Li-Lian Ang
speaker
308 appearances
1 recordings
1 series
first heard Apr 2026
last heard 2 Apr
Li-Lian Ang’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsRecordings per month over the last 12 months — 1 in all, peaking in Apr 2026 with 1.
Appearances
Where given that these two layers have lots of like quite big holes in them where things can just kind of still slide through, can we make sure that we have good layers at the end that kind of prevent the worst things from happening still?
Yeah, so the first layer is to prevent the training of dangerous AI systems.
The second is to be able to detect if these systems are
The second layer is to be able to detect if these systems have, in fact, still taken dangerous actions anyway, despite our efforts to try to train it out of them.
And then the third layer is to be able to withstand any kind of dangerous actions that we weren't able to stop in the first two layers.
Yeah.
So, like, earlier we talked a lot about, like, if all the development of AI systems were constrained to these, like, five companies, if we could control these five companies, then, you know, maybe things would be fine.
But then, you know, like, open source, open weight models, all these things add, like, an additional complexity to this because you don't even know, like, how powerful a person's model is because they have, like, you know, like...
there's a proliferation of these around.
And one of the proposals that I've heard around this is like,
in a more like decentralized fashion, can we like ensure that all the open weights models, like the people who maintain these open weights models are all like bought in on the idea that we need to do these kinds of like safeguard mechanisms on these.
And we have people who like go around and like test all the open weight models that are available and make sure that, you know, they like follow these kind of like guidelines and that they're always safe.
able to be safe.
But obviously, this is not a robust method at all.
And this is why I also think that as much as we try to minimize the risk in this area, we're still just going to need to build out the rest of the layers of defense.
Because it's just unrealistic to believe that we could ever get 100% guarantee on these earlier funds.
We can try to reduce it as much as possible.
which is why people try to do methods like trying to remove specific capabilities from AI systems, from doing cyber attacks, or to be able to build...
unlearning mechanisms to remove particular knowledge from them or to build it into their personalities such that they would never take these actions because they have it encoded in them that this is morally wrong or something but they're not 100% but they're helpful to a degree
I feel like the
Showing 221–240 of 308 · page 12 of 16
← Previous
Next →