Rob Wiblin
speaker
1,787 appearances
3 recordings
1 series
first heard Sep 2024
last heard May 2025
Rob Wiblin’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsNo recordings in the last 12 months.Older appearances are listed below; set an alert to hear about the next one.
Appearances
Uh things keep looking misaligned.
Yeah.
So you mentioned that uh people are adopting strategies for measuring alignment that are not very adversarially robust.
Um
Like how yeah, in in what specific way do you think that they would like fail to um to detect that it's it that a model is uh open to scheming?
So
Yeah, okay.
I guess we've been assuming this entire time that AIs can effectively randomize and th they they can know whether uh th th they can implement the strategy of only defecting like once every millionth uh times.
Can they do that?
not right now.
I see.
Um and do you think trying to trying to make that impossible is probably just not what not one of the best control mechanisms that you could you could implement?
I I don't
effort.
Okay.
What sort of I guess empirical evidence that you might get over the next year or two would make you more inclined to like double down on the AI control agenda versus think, oh, this isn't as promising as I as I feel right now?
Yeah.
So I would say the
risk posed by those AIs.
Uh and I guess might you want to double down on making sure that that remains the case, that we don't have them, you know, beginning to reason in code or
Showing 401–420 of 1,787 · page 21 of 90
← Previous
Next →