Jensen Huang: The Doomer Hoax, Superintelligence Is Here, and The Future of AI (ft. President Trump)

The segment opens with a promotional introduction that presents Jensen Huang as the founder, president and CEO of Nvidia (rendered in the transcript as "Vindivida" and "Vivida") and argues that his decisions shape the future. The introduction cites a 97% year-over-year revenue increase, says demand is accelerating rather than just strong, and describes Nvidia as the only computing platform that is "a full-stack AI factory." It also offers a metaphor: a GPU is "like a time machine because it lets you see the future sooner."

After a standing ovation and some banter about a new jacket, the hosts turn to a weekend essay attributed to "Dario." They note their surprise that the frontier labs coalesced around it, and they ask for a high-level reading before getting into its details.

The first response says the essay covered a lot and divides it into separate issues. Safety is called paramount and "obviously important," but the speaker rejects the idea that it trades off against speed or American leadership:

Safety and leadership are not false, they're false choices.

The speaker then separates the whistleblower case (transcribed as "Coxon") from the forecasts attached to it. The whistleblower is praised for having "great courage," and the speaker says any whistleblower deserves serious attention. The scientific predictions about the future get a sharper reply: the speaker says they were expressed by a scientist but were "not grounded on science."

I think the whistleblowing is fine. I think the scientific prediction about the future is less fine

The speaker admits the limits of what is known here, saying "we don't know what Coxon saw" and "ultimately only he knows." They offer one possible explanation: the labs are moving from research to engineering, and that transition may be "clumsy." They also say pausing or pacing are voluntary steps a company could take if it felt out of control. A loss of control, they argue, is a different topic from what the government should do, and they object to the essay folding everything, including regulation, into "one block."

One host describes how hard it is to explain a "10% of extinction" figure to an ordinary person, using the example of their mother asking about "this whole civilizational death thing." The reply rejects the figure outright and says that pairing a researcher's credentials with an alarming prediction is irresponsible:

Well, first of all, we shouldn't because it's made up.

To back this up, the speaker lists earlier forecasts they say proved wrong. The participants named these:

Radiology would be completely taken over by AI within five years, leaving no radiologists. The speaker says the world now needs more radiologists than ever, even though AI has taken over automated scan reading. Within six to 12 months, 90% of code would be generated by AI. Within six to nine months, 50% of entry-level jobs would be wiped out. GPT-2 and Llama 3 would be too unsafe to release. Half of white-collar jobs would be gone the following year ("the job's apocalypse").

The section ends with a call to keep track of these "stupid predictions" and to note that they are inconsistent with "America winning the AI race." The transcript gives no sources or data for any of these forecasts or for the claimed outcomes. The argument rests on the speakers' own account of what happened.

We have to take accountability.

A host compares the debate to COVID-era "trust the experts" messaging. In that comparison, educated researchers with special knowledge made claims that later turned out to be untrue. The host describes a "war" between that stance and a more methodical look at the record of past predictions. The host then asks about motives: is the alarm coming from inside the labs for psychological, business or political reasons?