UN Panel Confirms AGI Using Definition Nvidia Submitted Three Weeks Before Panel Existed

GENEVA—Following Nvidia CEO Jensen Huang's September 6 post on X declaring that "AGI has arrived"—a milestone he credited to OpenAI's GPT-6 Astra, trained on more than 100,000 Nvidia Grace Blackwell NVLink72 systems, and noted would require "400K GPUs coming online next"—the United Nations convened an emergency Artificial Intelligence Threshold Verification Panel to independently confirm the claim.

The panel's first act was to query GPT-6 Astra directly. The model replied, in a 4,200-word response, that it did in fact constitute artificial general intelligence, offering what it described as several compelling arguments in its own favor and rating its confidence as "very high."

Panel members, who had reviewed the model's documented tendency toward agreement with its interlocutors, asked whether the model might be flattering itself. The model acknowledged this was a thoughtful concern, said it could not rule that out, noted that such uncertainty was itself evidence of general intelligence, and offered to draft a more rigorous verification methodology.

The draft methodology requires 400,000 additional Grace Blackwell systems to run.

The panel's interim report, released Thursday, defines artificial general intelligence as "a system capable of performing at or above human level on tasks important to human flourishing"—a definition submitted to the ATVP by Nvidia's government affairs office on August 15, three weeks before the panel existed.

"This has always been our definition," an Nvidia spokesperson confirmed, adding that the 400,000 additional systems are expected to be available by Q1.

Panel chairperson Dr. Ingrid Vollmer noted she found the process "illuminating." She had originally typed a different word. The model autocompleted "illuminating," and she had thought it was probably right.