AGI Has Been Achieved Internally

🌈 Abstract

The article discusses the implications and potential consequences of someone, particularly Sam Altman of OpenAI, publicly claiming that AGI (Artificial General Intelligence) has been achieved internally. It explores the various reactions and debates that could arise from such a claim, as well as the challenges in verifying the claim and its impact on the world.

🙋 Q&A

[01] "AGI has been achieved internally"

1. What was the context around the "AGI has been achieved internally" statement?

• The statement was first tweeted by an account called "Jimmy Apples" on September 18th, and was called out as "bullshit".
• A week later, Sam Altman, the CEO of OpenAI, echoed the same statement in a Reddit comment, but later edited it to reveal he was just "memeing".
• The article suggests that while Altman's statement was a joke this time, there are indications that the next time he makes a similar claim, it may not be for the memes.

2. Why is the statement powerful, even if it's not true?

• If the statement comes from someone like Sam Altman, the CEO of a leading AI lab like OpenAI, it can have a significant impact, even if it's not true.
• Reactions can range from annoyance that the CEO would tease something so serious, to worry about the possibility of it being true, to excitement at the prospect of witnessing a historic event.

3. What are the implications if Altman were to make this claim in a more serious venue, without any hints of retraction?

• The article suggests that this scenario is likely to happen soon, and it would be a pivotal moment that would be "engraved in our memory".
• It would raise questions about the ability to verify the claim, the reactions from various stakeholders, and the impact on the world if the claim is accepted or rejected.

[02] Verifying the claim of AGI achievement

1. Who is most likely to make the claim of AGI achievement, and why?

• The article predicts that Sam Altman, as the visible head of OpenAI, the company best positioned to achieve AGI, is the most likely person to make such a claim.

2. What kind of evidence would OpenAI need to provide to make the claim credible?

• The article suggests that OpenAI would need to provide definitive proof, such as a technical report with results that are not immediately refutable, even by veteran AI researchers without access to the system.

3. What are the challenges in verifying the claim of AGI achievement?

• The article notes that it's likely OpenAI would not allow for independent external auditing of their system, making it difficult for others to verify the claim.
• Without the ability to independently verify the claim, there may not be a consensus on whether AGI has truly been achieved.

[03] Implications of the claim of AGI achievement

1. What are the different perspectives and reactions that could arise from the claim of AGI achievement?

• The article suggests that the claim would be met with strong reactions from insiders, competitors, and governments, and there may not be an immediate consensus on the validity of the claim.
• The public and those with the power to enforce the claim, such as governments, would have to grapple with the political, social, economic, and philosophical implications of the claim.

2. How does the article suggest we should redefine our AGI timelines?

• The article proposes that we should redefine our AGI timelines not for when we think AGI will be achieved, but for when we think it will be first claimed to have been achieved by a sufficiently influential person, in possession of strong enough evidence.

3. What is the key point the article makes about the impact of beliefs versus objective truths?

• The article suggests that the real effect will be the consequences of our beliefs, rather than the objective truths that ideally back them up. For any measure of impact on the world, whether AGI is real is arguably less important than whether people believe it is.