I think this is one of the most interesting lines as it basically directly implies that leadership thinks this won't be a winner take all market:
> Instead of our current complex capped-profit structure—which made sense when it looked like there might be one dominant AGI effort but doesn’t in a world of many great AGI companies—we are moving to a normal capital structure where everyone has stock. This is not a sale, but a change of structure to something simpler.
That is a very obvious thing for them to say though regardless of what they truly believe, because (a) it legitimizes removing the cap , making fundraising easier and (b) averts antitrust suspicions.
> "Our for-profit LLC, which has been under the nonprofit since 2019, will transition to a Public Benefit Corporation (PBC)–a purpose-driven company structure that has to consider the interests of both shareholders and the mission."
One remarkable advantage of being a "Public Benefit Corporation" is this it:
> prevent[s] shareholders from using a drop in stock value as evidence for dismissal or a lawsuit against the corporation[1]
In my view, it is their own shareholders that the directors of OpenAI are insulating themselves against.
(b) is true but no so much (a). If investors thought it would be winner take all and they thought ClosedAI would win they'd invest in ClosedAI only and starve competitors of funding.
Actually I'm thinking in a winner-takes-all universe, the right strategy would be to spread your bets on as many likely winners as possible.
That's literally the premise of venture capital. This is a scenario where we're assuming ALL our bets will go to zero, except one which will be worth trillions. In that case you should bet on everything.
It's only in the opposite scenario (where every bet pays off with varying ROI) that it makes sense to go all-in on whichever bet seems most promising.
As a deeper issue on "justification", here is something I wrote related to this in 2001 on the risks of non-profits engaging in self-dealing when they create artificial scarcity to enrich themselves:
"Consider this way of looking at the situation. A 501(c)3 non-profit creates a digital work which is potentially of great value to the public and of great value to others who would build on that product. They could put it on the internet at basically zero cost and let everyone have it effectively for free. Or instead, they could restrict access to that work to create an artificial scarcity by requiring people to pay for licenses before accessing the content or making derived works. If they do the latter and require money for access, the non-profit can perhaps create revenue to pay the employees of the non-profit. But since the staff probably participate in the decision making about such licensing (granted, under a board who may be all volunteer), isn't that latter choice still in a way really a form of "self-dealing" -- taking public property (the content) and using it for private gain? From that point of view, perhaps restricting access is not even legal?"
"Self-dealing might be clearer if the non-profit just got a grant, made the product, and then directly sold the work for a million dollars to Microsoft and put the money directly in the staff's pockets (who are also sometimes board members). Certainly if it was a piece of land being sold such a transaction might put people in jail. But because the content or software sales are small and generally to their mission's audience they are somehow deemed OK. The trademark-infringing non-profit-sheltered project I mention above is as I see it in large part just a way to convert some government supported PhD thesis work and ongoing R&D grants into ready cash for the developers. Such "spin-offs" are actually encouraged by most funders. And frankly if that group eventually sells their software to a movie company, say, for a million dollars, who will really bat an eyebrow or complain? (They already probably get most of their revenue from similar sales anyway -- but just one copy at a time.) But how is this really different from the self-dealing of just selling charitably-funded software directly to Microsoft and distributing a lump sum? Just because "art" is somehow involved, does this make everything all right? To be clear, I am not concerned that the developers get paid well for their work and based on technical accomplishments they probably deserve that (even if we do compete for funds in a way). What I am concerned about is the way that the proprietary process happens such that the public (including me) never gets full access to the results of the publicly-funded work (other than a few publications without substantial source)."
That said, charging to provide a service that costs money to supply (e.g. GPU compute) is not necessarily self-dealing. It is restricting the source code or using patents to create artificial scarcity around those services that could be seen that way.
Enlightening read, especially your last paragraph which touches on the nuance of the situation. It’s quite easy to end up on one side or the other when it comes to charity/nonprofits because the mission itself can be very motivating and galvanizing.
>"Self-dealing [...] convert some government supported PhD thesis work [...] the public (including me) never gets full access to the results of the publicly-funded work [...]
Your 2001 essay isn't a good parallel to OpenAI's situation.
OpenAI wasn't "publicly funded" i.e. with public donations or government grants.
The non-profit was started and privately funded by a small group of billionaires and other wealthy people (Elon Musk donates $44 million, Reid Hoffman, etc collectively pledging $1 billion of their own money).
They miscalculated in thinking their charity donations would be enough to recruit the PhD machine learning researchers and pay the high GPU costs to create the AI alternative to Google DeepMind, etc. Their 2015 assumptions about future AI development costs were massively underestimated and now they look like bad for trying to convert it to a for-profit enterprise. Instead of a big conversion to for-profit, they now will settle with keeping a subsidiary that's for-profit. Somewhat like other entities structured as a non-profit that owns for-profit subsidiaries such as Mozilla, Girl Scouts, Novo Nordisk, etc.
Obviously with hindsight... if they had to do it all over, they would just create the reverse structure of creating the OpenAI for-profit company as the "parent entity" that pledges to donate money to charities. E.g. Amazon Inc is the for-profit that donates to Housing Equity Fund for affordable housing.
>uncollected tax revenues for economically valuable activity.
Taxes are on profits not revenue. The for-profit OpenAI LLC subsidiary created in 2019 would have been the entity that owes taxes but it has been losing money and never made any profits to tax.
Yesterday's news about switching from for-profit LLC to for-profit PBC still leaves a business entity that's liable for future taxes on profits.
The value investor Mohnish Pabrai once talked about his observation that most companies with a moat pretend they don’t have one and companies without pretend they do.
A version of this is emphasized in the thielverse as well. Companies in heavy competition try to intersect all their qualities to appear unique. Dominant companies talk about their portfolio of side projects to appear in heavy competition (space flight, ed tech, etc).
Mohnish isn't a tech bro though, in my books. After selling his company, guy retreated away from the tech scene to get into Buffett-style value investing. And if you read his book, it's about glorifying the small businessmen running motels and garages, who invest bit by bit into the stock market.
There needs to be regulations about deceptive, indirect, purposefully ambiguous or vague public communication by corporations (or any entity). I'm not an expert in corporate law or finance, but the statement should be:
"Open AI for-profit LLC will become a Public Benefit Corporation (PBC)"
followed by: "Profit cap is hereby removed" and finally "The Open AI non-profit will continue to control the PBC. We intend it to be a significant shareholder of the PBC."
AGI can't really be a winner take all market. The 'reward' for general intelligence is infinite as a monopoly and it accelerates productivity.
Not only is there infinite incentive to compete, but theres decreasing costs to. The only world in which AGI is winner take all is a world in which it is extremely controlled to the point at which the public cant query it.
> AGI can't really be a winner take all market. The 'reward' for general intelligence is infinite as a monopoly and it accelerates productivity
The first-mover advantages of an AGI that can improve itself are theoretically unsurmountable.
But OpenAI doesn't have a path to AGI any more than anyone else. (It's increasingly clear LLMs alone don't make the cut.) And the market for LLMs, non-general AI, is very much not winner takes all. In this announcement, OpenAI is basically acknowledging that it's not getting to self-improving AGI.
> this has some baked assumptions about cycle time and improvement per cycle and whether there's a ceiling
To be precise, it assumes a low variability in cycle time and improvement per cycle. If everyone is subjected to the same limits, the first-mover advantage remains insurmountable. I’d also argue that whether there is a ceiling matters less than how high it is. If the first AGI won’t hit a ceiling for decades, it will have decades of fratricidal supremacy.
I think the foundation model companies are actually poorly situated to reach the leading edge of AGI first, simply because their efforts are fragmented across multiple companies with different specializations—Claude is best at coding, OpenAI at reasoning, Gemini at large context, and so on.
The most advanced tools are (and will continue to be) at a higher level of the stack, combining the leading models for different purposes to achieve results that no single provider can match using only their own models.
I see no reason to think this won't hold post-AGI (if that happens). AGI doesn't mean capabilities are uniform.
I find these assumptions curious. How so? What is the AGI going to do that captures markets? Even if it can take over all desk work, then what? Who is going to consume that? And further more (and perhaps more importantly), with it putting everyone out of work, who is going to pay for it?
I'm pretty sure today's models probably can be capable of self-improving. It's just that they are not yet as good as self-improving as the combinations of programmers improving them with the help of the models.
Agreed and, if anything, you are too generous. They aren’t just not “close”, they aren’t even working in the same category as anything that might be construed as independently intelligent.
I agree with you, but that’s kindof beside the point. Open AI’s thesis is that they will work towards AGI, and eventually succeed. In the context of that premise, Open AI still doesn’t believe AGI would be winner-takes-all. I think that’s an interesting discussion whether you believe the premise or not.
Differentiating between AGI and non-AGI, if we ever get remotely close, would be challenging, but for now it's trivial. The defining feature of AGI is recursive self improvement across any field. Without self improvement, you're just regurgitating. Humanity started with no advanced knowledge or even a language. In what should practically be a heartbeat at the speed of distributed computing with perfect memory and computation power, we were landing a man on the Moon.
So one fundamental difference is that AGI would not need some absurdly massive data dump to become intelligent. In fact you would prefer to feed it as minimal a series of the most primitive first principles as possible because it's certain that much of what we think is true is going to end up being not quite so -- the same as for humanity at any other given moment in time.
We could derive more basic principles, but this one is fundamental and already completely incompatible with our current direction. Right now we're trying to essentially train on the entire corpus of human writing. That is a defacto acknowledgement that the absolute endgame for current tech is simple mimicry, mistakes and all. It'd create a facsimile of impressive intelligence because no human would have a remotely comparable knowledge base, but it'd basically just be a glorified natural language search engine - frozen in time.
I mostly agree with you. But if you think about it mimicry is an aspect of intelligence. If I can copy you and do what you do reliably, regardless of the method used, it does capture an aspect of intelligence. The true game changer is a reflective AI that can automatically improve upon itself
Your quote is a non sequitur to your question. The reason you want to avoid massive data dumps is because there are guaranteed to be errors and flaws. See things like Alpha Go vs Alpha Go Zero. The former was trained on the entirety of human knowledge, the latter was trained entirely on itself.
The zero training version not only ended up dramatically outperforming the 'expert' version, but reached higher levels of competence exponentially faster. And that should be entirely expected. There were obviously tremendous flaws in our understanding of the game, and training on those flaws resulted in software seemingly permanently handicapping itself.
Minimal expert training also has other benefits. The obvious one is that you don't require anywhere near the material and it also enables one to ensure you're on the right track. Seeing software 'invent' fundamental arithmetic is somewhat easier to verify and follow than it producing a hundred page proof advancing, in a novel way, some esoteric edge theory of mathematics. Presumably it would also require orders of magnitude less operational time to achieve such breakthroughs, especially given the reduction in preexisting state.
The moment after human birth the human agent starts a massive information gathering process - that no other system really expects much output from in a coherent way - for 5-10 years. Aka “data dump” some of that data is good, and some of it is bad. This in turn leads to biases, it leads to poor thinking models; everything that you described, is also applicable to every intelligent system - including humans. So again you presupposing that there’s some kind of perfect information benchmark that couldn’t exist.
When that system comes out of the birth canal it already has embedded in it millions of years of encoded expectations predictability systems and functional capabilities that are going to grow independent of what the environment does (but will be certainly shaped in its interactions by the environment).
So no matter what, you have a structured system of interaction that must be loaded with previously encoded data (experience, transfer learning etc) with and it doesn’t matter what type of intelligent system you’re talking about there are foundational assumptions at the physical interaction layer that encode all previous times steps of evolution.
Said an easier way: a lobster, because of the encoded DNA that created it, will never have the same capabilities as a human, because it is structured to process information completely differently and their actuators don’t have the same type and level of granularity as human actuators.
Now assume that you are a lobster compared to a theoretical AGI in sensor-effector combination. Most likely it would be structured entirely differently than you are as a biological thing - but the mere design itself carries with it an encoding of structural information of all previous systems that made it possible.
So by your definition you’re describing something that has never been seen in any system and includes a lot of assumptions about how alternative intelligent systems could work - which is fair because I asked your opinion.
With due respect I do not think you're tackling the fundamental issue, which I do not think is particularly controversial: intelligence and knowledge are distinct things, with the latter created by the former. What we're aiming to do is to create an intelligent system, a system that can create fundamentally new knowledge, and not simply reproduce or remix it on demand.
The next time your in the wilds, it's quite amazing to consider that your ancestors - millennia past, would have looked at, more or less, these exact same wilds but with so much less knowledge. Yet nonetheless they would discover such knowledge - teaching themselves, and ourselves, to build rockets, put a man on the Moon, unlock the secrets of the atom, and so much more. All from zero.
---
What your example and elaboration focus on is the nature of intelligence, and the difficulty in replicating it. And I agree. This is precisely we want to avoid making the problem infinitely more difficult, costly, and time consuming by dumping endless amounts of knowledge in the equation.
Intelligence and knowledge being different things is quite the claim - namely it sounds like you’re stuck in the Cartesian dualist world and having transitioned into statistical empiricism.
I’m curious what epistemological grounding you are basing your claim on
I don't understand how you can equate the two and reconcile the past. The individuals who have pushed society forward in this domain or that scarcely, if ever, had any particular knowledge edge. Cases like Ramanujan [1] exemplify such to the point of absurdity.
If you took the average human from birth and gave them only 'the most primitive first principles', the chance that they would have novel insights into medicine is doubtful.
I also disagree with your following statement:
> Right now we're trying to essentially train on the entire corpus of human writing. That is a defacto acknowledgement that the absolute endgame for current tech is simple mimicry
At worst it's complex mimicry! But I would also say that mimicry is part of intelligence in general and part of how humans discover. It's also easy to see that AI can learn things - you can teach an AI a novel language by feeding in a fairly small amount of words and grammar of example text into context.
I also disagree with this statement:
> One fundamental difference is that AGI would not need some absurdly massive data dump to become intelligent
I don't think how something became intelligent should affect whether it is intelligent or not. These are two different questions.
> you can teach an AI a novel language by feeding in a fairly small amount of words and grammar of example text into context.
You didn't teach it, the model is still the same after you ran that. That is the same as a human following instructions without internalizing the knowledge, he forgets it afterward and didn't learn what he performed. If that was all humans did then there would be no point in school etc, but humans do so much more than that.
As long as LLM are like an Alzheimer's human they will never become a general intelligence. And following instructions is not learning at all, learning is building an internal model for those instructions that is more efficient and general than the instructions themselves, humans do that and that is how we manage to advance science and knowledge.
It depends what you count as learning - you told it something, and it then applied that new knowledge, and if you come back to that conversation in 10 years, it will still have that new knowledge and be able to use it.
Then when OpenAI does another training run it can also internalise that knowledge into the weights.
This is much like humans - we have short term memory (where it doesn't get into the internal model) and then things get baked into long term memory during sleep. AI's have context-level memory, and then that learning gets baked into the model during additional training.
Although whether or not it changed the weights IMO is not a prerequisite for whether something can learn something or not. I think we should be able to evaluate if something can learn by looking at it as a black-box, and we could make a black-box which would meet this definition if you spoke to a LLM and limited it to it's max context length each day, and then ran an overnight training run to incorporate learned knowledge into weights.
Why does the Author choose to ignore the "General" in AGI?
Can ChatGPT drive a car? No, we have specialized models for driving vs generating text vs image vs video etc etc. Maybe ChatGPT could pass a high school chemistry test but it certainly couldn't complete the lab exercises. What we've built is a really cool "Algorithm for indexing generalized data", so you can train that Driving model very similarly to how you train the Text model without needing to understand the underlying data that well.
The author asserts that because ChatGPT can generate text about so many topics that it's general, but it's really only doing 1 thing and that's not very general.
Generally speaking, anyone can learn to use any tool. This isn't true of generative AI systems which can only learn through specialized training with meticulously curated data sets.
People physically unable to use the tool can't learn to use it. This isn't necessarily my view, but one could make a pretty easy argument that the LLMs we have today can't drive a car only because they aren't physically able to control the car.
> but one could make a pretty easy argument that the LLMs we have today can't drive a car only because they aren't physically able to control the car.
Of course they can. We already have computer controlled car systems, the reason LLMs aren't used to drive them is because AI systems that specialize in text are a poor choice for driving - specialized driving models will always outperform them for a variety of technical reasons.
We have compute controlled automobiles, not LLM controlled automobiles.
That was my whole point. Maybe in theory an LLM could learn to drive a car, but they can't today because they don't physically have access to cars they could try to drive just like a person who can't learn to use a tool because they're physically limited from using it.
This isn't true. A curated data set can greatly increase learning efficiency in some cases, but it's not strictly necessary and represents only a fraction of how people learn. Additionally, all curated data sets were created by humans in the first place, a feat that language models could never achieve if we did not program them to do so.
Generality is a continuous value, not a boolean; turned out that "AGI" was poorly defined, and because of that most people were putting the cut-off threshold in different places.
Likewise for "intelligent", and even "artificial".
So no, ChatGPT can't drive a car*. But it knows more about car repairs, defensive driving, global road features (geoguesser), road signs in every language, and how to design safe roads, than I'm ever likely to.
* It can also run python scripts with machine vision stuff, but sadly that's still not sufficient to drive a car… well, to drive one safety, anyway.
Text can be a carrier for any type of signal. The problem gets reduced to that of an interface definition. It’s probably not going to be ideal for driving cars, but if the latency, signal quality, and accuracy is within acceptable constraints, what else is stopping it?
This doesn’t imply that it’s ideal for driving cars, but to say that it’s not capable of driving general intelligence is incorrect in my view.
You can literally today prompt ChatGPT with API instructions to drive a car, then feed it images of a car's window outlooks and have it generate commands for the car (JSON schema restricted structured commands if you like). Text can represent any data thus yes, it is general.
But it cannot think on it's own! Billions of years of evolution couldn't bring human level 'AGI' to many many species, and we think a mere LLM company could do so. AGI isn't just a language model, there's tons of things baked into dna(the way brain functions, it's structure when it grows etc). It's not simply neuron interactions as well. The complexity is mind boggling
Last time I checked, in an Anthropic paper, they asked the model to count something. They examined the logits and a graph showing how it arrived at the answer. Then they asked the model to explain its reasoning, and it gave a completely different explanation, because that was the most statistically probable response to the question. Does that seem like AGI to you?
There is no post factum rationalization here. If you ask a human to think about how they do something before they do it, there's no post factum rationalization. If you ask an LLM to do the same, it will give you a different answer. So, there is a difference. It's all about having knowledge of your internal state and being conscious of your actions and how you perform them, so you can learn from that knowledge. Without that, there is no real intelligence, just statistics.
Yes, humans can post rationalize. But an LLM do nothing but post rationalize, as you yourself admitted humans can think it through beforehand and then actually do what they planned, while an LLM wont follow that plan mentally.
It is easy to see why, since the LLM doesn't communicate what it thinks it communicates what it thinks a human would communicate. A human would explain their inner process, and then go through that inner process. An LLM would explain a humans inner process, and then generate a response using a totally different process.
So while its true that humans doesn't have perfect introspection, the fact that we have introspection about our own thoughts at all is extremely impressive. An LLM has no part that analyzes its own thoughts the way humans do, meaning it has no clue how it thinks.
I have no idea how you would even build introspection into an AI, like how are we able to analyze our own thoughts? What is even a thought? What would this introspection part of an LLM do, what would it look like, would it identify thoughts and talk about them the way we do? That would be so cool, but that is not even on the horizon, I doubt we will ever see that in our lifetime, it would need some massive insight changing the AI landscape at its core to get there.
But, once you have that introspection I think AGI will happen almost instantly. Currently we use dumb math to train the model, that introspection will let the model train itself in an intelligent way, just like humans do. I also think it will never fully replace humans without introspection, intelligent introspection seems like a fundamental part to general intelligence and learning from chaos.
I would argue that this is a fringe opinion that has been adopted by a mainstream scholar, not a mainstream opinion. That or, based on my reading of the article, this person is using a definition of AGI that is very different than the one that most people use when they say AGI.
... that was written in mid-2023. So that opinion piece is trying to redefine 2 year old LLMs like GPT-4 (pre-4o) as AGI. Which can only be described as an absolutely herculean movement of goalposts.
Please, keep telling people that. For my sake. Keep the world asleep as I take advantage of this technology which is literally General Artificial Intelligence that I can apply towards increasing my power.
AGI would mean something which doesn't need direction or guidance to do anything. Like us humans, we don't wait for somebody to give us a task and go do it as if that is our sole existence. We live with our thoughts, blank out, watch TV, read books etc. What we currently have and possibly in the next century as well will be nothing close to an actual AGI.
I don't know if it is optimism or delusions of grandeur that drives people to make claims like AGI will be here in the next decade. No, we are not getting that.
And what do you think would happen to us humans if such AGI is achieved? People's ability to put food on the table is dependent on their labor exchanged for money. I can guarantee for a fact, that work will still be there but will it be equitable? Available to everyone? Absolutely not. Even UBI isn't going to cut it because even with UBI people still want to work as experiments have shown. But with that, there won't be a majority of work especially paper pushing mid level bs like managers on top of managers etc.
If we actually get AGI, you know what would be the smartest thing for such an advanced thing to do? It would probably kill itself because it would come to the conclusion that living is a sin and a futile effort. If you are that smart, nothing motivates you anymore. You will be just a depressed mass for all your life.
I think there's a useful distinction that's often missed between AGI and artificial consciousness. We could conceivably have some version of AI that reliably performs any task you throw at it consistently with peak human capabilities, given sufficient tools or hardware to complete whatever that task may be, but lacks subjective experience or independent agency; I would call that AGI.
The two concepts have historically been inexorably linked in sci-fi, which will likely make the first AGI harder to recognize as AGI if it lacks consciousness, but I'd argue that simple "unconscious AGI" would be the superior technology for current and foreseeable needs. Unconscious AGI can be employed purely as a tool for massive collective human wealth generation; conscious AGI couldn't be used that way without opening a massive ethical can of worms, and on top of that its existence would represent an inherent existential threat.
Conscious AGI could one day be worthwhile as something we give birth to for its own sake, as a spiritual child of humanity that we send off to colonize distant or environmentally hostile planets in our stead, but isn't something I think we'd be prepared to deal with properly in a pre-post-scarcity society.
It isn't inconceivable that current generative AI capabilities might eventually evolve to such a level that they meet a practical bar to be considered unconscious AGI, even if they aren't there yet. For all the flak this tech catches, it's easy to forget that capabilities which we currently consider mundane were science fiction only 2.5 years ago (as far as most of the population was concerned). Maybe SOTA LLMs fit some reasonable definition of "emerging AGI", or maybe they don't, but we've already shifted the goalposts in one direction given how quickly the Turing test became obsolete.
Personally, I think current genAI is probably a fair distance further from meeting a useful definition of AGI than those with a vested interest in it would admit, but also much closer than those with pessimistic views of the consequences of true AGI tech want to believe.
One sci-fi example could be based on the replicators from Star Trek, who are able to synthesize any meals on demand.
It is not hard to imagine a "cooking robot" as a black box that — given the appropriate ingredients — would cook any dish for you. Press a button, say what you want, and out it comes.
Internally, the machine would need to perform lots of tasks that we usually associate with intelligence, from managing ingredients and planning cooking steps, to fine-grained perception and manipulation of the food as it is cooking. But it would not be conscious in any real way. Order comes in, dish comes out.
Would we use "intelligent" to describe such a machine? Or "magic"?
I immediately thought of Star Trek too, I think the ship's computer was another example of unconscious intelligence. It was incredibly capable and could answer just about any request that anyone made of it. But it had no initiative or motivation of its own.
Regarding "We could conceivably have some version of AI that reliably performs any task you throw at it consistently" - it is very clear to anyone who just looks at the recent work by Anthropic analyzing how their LLM "reasons" that such a thing will never come from LLMs without massive unknown changes - and definitely not from scale - so I guess the grandparent is absolute right that openai is nor really working on this.
I agree. AGI is meaningless as a term if it doesn't mean completely autonomous agentic intelligence capable of operating on long-term planning horizons.
Edit: because if "AGI" doesn't mean that... then what means that and only that!?
> Edit: because if "AGI" doesn't mean that... then what means that and only that!?
"Agentic AI" means that.
Well, to some people, anyway. And even then, people are already arguing about what counts as agency.
That's the trouble with new tech, we have to invent words for new stuff that was previously fiction.
I wonder, did people argue if "horseless carriages" were really carriages? And "aeroplane" how many argued that "plane" didn't suit either the Latin or Greek etymology for various reasons?
We never did rename "atoms" after we split them…
And then there's plain drift: Traditional UK Christmas food is the "mince pie", named for the filling, mincemeat. They're usually vegetarian and sometimes even vegan.
Agents can operate in narrow domains too though, so to fit the G part of AGI the agent needs to be non-domain specific.
It's kind of a simple enough concept... it's really just something that functions on par with how we do. If you've built that, you've built AGI. If you haven't built that, you've built a very capable system, but not AGI.
> Agents can operate in narrow domains too though, so to fit the G part of AGI the agent needs to be non-domain specific.
"Can", but not "must". The difference between an LLM being harnessed to be a customer service agent, or a code review agent, or a garden planning agent, can be as little as the prompt.
And in any case, the point was that the concept of "completely autonomous agentic intelligence capable of operating on long-term planning horizons" is better described by "agentic AI" than by "AGI".
> It's kind of a simple enough concept... it's really just something that functions on par with how we do.
"On par with us" is binary thinking — humans aren't at the same level as each other.
The problem we have with LLMs is the "I"*, not the "G". The problem we have with AlphaGo and AlphaFold is the "G", not the ultimate performance (which is super-human, an interesting situation given AlphaFold is a mix of Transformer and Diffusion models).
For many domains, getting a degree (or passing some equivalent professional exam) is just the first step, and we have a long way to go from there to being trusted to act competently, let alone independently. Someone who started a 3-year degree just before ChatGPT was released, will now be doing their final exams, and quite a lot of LLMs operate like they have just about scraped through degrees in almost everything — making them wildly superhuman with the G.
The G-ness of an LLM only looks bad when compared to all of humanity collectively; they are wildly more general in their capabilities than any single one of us — there are very few humans who can even name as many languages as ChatGPT speaks, let alone speak them.
* they need too many examples, only some of that can be made up for by the speed difference that lets machines read approximately everything
Think about it - the original definition of AGI was basically a machine that can do absolutely anything at a human level of intelligence or better.
That kind of technology wouldn't just appear instantly in a step change. There would be incremental progress. How do you describe the intermediate stages?
What about a machine that can do anything better than the 50th percentile of humans? That would be classified as "Competent AGI", but not "Expert AGI" or ASI.
> fancy search engine/auto completer
That's an extreme oversimplification. By the same reasoning, so is a person. They are just auto completing words when they speak. No that's not how deep learning systems work. It's not auto complete..
It's really not. The Space Shuttle isn't an emerging interstellar spacecraft, it's just a spacecraft. Throwing emerging in front of a qualifier to dilute it is just bullshit.
> By the same reasoning, so is a person. They are just auto completing words when they speak.
We have no evidence of this. There is a common trope across cultures and history of characterising human intelligence in terms of the era's cutting-edge technology. We did it with steam engines [1]. We did it with computers [2]. We're now doing it with large language models.
Technically it is a refinement, as it distinguishes levels of performance.
The General Intelligence part of AGI refers to its ability to solve problems that it was not explicitly trained to solve, across many problem domains. We already have examples of the current systems doing exactly that - zero shot and few shot capabilities.
> We have no evidence of this.
That's my point. Humans are not "autocompleting words" when they speak.
> Technically it is a refinement, as it distinguishes levels of performance
No, it's bringing something out of scope into the definition. Gluten-free means free of gluten. Gluten-free bagel verus sliced bread is a refinement--both started out under the definition. Glutinous bread, on the other hand, is not gluten free. As a result, "almost gluten free" is bullshit.
> That's my point. Humans are not "autocompleting words" when they speak
Humans are not. LLMs are. It turns out that's incredibly powerful! But it's also limiting in a way that's fundamentally important to the definition of AGI.
LLMs bring us closer to AGI in the way the inventions of writing, computers and the internet probably have. Calling LLMs "emerging AGI" pretends we are on a path to AGI in a way we have zero evidence for.
Bad analogy. That's a binary classification. AGI systems can have degrees of performance and capability.
> Humans are not. LLMs are.
My point is that if you oversimplify LLMs to "word autocompletion" then you can make the same argument for humans. It's such an oversimplification of the transformer / deep learning architecture that it becomes meaningless.
> That's a binary classification. AGI systems can have degrees of performance and capability
The "g" in AGI requires the AI be able to perform "the full spectrum of cognitively demanding tasks with proficiency comparable to, or surpassing, that of humans" [1]. Full and not full are binary.
> if you oversimplify LLMs to "word autocompletion" then you can make the same argument for humans
No, you can't, unless you're pre-supposing that LLMs work like human minds. Calling LLMs "emerging AGI" pre-supposes that LLMs are the path to AGI. We simply have no evidence for that, no matter how much OpenAI and Google would like to pretend it's true.
Why are you linking a Wikipedia page like it's the ground zero for the term? Especially when neither article the page link to justify that definition see the term as a binary accomplishment.
The g in AGI is General. I don't what world you think Generality isn't a spectrum, but it's sure as hell isn't this one.
That's right, and the Wikipedia page refers to the classification system:
"A framework for classifying AGI by performance and autonomy was proposed in 2023 by Google DeepMind researchers. They define five performance levels of AGI: emerging, competent, expert, virtuoso, and superhuman"
In the second paragraph:
"Some researchers argue that state‑of‑the‑art large language models already exhibit early signs of AGI‑level capability, while others maintain that genuine AGI has not yet been achieved."
The entire article makes it clear that the definitions and classifications are still being debated and refined by researchers.
Then you are simply rejecting any attempts to refine the definition of AGI. I already linked to the Google DeepMind paper. The definition is being debated in the AI research community. I already explained that definition is too limited because it doesn't capture all of the intermediate stages. That definition may be the end goal, but obviously there will be stages in between.
> No, you can't, unless you're pre-supposing that LLMs work like human minds.
You are missing the point. If you reduce LLMs to "word autocompletion" then you completely ignore the the attention mechanism and conceptual internal representations. These systems have deep learning models with hundreds of layers and trillions of weights. If you completely ignore all of that, then by the same reasoning (completely ignoring the complexity of the human brain) we can just say that people are auto-completing words when they speak.
> I already linked to the Google DeepMind paper. The definition is being debated in the AI research community
Sure, Google wants to redefine AGI so it looks like things that aren’t AGI can be branded as such. That definition is, correctly in my opinion, being called out as bullshit.
> obviously there will be stages in between
We don’t know what the stages are. Folks in the 80s were similarly selling their expert systems as a stage to AGI. “Emerging AGI” is a bullshit term.
> If you reduce LLMs to "word autocompletion" then you completely ignore the the attention mechanism and conceptual internal representations. These systems have deep learning models with hundreds of layers and trillions of weights
It is not a redefinition. It's a classification for AGI systems. It's a refinement.
Other researchers are also trying to classify AGI systems. It's not just Google. Also, there is no universally agreed definition of AGI.
> We don’t know what the stages are. Folks in the 80s were similarly selling their expert systems as a stage to AGI. “Emerging AGI” is a bullshit term.
Generalization is a formal concept in machine learning. There can be degrees of generalized learning performance. This is actually measurable. We can compare the performance of different systems.
While I also hold a peer comment's view that the Turing Test is meaningless, I would further add that even that has not been meaningfully beaten.
In particular we redefined the test to make it passable. In Turing's original concept the competent investigator and participants were all actively expected to collude against the machine. The entire point is that even with collusion, the machine would be able to pass. Instead modern takes have paired incompetent investigators alongside participants colluding with the machine, probably in an effort to be part 'of something historic'.
In "both" (probably more, referencing the two most high profile - Eugene and the large LLMs) successes, the interrogators consistently asked pointless questions that had no meaningful chance of providing compelling information - 'How's your day? Do you like psychology? etc' and the participants not only made no effort to make their humanity clear, but often were actively adversarial obviously intentionally answering illogically, inappropriately, or 'computery' to such simple questions. And the tests are typically time constrained by woefully poor typing skills (this the new normal in the smartphone gen?) to the point that you tend to get anywhere from 1-5 interactions of a few words each.
The problem with any metric for something is that it often ends up being gamed to be beaten, and this is a perfect example of that.
I mean, I am pretty sure that I won't be fooled by a bot, if I get the time to ask the right questions.
And I did not looked into it (I also don'think the test has too much relevance), but fooling the average person sounds plausible by now.
Now sounding plausible is what LLMs are optimized for and not being plausible, still, I would not have thought we get so far so quick 10 years ago. So I am very hesistant about the future.
The very people whose theories about language are now being experimentally verified by LLMs, like Chomsky, have also been discrediting the Turing test as pseudoscientific nonsense since early 1990s.
It's one of those things like the Kardashev scale, or Level 5 autonomous driving, that's extremely easy to define and sounds very cool and scientific, but actually turns out to have no practical impact on anything whatsoever.
I feel like, if nothing else, this new wave of AI products is rapidly demonstrating the lack of faith people have in their own intelligence -- or maybe, just the intelligence of other human beings. That's not to say that this latest round of AI isn't impressive, but legions of apologists seem to forget that there is more to human cognition than being able to regurgitate facts, write grammatically-correct sentences, and solve logical puzzles.
> legions of apologists seem to forget that there is more to human cognition than being able to regurgitate facts, write grammatically-correct sentences, and solve logical puzzles
To be fair, there is a section of the population whose useful intelligence can roughly be summed up as that or worse.
I think this takes an unnecessarily narrow view of what "intelligence" implies. It conflates "intelligence" with fact-retention and communicative ability. There are many other intelligent capabilities that most normally-abled human beings possess, such as:
- Processing visual data and classifying objects within their field of vision.
- Processing auditory data, identifying audio sources and filtering out noise.
- Maintaining an on-going and continuous stream of thoughts and emotions.
- Forming and maintaining complex memories on long-term and short-term scales.
- Engaging in self-directed experimentation or play, or forming independent wants/hopes/desires.
I could sit here all day and list the forms of intelligence that humans and other intelligent animals display which have no obvious analogue in an AI product. It's true that individual AI products can do some of these things, sometimes better than humans could ever, but there is no integrated AGI product that has all these capabilities. Let's give ourselves a bit of credit and not ignore or flippantly dismiss our many intelligent capabilities as "useless."
> It conflates "intelligence" with fact-retention and communicative ability
No, I’m using useful problem solving as my benchmark. There are useless forms of intelligence. And that’s fine. But some people have no useful intelligence and show no evidence of the useless kind. They don’t hit any of the bullets you list, there just isn’t that curiosity and drive and—I suspect—capacity to comprehend.
I don’t think it’s intrinsic. I’ve seen pets show more curiosity than some folk. But due to nature and nurture, they just aren’t intelligent to any material stretch.
Remember however that their charter specifies: "If a value-aligned, safety-conscious project comes close to building AGI before we do, we commit to stop competing with and start assisting this project"
It does have some weasel words around value-aligned and safety-conscious which they can always argue but this could get interesting because they've basically agreed not to compete. A fairly insane thing to do in retrospect.
Who defines "value-aligned, safety-conscious project"?
"Instead of our current complex non-competing structure—which made sense when it looked like there might be one dominant AGI effort but doesn’t in a world of many great AGI companies—we are moving to a normal competing structure where ..." is all it takes
AGI could be a winner-take-all market... for the AGI, specifically for the first one that's General and Intelligent enough to ensure its own survival and prevent competing AGI efforts from succeeding...
How would an AGI prevent others from competing? Sincere question. That seems like something that ASI would be capable of. If another company released an AGI, how would the original stifle it? I get that the original can self-improve to try to stay ahead, but that doesn't necessarily mean it self-improves the best or most efficiently, right?
AGI used to be synonymous with ASI; it's still unclear to me it's even possible to build a sufficiently general AI - that is, as general as humans - without it being an ASI just by virtue of being in silico, thus not being constrained in scale or efficiency like our brains are.
If it was first, it could have self-improved more, to the point that it has the capacity to prevent competition, while the competition does not have the capacity to defend itself against superior AGI. This all is so hypothetical and frankly far from what we're seeing in the market now. Funny how we're all discussing dystopian scifi scenarios now.
Homo Sapiens wiped out every other intelligent hominid and every other species on Earth exists at our mercy. That looks a lot like the winners (humans) taking all.
Well, yeah, the world in which it is winner take all is the one where it accelerates productivity so much such that the first firm to achieve it doesn't provide access to its full capabilities directly to oursiders but uses it themselves and conquers every other field of endeavor.
That's always been pretty overtly the winner-take-all AGI scenario.
AGI might not be fungible. From the trends today it's more likely there will be multiple AGIs with different relative strengths and weakness, different levels of accessibility and compliance, different development rates, and different abilities to be creative and surprising.
OpenAI is winning in a similar way that Apple is winning in smartphones.
OpenAI is capturing most of the value in the space (generic LLM models), even though they have competitors who are beating them on price or capabilities.
I think OpenAI may be able to maintain this position at least for the medium term because of their name recognition/prominence and they are still a fast mover.
I also think the US is going to ban all non-US LLM providers from the US market soon for "security reasons."
Apple is not the right analogy. OpenAI has first mover advantage and they have a widely recognized brand name — ChatGPT — and that’s kind of it. Anyone (with very deep pockets) can buy Nvidia chips and go to town if they have a better or equivalent idea. There was a brief time (long before I was born) when “Univac” was synonymous with “computer.”
> I also think the US is going to ban all non-US LLM providers from the US market soon for "security reasons."
Well Trump is interested in tariffing movies and South Korea took DeepSeek off mobile app stores, so they certainly may try. But for high-end tasks, DeepSeek R1 671B is available for download, so any company with a VPN to download it and the necessary GPUs or cloud credits can run it. And for consumers, DeepSeek V3's distilled models are available for download, so anyone with a (~4 year old or newer) Mac or gaming PC can run them.
If the only thing keeping these companies valuations so high is banning the competition, that's not a good sign for their long-term value. If you have to ban the competition, you can't be feeling good about what you're making.
For what it's worth, I think GPT o3 and o1, Gemini 2.5 Pro and Claude 3.7 Sonnet are good enough to compete. DeepSeek R1 is often the best option (due to cost) for tasks that it can handle, but there are times where one of the other models can achieve a task that it can't.
But if the US is looking to ban Chinese models, then that could suggest that maybe these models aren't good enough to raise the funding required for newer, significantly better (and more expensive) models. That, or they just want to stop as much money as possible from going to China. Banning the competition actually makes the problem worse though, as now these domestic companies have fewer competitors. But I somewhat doubt there's any coherent strategy as to what they ban, tariff, etc.
What do you consider an "LLM provider"? Is it a website where you interact with a language model by uploading text or images? That definition might become too broad too quickly. Hard to ban.
the bulk of money comes from enterprise users. Just need to call 500 CEOs from the S&P500 list, and enforce via "cyber data safety" enforcement via SEC or something like that.
everyone will roll over if all large public companies roll over (and they will)
IE once captured all of the value in browserland, with even much higher mindshare and market dominance than OpenAI has ever had. Comparing with Apple (= physical products) is Apples to oranges (heh).
Their relationship with MS breaking down is a bad omen. I'm already seeing non-tech users who use "Copilot" because their spouse uses it at work. Barely knowing it's rebadged GPT. You think they'll switch when MS replaces the backend with e.g. Anthropic? No chance.
MS, Google and Apple and Meta have gigantic levers to pull and get the whole world to abandon OpenAI. They've barely been pulling them, but it's a matter of time. People didn't use Siri and Bixby because they were crap. Once everyone's Android has a Gemini button that's just as good as GPT (which it already is (it's better) for anything besides image generation), people are going to start pressing them. And good luck to OpenAI fighting that.
Companies that are contractors with the US government already aren’t allowed to use Deepseek even if its an airgapped R1 model is running on our own hardware. Legal told us we can’t run any distills of it or anything. I think this is very dumb.
Please promise to come back to this comment in 2030 and playfully mock me for ever being worried and I will buy you a coffee. If AGI is invented before 2030 please buy me one and let me mock you playfully.
to me it sounds like an admission that AGI is bullshit! AGI would be so disruptive to the current economic regime that "winner takes all" barely covers it, I think. Admitting they will be in normal competition with other AI companies implies specializations and niches to compete, which means Artificial Specialized Intelligence, NOT general intelligence!
and that makes complete sense if you don't have a lay person's understanding of the tech. Language models were never going to bring about "AGI."
If they think AGI is imminent the value of that payday is very limited. I think the grandparent is more correct: OpenAI is admitting that near term AGI - which, being that the only one anyone really cares about is the case with exponential self improvement - isn't happening any time soon. But that much is obvious anyway despite the hyperbolic nonsense now common around AI discussions.
If I were a person like several of the people working on AI right now (or really, just heading up tech companies), I could be the kind to look at a possible world-ending event happening in the next - eh, year, let's say - and just want to have a party at the end of the world.
It will likely require research breakthroughs, significant hardware advancement, and anything from a few years to a few decades. But it's coming.
ChatGPT was released 2.5 years ago, and look at all the crazy progress that has been made in that time. That doesn't mean that the progress has to continue, we'll probably see a stall.
But AIs that are on a level with humans for many common tasks is not that far off.
Either that, or this AI boom mirrors prior booms. Those booms saw a lot of progress made, a lot of money raised, then collapsed and led to enough financial loss that AI went into hibernation for 10+ years.
There's a lot of literature on this, and if you've been in the industry for any amount of time since the 1950s, you have seen at least one AI winter.
probably true but this statement would be true if when is 2308 which would defeat the purpose of the statement. when first cars started rolling around some mates around the campfire we saying “not if but when” we’ll have flying cars everywhere and 100 years later (with amazing progress in car manufacturing) we are nowhere near… I think saying “when, not if” is one of those statements that while probably indisputable in theory is easily disputable in practice. give me “when” here and I’ll put up $1,000 to a charity of your choice if you are right and agree to do the same thing if wrong
you can see a pattern of fairly steady progress in different aspects, like they matched humans for image recognition around 2015 but 'complex reasoning' is still much worse than humans but rising.
Looking at the graph, I'd guess maybe five years before it can do all human skills which is roughly AGI?
I've got a personal AGI test of being able to fix my plumbing, given a robot body. Which they are way off just now.
It is already here, kinda. I mean look at how it passes the bar exam, solves math olympiad level questions, generates video, art, music. What else are you looking for? It already has penetrated into job market causing significant disruption in programming. We are not seeing flying cars but we are witnessing things even not talked about around campfire. Seriously even 4 years ago, would you think all these would happen?
To begin with, systems that don't tell people to use elmer's glue to keep the cheese from sliding off the pizza, displaying a fundamental lack of understanding of.. everything. At minimum it needs to be able to reliably solve hard, unique, but well-defined problems like a group of the most cohesive intelligent people could. It's certainly not AGI until it can do a better job than the most experienced, talented, and intelligent knowledge workers out there.
Every major advancement (which LLMs certainly are) has caused some disruption in the fields it affected, but that isn't useful criteria that can differentiate between "crude but useful tool" from "AGI".
Majority of people on earth don't solve hard, unique, but well-defined problems, do we? I dont expect AGI to to solve one of Hilbert's list of problems (yet). Your definition of AGI is a bit too imposing.
Saying that I believe you would get answers from an LLM better than most of the answers you would get from an average human.
IMHO the trend is obvious and we will see if it stalls or keeps the pace.
I don't mean "hard" in the sense that it can easily solve novel problems that no living human knows how to solve, although any "general" intelligence should certainly be capable of learning and making progress on these just like human would, but without limitations of human memory, attention span, relatively short lifetime, and other human needs.
I mean "hard" in the sense that it can reliably replace the best software developers, civil engineers, lawyers, diagnosticians. Not just in economic sense, but by reliably matching the quality of their work 100% of the time.
It should be capable of methodically and reliably arriving at correct answers without expert intervention. It shouldn't be the case that some people claim that they don't know how to code and the LLM generated an entire project for them, while I can confidently claim that LLMs fall flat on their face almost every time I try to use them for more delicate business logic.
Progress is not just a function of technical possibility( even if it exists) it is also economics.
It has taken tens to hundred of billions of dollars without equivalent economic justification(yet) before to reach here. I am not saying economic justification doesn't exist or wont come in the future, just that the upfront investment and risk is already in order of magnitude of what the largest tech companies can expend.
If the the next generation requires hundreds of billions or trillions [2] upfront and a very long time to make returns, no one company (or even country) could allocate that kind of resources.
Many cases of such economically limited innovations[1], nuclear fusion is the classic always 20 years away example. Another close one is anything space related, we cannot replicate in next 5 years what we already achieved from 50 years ago of say landing on the moon and so on.
From a just a economic perspective it is a definitely a "If", without even going into the technology challenges.
[1]Innovations in cost of key components can reshape economics equation, it does happen (as with spaceX) but it also not guaranteed like in fusion.
[2] The next gen may not be close enough to AGI. AGI could require 2-3 more generations ( and equivalent orders of magnitude of resources), which is something the world is unlikely to expend resources on even if it had them.
I agree that LLMs are hurting the general population’s capacity to think (assuming they use it often. I’ve certainly noticed a slight trend among students I’ve taught to use less effort, and myself to some extent).
I don’t agree that this will affect ML progress much, since the general population isn’t contributing to core ML research.
I think this is right but also missing a useful perspective.
Most HN people are probably too young to remember that the nanotech post-scarcity singularity was right around the corner - just some research and engineering way - which was the widespread opinion in 1986 (yes, 1986). It was _just as dramatic_ as today's AGI.
That took 4-5 years to fall apart, and maybe a bit longer for the broader "nanotech is going to change everything" to fade. Did nanotech disappear? No, but the notion of general purpose universal constructors absolutely is dead. Will we have them someday? Maybe, if humanity survives a hundred more years or more, but it's not happening any time soon.
There are a ton of similarities between nanotech-nanotech singularity and the moderns LLM-AGI situation. People point(ed) to "all the stuff happening" surely the singularity is on the horizon! Similarly, there was the apocalytpic scenario that got a ton of attention and people latching onto "nanotech safety" - instead of runaway AI or paperclip engines, it was Grey Goo (also coined in 1986).
The dynamics of the situation, the prognostications, and aggressive (delusional) timelines, etc. are all almost identical in a 1:1 way with the nanotech era.
I think we will have both AGI and general purpose universal constructors, but they are both no less than 50 years away, and probably more.
So many of the themes are identical that I'm wondering if it's a recurring kind of mass hysteria. Before nanotech, we were on the verge of genetic engineering (not _quite_ the same level of hype, but close, and pretty much the same failure to deliver on the hype as nanotech) and before that the crazy atomic age of nuclear everything.
Yes, yes, I know that this time is different and that AI is different and it won't be another round of "oops, this turned out to be very hard to make progress on and we're going to be in a very slow, multi-decade slow-improvement regime, but that has been the outcome of every example of this that I can think of.
I won't go too far out on this limb, because I kind of agree with you... but to be fair -- 1980s-1990s nanotech did not attract this level of investment, nor was it visible to ordinary people, nor was it useful to anyone except researchers and grant writers.
It seems like nanotech is all around us now, but the term "nanotech" has been redefined to mean something different (larger scale, less amazing) from Drexler's molecular assemblers.
> Did nanotech disappear? No, but the notion of general purpose universal constructors absolutely is dead. Will we have them someday? Maybe, if humanity survives a hundred more years or more,
I thought this was a "we know we can't" thing rather than a "not with current technology" thing?
Specific cases are probably impossible, though there's always hope. After all, to ue the example the nanotech people loved: there are literal assemblers all around you. Whether we can have singular device that can build anything (probably not - energy limits and many many other issues) or factories that can work on atomic scale (maybe) is open, I think. The idea of little robots was kind of visibly silly even at the peak.
The idea of scaling up LLMs and hoping is .. pretty silly.
Every consumer has very useful AI at their fingertips right now. It's eating the software engineering world rapidly. This is nothing like nanotech in the 80s.
Sure. But fancy autocomplete for a very limited industry (IT) plus graphics generation and a few more similar items, are indeed useful. Just like "nanotech" coating of say optics or in the precise machinery or all other fancy nano films in many industries. Modern transistors are close to nano scale now, etc.
The problem is that the distance between a nano thin film or an interesting but ultimately rigid nano scale transistor and a programmable nano level sized robot is enormous, despite similar sizes. Same like the distance between an autocomplete heavily relying on the preexisting external validators (compilers, linters, static code analyzers etc.) and a real AI capable of thinking is equally enormous.
Could you elaborate on the progress that has been made?
To me, it seems only small/incremental changes are made between models with all of them still hallucinating.
I can see no clear steps towards AGI.
"X increased exponentially in the past, therefore it will increase exponentially in the same way in the future" is fallacious. There is nothing guaranteeing indefinite uncapped growth in capabilities of LLMs. An exponential curve and a sigmoidal curve look the same until a certain point.
Yeah, it is a pretty good bet that any real process that produces something that looks like an exponential curve over time is the early phase of a sigmoid curve, because all real processes have constraints.
And if we apply the 80/20 rule, feels like we're at about 50-75% right now. So we're almost getting close to done with the easy parts. Then come the hard parts.
I don’t think that’s a safe foregone conclusion. What we’ve seen so far is very very powerful pattern matchers with emergent properties that frankly we don’t fully understand. It very well may be the road to AGI, or it may stop at the kind of things we can do in our subconscious—but not what it takes to produce truly novel solutions to never before seen problems. I don’t think we know.
I don't read it that way. It reads more like AGIs will be like very smart people and rather than having one smart person/AGI, everyone will have one. There's room for both Beethoven and Einstein although they were both generally intelligent.
> Instead of our current complex capped-profit structure—which made sense when it looked like there might be one dominant AGI effort but doesn’t in a world of many great AGI companies—we are moving to a normal capital structure where everyone has stock. This is not a sale, but a change of structure to something simpler.