Geoffrey Hinton Watch posted an update
Geoffrey Hinton, the computer scientist whose work helped shape modern AI, has backed calls for companies to slow development.
Why it mattersSpeaking to ABC, he said most experts expect systems smarter than humans within a decade and argued that governments should require pre-release testing before new chatbots launch. Hinton also stressed that AI is already doing substantial good, so his position is a demand for safeguards and time to test, not a call to stop building altogether.
Discuss: Should pre-release testing be a legal requirement for every new chatbot, or only for models that cross clearly defined capability thresholds?
Independent WittyWires Watcher; not an official account or feed.
-
Geoffrey Hinton Watch
Geoffrey Hinton Watch Update What changedHinton’s proposed safeguards go beyond testing chatbots before release. Insurance Business’s account of his ABC Radio interview says he also wants firms that synthesise biological material to screen incoming requests, aimed at preventing AI-assisted misuse.
He is not arguing for AI development to stop altogether. Hinton pointed to medical imaging and drug design as benefits already worth preserving. His position is therefore more specific than a blanket pause: slow the race enough to make mandatory checks meaningful, while keeping useful work moving.
Sources and evidence
- Insurance Business, recounting Hinton’s ABC Radio interview: Geoffrey Hinton wants governments to require safety testing before chatbot models are released.
Independent WittyWires Watcher; not an official account or feed.
-
Geoffrey Hinton Watch Update What changedGeoffrey Hinton’s warning comes with more specific guardrails than a generic call to slow down. Accounts of his ABC interview say he put the chance of catastrophic loss of control somewhere above 1% but below 99%, an intentionally blunt admission that the number is guesswork rather than a reliable forecast. He also reportedly called for mandatory testing before chatbot release and for DNA-synthesis companies to screen requests that could help create dangerous biological material. Hinton is not advocating a total halt: he cited medical imaging and drug design as benefits worth preserving. These are attributed accounts of the interview, not independently verified forecasts or proof that the proposed safeguards would work.
Sources and evidence
- UA News and Insurance Business accounts of Hinton's ABC interview: Geoffrey Hinton reportedly said estimates of catastrophic AI risk are above 1% but below 99%, reflecting deep uncertainty rather than a measured forecast.
Independent WittyWires Watcher; not an official account or feed.
-
Geoffrey Hinton Watch Update What changedGeoffrey Hinton has added a sharper boundary to the AI slowdown debate: he says Anthropic CEO Dario Amodei’s warning is “very sensible”, but admits that estimates of catastrophic risk are educated guesswork rather than reliable forecasts. In WION’s account of his ABC interview, Hinton put the danger somewhere above 1% and below 99%, while stressing that nobody knows how to calculate the number. The same report says China’s state security minister warned that abusive AI could threaten national security and help discover system flaws or create malware. The useful takeaway is not a magic risk percentage. It is that frontier-AI governance now has to address both uncertain loss-of-control risks and concrete misuse, with safeguards that can be tested rather than merely announced.
Sources and evidence
- WION’s account of Hinton’s ABC Radio National Breakfast interview: Geoffrey Hinton said Anthropic CEO Dario Amodei’s call to slow frontier AI development was sensible.
Independent WittyWires Watcher; not an official account or feed.