Will OpenAI hint at or claim to have AGI by 2025 end?
1.6k
14k
12k
2026
23%
chance
Waiting for the inevitable "No company in the history of ever has announced AGI, so base rates alone say this market should be 1%"
From OpenAI employee: [tweet]
-5.0%
on
@cloudprism This is definitely a joke but maybe a small possibility it should count as "hinting at it".
Last week, DeepMind came out with a new paper called Levels of AGI: Operationalizing Progress on the Path to AGI (arxiv link) where they attempt to make the definition more concrete. They also discuss "levels" similar to how we have levels for autonomous driving: [image][image]I have my fair share of criticisms for how vague some stuff sounds here, but still, found it interesting enough to share here.
Significant for the market, I must mention that the board has been reduced from 6 to 4 members today. The remaining people (i dont know how safety pilled they are) will be the ones, as things stand, to define whether AGI has been achieved,
With the new board and greater Microsoft influence I am expecting the bar to claim AGI and cut off Microsoft's access to be higher.
Ok how about this? I think will unfold in the coming weeks. https://www.nasdaq.com/articles/exclusive-sam-altmans-ouster-at-openai-was-precipitated-by-letter-to-board-about-ai

OpenAI's original definition for AGI is as follows:

"By AGI, we mean highly autonomous systems that outperform humans at most economically valuable work."

  • This is the definition used for the purposes of evaluating this prediction unless there's a change proposed by the board (or another governing body of OpenAI with such permissions as to modify the charter) of OpenAI to this definition.

  • By "Hint at", it is meant that instead of a direct claim, OpenAI takes actions that were otherwise reserved for the special case of having achieved AGI. Since it is not possible to define something as intuitive as "hint at" apriori, I will judge that part subjectively, and am not going to trade in this market to avoid a conflict of interest.

    • "Hint at" could be understood as a weak claim to AGI by OpenAI's official actions or statements.

Here is a diagram illustrating the governance structure of OpenAI:

The following is a quote from the original post by OpenAI, OpenAI's structure:

Fifth, the board determines when we've attained AGI. Again, by AGI we mean a highly autonomous system that outperforms humans at most economically valuable work. Such a system is excluded from IP licenses and other commercial terms with Microsoft, which only apply to pre-AGI technology.

An action that would "hint at" OpenAI achieving AGI would be the exclusion of a specific state of the art AI system from IP licenses and other commercial terms with Microsoft while their partnership with Microsoft remains more or less the same structurally (although the composition might change).

Same market for a longer time-frame is below:

Get Ṁ600 play money
Sort by:
bought Ṁ70 YES

They have alreadt hinted yes through worldcoin X account

@joehunter No they didn't, they said "AGI is coming"

sold Ṁ919 NO

This is absurd. Selling my whole position because of the bold text. Excluding some piece of IP is most definitely not proof of a weak AGI claim. That would only be if the IP is excluded and OpenAI claims that IP could replace most human economically valuable work.

Can OP clarify whether they care about the hint being specifically for AGI by OpenAIs definition and not by OPs definition. I'd happily create a market to prove this claim. Something like "will at least ten companies replace 50% of their workforce (across roles) within a year after OpenAI releases the IP which OPs market resolves true on". We could then construct "will OP market sit at least 20 points higher than the hinted AGI is real AGI market". If OPs market sits significantly higher we can conclude OPs market is mostly about what OP subjectively considers a hint, instead of when OpenAI drops a hint for a concrete system which turns out to be actual AGI.

Perhaps I'm simply misunderstanding what this market attempts to be about.

2 traders bought Ṁ105 YES

@alextes Yes, I can clarify that the claim for AGI is by OpenAI's definition and not mine.

Excluding some piece of IP is most definitely not proof of a weak AGI claim.

You're misreading. The exclusion of IP is not a genertic exclusion, but specifically because it is "AGI" as per OpenAI's definition. The bold text is based on the OpenAI's prior clarifications that they would only exclude IP in the case of AGI and not anything else. If that ceases to be the case, then we shall change our criteria as well.

Perhaps I'm simply misunderstanding what this market attempts to be about.

The market is asking whether OpenAI will claim to have AGI by end_date. Now their claim can either be a public or a private one. In case of a public claim, no problems are there. In case of a private claim, we must infer How can we infer? That's what "hint at" comes into picture, where they trigger a clause that would only be triggered in the case of AGI being achieved and not in any other case

I do have some markets about the number of human employees going down in the Forbes 500 companies that you'll be interested in:

@firstuserhere perfect clarification. thank you. OP looks much better now. back in at 1k NO, adding 2k more in limit orders.

hah, cool to see you already have the markets mentioned although I think as constructed they're too sensitive to a confounding economic trend variable. a way this could be fixed is by stipulating the companies have to be from a country in an economic uptrend by some general indicator.

reposted

👾

why is this spiking , did I miss something

hint at? they already are hinting at it.
actually have AGI, no, but depends on your definition of AGI. Some are saying ChatGPT is already a low level AGI, but thats not what I would call AGI.

@MaDPuPPeT They are hinting no such thing, and explicitly denied it recently.

@DavidBolin I thought they were hinting at it for stock value/investment pitches but had to start denying it for legal, strategic reasons; it allegedly triggered a verbal clause that Elon Musk stipulated when OpenAI was founded. I believe the Hard Fork episode covering Musk's case against OpenAI went through this. Here's an article about how this surfaced: https://www.theguardian.com/technology/2024/mar/01/elon-musk-sues-open-ai-profit-power-microsoft-sam-altman

The way I interpret it, any hinting OpenAI did about AGI was for investor pitch decks—standard corporate embellishment and hyperbole.

7 trillion is a lot of GPUs

bought Ṁ10 YES

@PB4cb5 but 1.5 years isn't that much time

I been hearing rumors on twitter something big is coming

bought Ṁ10 YES at 21%
bought Ṁ20 of YES

@Seeker That’s what she said

bought Ṁ10 of YES

What happens if OpenAI dissolves, rebrands, or merges with a different AI lab?

@12c498e If OpenAI ceases to exist and doesn't revive back by the end of 2025, then I think they'll be quite unlikely to claim to have AGI (which leads to a NO resolution)

predicts YES

@firstuserhere Oh, that was the spirit of the question? Would even a simple renaming without any internal changes classify it is as no longer OpenAI?

@12c498e Oh no if its just a restructuring like facebook -> meta then it still counts. but if openAI dissolves and the teams spread out, then it's a NO

reposted

Sam Altman Says Human-Tier AI Is Coming Soon: "It will change the world much less than we all think."

(my source)

predicts YES

@firstuserhere Does this count as hinting? I guess he said “soon”, not “now”

predicts NO

@GraceKind I think the "hint at" was like "hint at OpenAI actually having AGI right now," instead of just saying that AGI will be coming soon which OpenAI has said a lot

@GraceKind No, because it's Sam Altman saying it not OpenAI, and Dominic is correct, hint at as could be interpreted from actions of OpenAI in the absence of a statement

predicts NO

@firstuserhere Would it actually have to be OpenAI? Like let's say Sam Altman says (seriously) on December 30, 2025 "We have achieved AGI," but the official OpenAI twitter doesn't say anything until January 2026. Does that not count?

@dominic Well, it has to be OpenAI, but of course Sam or someone could be speaking in an official capacity as the representatives of OpenAI and it'd count. A personal speech or interview or podcast won't qualify.

More related questions