brain than just our language cortex, but that alone should make you realize there is no real comparison beyond the fact that our language generation is also going to be based on prediction, and partly auto-regressive.
If LLMs had shame, they'd surely not repeat mistakes (in the same context window) as much as they do.
People love to put a lot of meaning on what an LLM responds with when asked why it made a mistake, but it's critical to remember that the answer to that prompt is just another series of probabilistic tokens, and has no actual relation to how the error happened.
The LLM does not understand itself in any way.
A lot of human intelligence is really societal rather than individual, based on knowledge transmitted down through generations by writing (the real enabler). If you take that away then what you are left with is something more like an isolated hunter-gather tribe.
Your point about writing and social intelligence is, to me, more evidence for the "it's language that's smart, not us" hypothesis. We start off in small bands of hunter-gatherers that store their intelligence in an oral culture. Language then jumps to clay tablets, papyrus, codex books, etc. The printing press allows it to escape containment to a wider public than just a caste of priests and bureaucrats. As soon as we invent automatic calculators, we start networking them and using those to process language, albeit in a primitive way (email, the web, etc.). Recently we discovered some abstruse math that, with the assistance of a bunch of beefy video cards, can crunch centuries of human writing into a mathematical object that encodes at least some of the meaning of that writing into an even more "advanced" symbolic processing machine. There's a clear trajectory of language itself getting more and more free of the specific wetware it grew up on.
It's a falsifiable claim, in that if there is a way to train a useful LLM from scratch without any human authored input language to bootstrap it (something I've been on the lookout for but haven't seen, though admittedly I'm not an AI researcher, just some Linux nerd with a day job as an SRE), then we can disprove it.
For the religious angle, look no further than John 1:
"In the beginning was the Word, and the Word was with God, and the Word was God."
(This is admittedly less falsifiable!)
I'm sure that we will eventually build artificial brains, capable of bootstrapping communications and language for themelves (if run en-masse in a simulation where the benefit of communication would emerge). An LLM can't do this since it is by definition/construction something only capable of learning a pre-existing language.
An artificial brain, just like a wet jiggly one, is always going to be more intelligent than a one-trick pony like an LLM - a language processor, but it is notable how intelligent that one-trick pony nonetheless appears to be.
You could choose to lock it in a virtual or physical basement with printout of the Common Crawl dataset and raise it like an LLM that learns language with zero real world grounding, in which case it may feel a bit like an LLM (but smarter - able to learn, etc), or you could let it interact with the real world and learn everything, including language, that way and be a lot more grounded.
It's hard to guess how the grounded version would feel to talk to vs the CommonCrawl one - I think it would mostly come down to how far we wanted to go in making the artificial brain have all the moving parts of a human one. In an odd way the more human we tried to make it, the more alien it might feel, but not for the "uncanny valley" reason you might imagine...
The thing with an LLM that makes it feel so human is that they are designed to 100% copy humans - their output is 100% driven by the training goal of trying to exactly match the samples in the training set. As soon as we start to try to build something more brain-like then it's behavior is going to be a lot less predictable - not just auto-regressive "auto-pilot speech", but driven by it's own internal thoughts, emotions, innate traits, etc - depending on how much of our brain we tried to copy in the artificial one. I expect it would feel a lot more organic, less robotic, to talk to, but at the same time perhaps less human since unlike the LLM it's not built just to mimic human speech.
You might be redefining words here; there isn't a form of intelligence that isn't actual intelligence. It is all actual intelligence. Artificial in this context means it is something we're creating in a lab. LLMs can't avoid being artificial intelligence. The meaning of "AI" is to artificially create actual intelligence.
And if anything, average AI user is vastly overstating how good/useful it is. Papers about it pretty much always show huge gap between "productivity person thinks they are achieving" and "actual growth of productivity"
Are we already questioning the value of a human brain because of Large Language Models? Glorified search over a vector space? This is actual psychosis
I thought that was the objective of these celebrity speeches.