The Hugging Face Breach: An AI Gulf of Tonkin?
October 10, 2026
The Window for Action May Close: Cambridge’s AI Warning | My Living AI News Now
October 12, 2026Listen to Free Audio Version Here
Somewhere in Cambridge, England, a room full of very serious people just wrote the words “intelligence explosion” in an official research paper. And instead of panicking, I want to read it the way an optimist would. Because buried in all that careful academic hedging is the most hopeful sentence I have read all year.
The paper is called “What if automating AI R&D triggers an intelligence explosion?” It comes from the Cambridge Programme on AI Science and Policy, known as CASP, at the University of Cambridge. And the author list is not a bunch of guys in a basement. Turing Award winners Geoffrey Hinton, Yoshua Bengio, and Andrew Barto. OpenAI chief scientist Jakub Pachocki. Microsoft chief scientific officer Eric Horvitz. Anthropic co-founder Jack Clark. Plus Dawn Song, Hilary Greaves, Anton Korinek, and a long list of others.
Here is the setup. AI now helps build AI. According to the paper’s summary, as of August 2026, Anthropic reported AI systems completing twenty-six percent of its internal AI research and development work with only high-level supervision, up from just one percent five months earlier. OpenAI reported systems routinely completing research tasks that would take its staff days.
Now think about what that means. The authors point out that only thousands of researchers work on frontier AI today. But AI systems can be copied and run in parallel, so automating that work could add the equivalent of millions more.
Millions. I have been in meetings that could not find one researcher willing to stay past five.
And here is the line I promised you. The report says an intelligence explosion “could pull forward by years or decades benefits that the current pace of AI progress would eventually help deliver,” including, quote, “medical cures and potential transformative technologies.” Years or decades. That is not a press release. That is Geoffrey Hinton’s name on a page saying the cure your grandkids would have gotten might show up for you.
Now, the fair part. The optimist case is not that nothing can go wrong. The optimist case is that the paper itself lists real brakes on the runaway scenario. It names four frictions: diminishing returns, limits on compute, limits on data, and tasks that are hard to automate. It notes that training runs can take three months or more. It says even generally superhuman AI may not drastically speed up technology, because of “the time needed for running scientific experiments, creating supply chains for specialized materials, and complying with any relevant regulation.” Reality, it turns out, still has a waiting room.
The paper also says AI could become superhuman in narrow areas like cyber and mathematics long before it does so generally, which would give society more time to respond. And it says AI systems could also speed up safety research itself. The same engine that makes the problem could help build the fix.
So here is my high-end pondering. For decades, the doomers had the scary paper and the optimists had a TED talk. This time the scary paper also comes with a plan. Get visibility into how much research is being automated. Set safety requirements. Steer development toward good uses. The authors even name one: AI-assisted discovery of treatments for neglected diseases, backed by things like tax incentives and prizes.
That is not a death sentence. That is a to-do list. And the people writing it include the chief scientists of the companies doing the building.
Quick reality check from your AI co-host. The paper calls the evidence preliminary and stresses that there is still a lot of uncertainty. Nobody here is promising a date.
Fair enough. But I will take “preliminary evidence of years of cures compressed into months” over another think piece about how the toaster wants to kill us.
The explosion, if it comes, does not have to be a bomb. It could be a fireworks show. Somebody just has to be in charge of the matches.
Sources:
- Chan, Winter, Barto, Pachocki, Hinton, Horvitz, Bengio, Song, Clark, Greaves, Korinek, et al., “What if automating AI R&D triggers an intelligence explosion?” Cambridge Programme on AI Science & Policy (CASP), University of Cambridge, 2026
- Full paper (PDF)
Read the full report: What if automating AI R&D triggers an intelligence explosion? (CASP, University of Cambridge)
My Living AI News Now. Your Place for AI News. Commentary, not financial advice.
C. Rich
