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
So we've got a thing where, you know, you can destroy style somewhat by uh jittering it, changing it into a different language or whatever.
But if it's the core message of the thing that it will definitely be transmitted and and maintained, then like then then you're in trouble.
Oh, okay.
Okay.
So this is a mechanism that like maybe you think it like halves the I guess if you think it's unclear, maybe it like halves the halves the risk on average.
How how useful is halving halving the risk?
Is this a case where muddling through and kind of like, you know, and making it more difficult for for the for the AI to out with you here and there is is actually quite useful incrementally, or do we have to win or do we have to succeed overwhelmingly uh or or or like we basically lose the game?
Yeah.
I mean,
Got it.
Okay.
So this like hinges quite centrally on whether if we catch the model, we are gonna do things differently next time, not not trust the model, not deploy it as readily, uh come up with things that would stop it.
Is that true?
Well, uh I mean I didn't say you have to stop deploying
for
Yeah, let's okay.
So I ma I guess l let's imagine the worst case where you do feel an enormous amount of pressure to continue using the model.
Um in in that case, like how useful is it to catch them red handed?
I guess other people who were incredibly skeptical to begin with, they thought there was a one in a thousand chance.
I guess maybe they'll update people have gotten wrecked by the by the evidence, yes.
Showing 221–240 of 1,787 · page 12 of 90
← Previous
Next →