Artificial General Intelligence, AGI, superintelligence, powerful AI, Nobel-level performance, X, FINRA-style regulator, frontier AI, Bloomberg.

Until recently, Sam Altman and Dario Amodei had for years been the two voices most responsible for putting the term Artificial General Intelligence (AGI) on the map.

They used the term so often, and with such conviction, that it became shorthand for the entire industry’s promise: a machine that could think, reason, and create like a human, only faster and without fatigue.  Investors, regulators, and the public came to expect AGI as the finish line, a shared destination that every major lab was racing toward.

Now, that shared vocabulary is fracturing:

  • Altman has begun calling AGI “a very sloppy term”, swapping it for talk of superintelligence arriving in “a few thousand days”
  • Amodei, meanwhile, has all but dropped the phrase, preferring “powerful AI” and concrete benchmarks tied to Nobel-level performance. His working definition is narrower and more tangible: intellectual capabilities on par with or surpassing those of Nobel laureates across disciplines from biology to engineering

The two men who did more than anyone to make AGI a household acronym are now pulling in opposite directions — not just on what to call the technology, but on how to convince a skeptical world that it is both real and safe.

The lack of agreement on what AGI means is feeding directly into a broader crisis of trust. In an August exchange on X, Amodei had acknowledged that “ordinary people don’t trust companies, governments, or the tech industry and always suspect that we are cooking up some new way to screw them over,” admitted that AI firms “haven’t yet delivered on our big promises to benefit the world,” and voiced support for an independent, FINRA-style regulator for frontier AI. 

As Bloomberg’s Austin Carr noted, for many researchers, AGI has become less a measurable milestone than “a spiritual concept” — a horizon that seems to retreat as quickly as the industry approaches it.

Until a laboratory commits to a public, third-party benchmark with a concrete date attached, every timeline in this debate remains a promise made by those who stand to profit most from it being believed.