NEW YORK CITY (PIX11) -- Former AI researchers painted a bleak picture of an industry already far out of control from rogue AI agents to a huge lack of human understanding.
New York City Council on Monday convened the first-of-its-kind government hearing on AI oversight, including testimony from whistleblowers and high-ranking representatives of OpenAI, Google, Meta and Anthropic.
Discover more local reporting on our homepage
Advertisement
The whistleblowers started the hearing off, setting the stakes very high.
"I myself would guess AI takeover chances at roughly one in three. This logic may sound sci-fi, but it is a mainstream scientific concern that has been discussed for many years," said Alex Turner, a former Google DeepMind researcher. The City Council issued a subpoena for Turner's testimony.
The other two whistleblowers, former OpenAI researcher Daniel Kokotajlo and former Anthropic researcher Jacob Coxton, shared similar concerns about a lack of human control over AI development. All three pointed to what they see as the key concern: "recursive self-improvement."
It's a process in which AI systems become responsible for the coding, training and maintenance of itself and new systems. The whistleblowers said companies are all racing to perfect this model.
The transition is already underway, according to Kokotajlo and a written testimony he shared from current OpenAI researcher Dan Selsam. Researchers already rely on AI to do much of the coding and act more like a...
Hospitals will likely spend more to protect themselves from the Centers for Medicare and Medicaid Services’ expanded powers to kick them out of government health programs, which would be ruinous f...