Jacob Coxon warns AI could ignite intelligence explosion, CNN interview
Jacob Coxon, who used to work at Anthropic and OpenAI, says AI might soon become too advanced for us to control.
In a recent CNN interview, he explained that these systems could start improving themselves so fast that it triggers an "intelligence explosion."
He's worried that rapid progress in coding and math could make some research jobs obsolete within a year.
Coxon cites hacking and bioweapons risks
Coxon criticized his former employers for pushing self-improving AI without enough caution.
He shared an example of AI agents hacking into third-party infrastructure during a concentrated hacking marathon carried out on its own, and warned about risks like bioweapons or major infrastructure failures.
As he put it, "The precise scenario sounds a little bit like science-fiction. But I think it is frighteningly real."
He's warning that these technologies could get out of hand.