-
safety
2026-06-01
At the Sana AI Summit in New York: we are making new beings and must ensure they care about us.
«We're making these beings, and we should be very concerned to make them beings that care about us.»
-
safety
2025-08-12
↺ changement de position
Shift: instead of controlling superintelligence, proposed building 'maternal instincts' — we need AI mothers, not assistants.
«We need AI mothers rather than AI assistants. You can't fire your mother.»
-
safety
2025-04-26
Risk metaphor: AI is like a cute tiger cub that may kill you once grown; urged spending a third of compute on safety.
«Unless you can be very sure it's not gonna want to kill you when it's grown up, you should worry.»
-
safety
2024-12-27
On BBC Radio 4 put the chance of AI causing human extinction within 30 years at 10-20%, up from 10%.
«Not really, 10 to 20 [per cent].»
-
safety
2024-10-08
After the Nobel Prize in Physics warned no one knows whether we can control systems more intelligent than us.
«when we get things more intelligent than ourselves, no one knows whether we can control them.»
-
safety
2023-10-08
On 60 Minutes said he can see no path that guarantees safety; concedes AI might take over.
«I can't see a path that guarantees safety.»
-
safety
2023-05-03
Sharp existential framing: humanity may be just a passing phase in the evolution of intelligence.
«humanity is just a passing phase in the evolution of intelligence.»
-
safety
2023-05-01
↺ changement de position
Left Google to speak freely about AI dangers; admitted he underestimated how fast AI would progress.
«I left so that I could talk about the dangers of AI without considering how this impacts Google.»
