Leaked Documents Show OpenAI Has a Very Clear Definition of ‘AGI’
We finally have a real definition of the elusive "AGI."
AGI (artificial general intelligence) will be achieved once OpenAI has developed an AI system that can generate at least $100 billion in profits
nothing to do with actual capabilities.. just the ability to make piles and piles of money.
The same way these capitalists evaluate human beings.
Guess we're never getting AGI then, there's no way they end up with that much profit before this whole AI bubble collapses and their value plummets.
That's an Onion level of capitalism
The context here is that OpenAI has a contract with Microsoft until they reach AGI. So it's not a philosophical term but a business one.
For fun, I used this idea in chatgpt using similar logic:
Hello, can you explain to me how a strict profit generated number would result in a specific goal, such as if I make 1000 dollars I would suddenly be capable of owning a house. I find the idea strange and backwards.
You're correct to find this idea odd—it reflects a misunderstanding of how financial goals, like owning a house, relate to income. Here's why this thinking is flawed:
Example of Goal-Oriented Thinking:
If you wanted to own a $200,000 house:
You might need a 10% down payment ($20,000).
Closing costs could add $5,000–$10,000.
Monthly income would need to support a mortgage payment (let’s say $1,200/month).
Steps would include saving regularly, reducing debt, and improving credit, rather than focusing on an isolated profit milestone like $1,000.
Summary:
Focusing on a single, arbitrary profit number like $1,000 doesn’t align with the comprehensive planning required for significant financial goals like home ownership. Instead, success depends on a holistic view of your finances and structured goal-setting.
Lol. We're as far away from getting to AGI as we were before the whole LLM craze. It's just glorified statistical text prediction, no matter how much data you throw at it, it will still just guess what's the next most likely letter/token based on what's before it, that can't even get it's facts straith without bullshitting.
If we ever get it, it won't be through LLMs.
I hope someone will finally mathematically prove that it's impossible with current algorithms, so we can finally be done with this bullshiting.
There are already a few papers about diminishing returns in LLM.
I hope someone will finally mathematically prove that it's impossible with current algorithms, so we can finally be done with this bullshiting.
They did! Here's a paper that proves basically that:
van Rooij, I., Guest, O., Adolfi, F. et al. Reclaiming AI as a Theoretical Tool for Cognitive Science. Comput Brain Behav 7, 616–636 (2024). https://doi.org/10.1007/s42113-024-00217-5
Basically it formalizes the proof that any black box algorithm that is trained on a finite universe of human outputs to prompts, and capable of taking in any finite input and puts out an output that seems plausibly human-like, is an NP-hard problem. And NP-hard problems of that scale are intractable, and can't be solved using the resources available in the universe, even with perfect/idealized algorithms that haven't yet been invented.
This isn't a proof that AI is impossible, just that the method to develop an AI will need more than just inferential learning from training data.
Doesn't that just say that AI will never be cheap? You can still brute force it, which is more or less how back propagation works.
I don't think "intelligence" needs to have a perfect "solution", it just needs to do things well enough to be useful. Which is how human intelligence developed, evolutionarily - it's absolutely not optimal.
Thank you, it was an interesting read.
Unfortunately, as I was looking more into it, I've stumbled upon a paper that points out some key problems with the proof. I haven't looked into it more and tbh my expertise in formal math ends at vague memories from CS degree almost 10 years ago, but the points do seem to make sense.
Roger Penrose wrote a whole book on the topic in 1989. https://www.goodreads.com/book/show/179744.The_Emperor_s_New_Mind
His points are well thought out and argued, but my essential takeaway is that a series of switches is not ever going to create a sentient being. The idea is absurd to me, but for the people that disagree? They have no proof, just a religious furver, a fanaticism. Simply stated, they want to believe.
All this AI of today is the AI of the 1980s, just with more transistors than we could fathom back then, but the ideas are the same. After the massive surge from our technology finally catching up with 40-60 year old concepts and algorithms, most everything has been just adding much more data, generalizing models, and other tweaks.
What is a problem is the complete lack of scalability and massive energy consumption. Are we supposed to be drying our clothes at a specific our of the night, and join smart grids to reduce peak air conditioning, to scorn bitcoin because it uses too much electricity, but for an AI that generates images of people with 6 fingers and other mangled appendages, that bullshit anything it doesn't know, for that we need to build nuclear power plants everywhere. It's sickening really.
So no AGI anytime soon, but I am sure Altman has defined it as anything that can make his net worth 1 billion or more, no matter what he has to say or do.
Until you can see the human soul under a microscope, we can't make rocks into people.
What do you think Sam Altman's net worth is currently?
I mean, human intelligence is ultimately too "just" something.
And 10 years ago people would often refer to "Turing test" and imitation games in the sense of what is artificial intelligence and what is not.
My complaint to what's now called AI is that it's as similar to intelligence as skin cells grown in the form of a dck are similar to a real dck with its complexity. Or as a real-size toy building is similar to a real building.
But I disagree that this technology will not be present in a real AGI if it's achieved. I think that it will be.
I'm not sure that not bullshitting should be a strict criterion of AGI if whether or not it's been achieved is gauged by its capacity to mimic human thought
It's impossible to disprove statements that are inherently unscientific.
Lol. We’re as far away from getting to AGI as we were before the whole LLM craze. It’s just glorified statistical text prediction, no matter how much data you throw at it, it will still just guess what’s the next most likely letter/token based on what’s before it, that can’t even get it’s facts straith without bullshitting.
This is correct, and I don't think many serious people disagree with it.
If we ever get it, it won’t be through LLMs.
Well... depends. LLMs alone, no, but the researchers who are working on solving the ARC AGI challenge, are using LLMs as a basis. The one which won this year is open source (all are if are eligible for winning the prize, and they need to run on the private data set), and was based on Mixtral. The "trick" is that they do more than that. All the attempts do extra compute at test time, so they can try to go beyond what their training data allows them to do "fine". The key for generality is trying to learn after you've been trained, to try to solve something that you've not been prepared for.
Even OpenAI's O1 and O3 do that, and so does the one that Google has released recently. They are still using heavily an LLM, but they do more.
I hope someone will finally mathematically prove that it’s impossible with current algorithms, so we can finally be done with this bullshiting.
I'm not sure if it's already proven or provable, but I think this is generally agreed. just deep learning will be able to fit a very complex curve/manifold/etc, but nothing more. It can't go beyond what was trained on. But the approaches for generalizing all seem to do more than that, doing search, or program synthesis, or whatever.
I’m pretty sure the simplest way to look at is an LLM can only respond, not generate anything on its own without prompting. I wish humans were like that sometimes, especially a few in particular. I would think an AGI would be capable of independent thought, not requiring the prompt.
What is your brain doing if not statistical text prediction?
The show Westworld portrayed it pretty good. The idea of jumping from text prediction to conscience doesn't seem that unlikely. It's basically text prediction on a loop with some exterior inputs to interact.
Human brains also do processing of audio, video, self learning, feelings, and many more that are definitely not statistical text. There are even people without "inner monologue" that function just fine
Some research does use LLM in combination with other AI to get better results overall, but purely LLM isn't going to work.
How to tell me you're stuck in your head terminally online without telling me you're stuck in your head terminally online.
conscience
ok buddy
It's "free will". They chose to say what they wanted.
At least this is what the old religions teach. I don't know what AI preachers you're learning this nonsense from.
We taught sand to do math
And now we're teaching it to dream
All the stupid fucks can think to do with it
Is sell more cars
Cars, and snake oil, and propaganda
Then we untaught it to do math.
https://medium.com/tomorrows-jobs-today/why-chatgpt-cant-do-basic-math-fc1a132e9773
I dunno, I don't do math very well when I dream.
this is almost a haiku
"It's at a human-level equivalent of intelligence when it makes enough profits" is certainly an interesting definition and, in the case of the C-suiters, possibly not entirely wrong.
We've had definition for AGI for decades. It's a system that can do any cognitive task as well as a human can or better. Humans are "Generally Intelligent" replicate the same thing artificially and you've got AGI.
So if you give a human and a system 10 tasks and the human completes 3 correctly, 5 incorrectly and 3 it failed to complete altogether... And then you give those 10 tasks to the software and it does 9 correctly and 1 it fails to complete, what does that mean. In general I'd say the tasks need to be defined, as I can give very many tasks to people right now that language models can solve that they can't, but language models to me aren't "AGI" in my opinion.
Agree. And these tasks can't be tailored to the AI in order for it to have a chance. It needs to drive to work, fix the computers/plumbing/whatever there, earn a decent salary and return with some groceries and cook dinner. Or at least do something comparable to a human. Just wording emails and writing boilerplate computer-code isn't enough in my eyes. Especially since it even struggles to do that. It's the "general" that is missing.
Its a definition, but not an effective one in the sense that we can test and recognize it. Can we list all cognitive tasks a human can do? To avoid testing a probably infinite list, we should instead understand what are the basic cognitive abilities of humans that compose all other cognitive abilities we have, if thats even possible. Like the equivalent of a turing machine, but for human cognition. The Turing machine is based on a finite list of mechanisms and it is considered as the ultimate computer (in the classical sense of computing, but with potentially infinite memory). But we know too little about whether the limits of the turing machine are also limits of human cognition.
I wonder if we'll get something like NP Complete for AGI, as in a set of problems that humans can solve, or that common problems can be simplified down/converted to.
But we know too little about whether the limits of the turing machine are also limits of human cognition.
Erm, no. Humans can manually step interpreters of Turing-complete languages so we're TC ourselves. There is no more powerful class of computation, we can compute any computable function and our silicon computers can do it as well (given infinite time and scratch space yadayada theoretical wibbles)
The question isn't "whether", the answer to that is "yes of course", the question is first and foremost "what" and then "how", as in "is it fast and efficient enough".
As with many things, it’s hard to pinpoint the exact moment when narrow AI or pre-AGI transitions into true AGI. However, the definition is clear enough that we can confidently look at something like ChatGPT and say it’s not AGI - nor is it anywhere close. There’s likely a gray area between narrow AI and true AGI where it’s difficult to judge whether what we have qualifies, but once we truly reach AGI, I think it will be undeniable.
I doubt it will remain at "human level" for long. Even if it were no more intelligent than humans, it would still process information millions of times faster, possess near-infinite memory, and have access to all existing information. A system like this would almost certainly be so obviously superintelligent that there would be no question about whether it qualifies as AGI.
I think this is similar to the discussion about when a fetus becomes a person. It may not be possible to pinpoint a specific moment, but we can still look at an embryo and confidently say that it’s not a person, just as we can look at a newborn baby and say that it definitely is. In this analogy, the embryo is ChatGPT, and the baby is AGI.
Oh yeah!? If I'm so dang smart why am I not generating 100 billion dollars in value?
So then how do we define natural general intelligence? I'd argue it's when something can do better than chance at solving a task without prior training data particular to that task. Like if a person plays tetris for the first time, maybe they don't do very well but they probably do better than a random set of button inputs.
Likewise with AGI - say you feed an LLM text about the rules of tetris but no button presses/actual game data and then hook it up to play the game. Will it do significantly better than chance? My guess is no but it would be interesting to try.
Any or every task?
It should be able to perform any cognitive task a human can. We already have AI systems that are better at individual tasks.
That's kind of too broad, though. It's too generic of a description.
The key word here is general friend. We can't define general anymore narrowly, or it would no longer be general.
That's the idea, humans can adapt to a broad range of tasks, so should AGI. Proof of lack of specilization as it were.
This is just so they can announce at some point in the future that they've achieved AGI to the tune of billions in the stock market.
Except that it isn't AGI.
But OpenAI has received more than $13 billion in funding from Microsoft over the years, and that money has come with a strange contractual agreement that OpenAI would stop allowing Microsoft to use any new technology it develops after AGI is achieved
The real motivation is to not be beholden to Microsoft
That's not a bad way of defining it, as far as totally objective definitions go. $100 billion is more than the current net income of all of Microsoft. It's reasonable to expect that an AI which can do that is better than a human being (in fact, better than 228,000 human beings) at everything which matters to Microsoft.
Good observation. Could it be that Microsoft lowers profits by including unnecessary investments like acquisitions?
So it'd take a 100M users to sign up for the $200/mo plan. All it'd take is for the US government to issue vouchers for video generators to encourage everyone to become a YouTuber instead of being unemployed.
If they actually achieve AGI I don't understand what money would even mean anymore. It essentially is just a mechanism for getting people to do things they don't otherwise want to do, if the AI can do it just as well as the human, but for free other than the electricity costs, why the hell would you pay a human to do it?
It's like saving up money, in case of nuclear war. There are a few particular moments in history where the state of the world on the far side of the event is so different to the world on this side of the event that there's no point making any kind of plans based on today systems.
I see what you're saying and I agree that if, for example, we get an AI god then money won't be useful. However, that's not the only possible near-future outcome and if the world as we know it doesn't end then money can be used by AIs to get other AIs to do something they don't otherwise want to do.
So they don't actually have a definition of a AGI they just have a point at which they're going to announce it regardless of if it actually is AGI or not.
Great.
I'm gonna laugh when Skynet comes online, runs the numbers, and find that starvation issues in the country can be solved by feeding the rich to the poor.
Why does OpenAI "have" everything and they just sit on it, instead of writing a paper or something? They have a watermarking solution that could help make the world a better place and get rid of some of the Slop out there... They have a definition of AGI... Yet, they release none of that...
Some people even claim they already have a secret AGI. Or at least ChatGPT 5 sure will be it. I can see how that increases the company's value, and you'd better not tell the truth. But with all the other things, it's just silly not to share anything.
Either they're even more greedy than the Metas and Googles out there, or all the articles and "leaks" are just unsubstantiated hype.
Because OpenAI is anything but open. And they make money selling the idea of AI without actually having AI.
Because they don’t have all the things they claim to claim to have, or it’s with significant caveats. These things are publicised to fuel the hype which attracts investor money. Pretty much the only way they can generate money, since running the business is unsustainable and the next gen hardware did not magically solve this problem.
They don't have AGI. AGI also won't happen for another laege amount of years to come
What they currently have is a bunch of very powerful statistical probability engines that can predict the next word or pixel. That's it.
AGI is a completely different beast to the current LLM flower leaves
You're right. The current LLM approach has some severe limitations. If we ever achieve AGI, it'll probably be something which hasn't been invented yet. Seems most experts also predict it'll take some years and won't happen over night. I don't really agree with the "statistical" part, though. I mean that doesn't rule anything out... I haven't seen any mathematical proof that a statistical predictor can't be AGI or anything... That's just something non-expert people often say... But the current LLMs have other/proper limitations as well.
Plus, I don't have that much use for something that does the homework assignments for me. If we're dreaming about the future anyways: I'm waiting for an android that can load the dishwasher, dust the shelves and do the laundry for me. I think that'd be massively useful.
Does anyone have a real link to the non-stalkerware version of:
https://www.theinformation.com/articles/microsoft-and-openais-secret-agi-definition
-and the only place with the reference this article claims to cite but doesn't quote?
