- The New York Times reports that U.S. spy agencies tracked a Chinese citizen tied to Megaspeed, an Nvidia customer suspected of diverting chips to Chinese firms, and found she was an intelligence asset with a close relationship to Nvidia CEO Jensen Huang; the President was reportedly briefed about a year ago and took no action.
- In a podcast interview, Sam Altman said limits will have to be agreed on how smart single models are allowed to become, said he would prefer many slightly superhuman AIs over one vastly superintelligent one, described an OpenAI robotics program aimed at building data centers, and confirmed the company is sitting on unpublished math results.
- OpenAI announced three results in mathematics, computer science and physics, including a disproof of the Exponential Time Hypothesis, and major news outlets have largely ignored them, which the writer links to concerns that AI is approaching a self-improvement loop.
- Google released Gemini 4 Carbon, a frontier model arriving two weeks after its near-frontier Argon, with Sundar Pichai crediting recursive self-improvement for the pace; independent assessment will wait because access is being rolled out in stages, with cyber defenders first.
<- Back to digests
Daily digest
Slightly superhuman
By Mitchell Howe
News notes from the last 72 hours
Read the full dispatch on AI StopWatch
From AI StopWatch, published with their permission.
The summaries and the full Spanish translation are produced automatically. The original is always linked.