
Helen Toner
Executive director, CSET
Worried, week of Nov 8
Capability 4Outlook 2
Where they stand
Against all 37 people scored in Week 46. Bars count people at each score; black is Toner, the dashed line is the average.
What they've said
- Saturday, Nov 14Week 46
On Decoder, Helen Toner said the leading labs are openly trying to reach superintelligence through recursive self-improvement, that getting AI to do what users want is
not on track to be solved
, and thatWhen the people who are building the systems tell us, "Oh, we have to solve this within 12 to 24 months," I think that's terrifying and I hope they're wrong.
The former OpenAI board member, who now runs Georgetown's Center for Security and Emerging Technology, said she would vote to remove Sam Altman again and that
if our hopes for the future are pinned on any one individual being trustworthy, then we're pretty screwed.
She said her own guess is thatwe probably have longer
, but that the problem is whether alignment issolvable in the time that we have.
She floated decelerating the growth of pure capabilities while speeding up uses of today's models.Details and sources
Toner also said the labs
Decoder, The Vergeare increasingly saying that they're seeing acceleration
from using AI in their own research, that giving AI agents more autonomydefinitely makes me nervous
, and that the US and China have over the past one to three months begun to see more in common on loss-of-control risks. Nilay Patel notes the interview was recorded before Nvidia opened its investigation into Jensen Huang.