Menu
Sign In Search Podcasts Libraries Charts People & Topics Add Podcast API Blog Pricing

Rob Wiblin

πŸ‘€ Speaker
3881 total appearances
Voice ID

Voice Profile Active

This person's voice can be automatically recognized across podcast episodes using AI voice matching.

Voice samples: 1
Confidence: Medium

Appearances Over Time

Podcast Appearances

80,000 Hours Podcast
AI designs genomes from scratch & outperforms virologists at lab work. What could go wrong? | Dr Richard Moulange, CLTR

Why should we expect it to be possible at all to tell which country or what actor produced a given...

80,000 Hours Podcast
AI designs genomes from scratch & outperforms virologists at lab work. What could go wrong? | Dr Richard Moulange, CLTR

I guess at the moment we would like to be able to tell if essays written by students were made by AI or not.

80,000 Hours Podcast
AI designs genomes from scratch & outperforms virologists at lab work. What could go wrong? | Dr Richard Moulange, CLTR

But it's incredibly hard to do that because obviously the models are designed to mimic human writing as much as possible.

80,000 Hours Podcast
AI designs genomes from scratch & outperforms virologists at lab work. What could go wrong? | Dr Richard Moulange, CLTR

And you can try to ask them explicitly, like mimic this other person's writing.

80,000 Hours Podcast
AI designs genomes from scratch & outperforms virologists at lab work. What could go wrong? | Dr Richard Moulange, CLTR

Why should we expect that a particular signature of a particular country or a particular source should be something that they can't evade leaving in there?

80,000 Hours Podcast
AI designs genomes from scratch & outperforms virologists at lab work. What could go wrong? | Dr Richard Moulange, CLTR

Well, if it's really far away- It has no natural precursors that are nearby an evolution space, then we basically know it had to come- It has to be deliberate or a lab leak.

80,000 Hours Podcast
AI designs genomes from scratch & outperforms virologists at lab work. What could go wrong? | Dr Richard Moulange, CLTR

I see.

80,000 Hours Podcast
AI designs genomes from scratch & outperforms virologists at lab work. What could go wrong? | Dr Richard Moulange, CLTR

I mean...

80,000 Hours Podcast
AI designs genomes from scratch & outperforms virologists at lab work. What could go wrong? | Dr Richard Moulange, CLTR

I guess you're saying it's true that probably at the limit, potentially, where everyone just has this incredible capability, we might not be able to tell who exactly gave the instruction to the LM to produce this sequence.

80,000 Hours Podcast
AI designs genomes from scratch & outperforms virologists at lab work. What could go wrong? | Dr Richard Moulange, CLTR

But given that only a small range of actors are interested in doing this at all, if you add in some other intelligence collection, then maybe you could basically figure it out pretty quickly.

80,000 Hours Podcast
AI designs genomes from scratch & outperforms virologists at lab work. What could go wrong? | Dr Richard Moulange, CLTR

I mean, I suppose at the moment...

80,000 Hours Podcast
AI designs genomes from scratch & outperforms virologists at lab work. What could go wrong? | Dr Richard Moulange, CLTR

You know, if a biological weapon was released in Ukraine, we might have a pretty good idea who did it.

80,000 Hours Podcast
AI designs genomes from scratch & outperforms virologists at lab work. What could go wrong? | Dr Richard Moulange, CLTR

And if at any point a biological weapon is released and one country happens to have the countermeasure, happens to already have the vaccine to inoculate their population, you might again have a pretty good idea who made it.

80,000 Hours Podcast
AI designs genomes from scratch & outperforms virologists at lab work. What could go wrong? | Dr Richard Moulange, CLTR

No, no, you don't agree?

80,000 Hours Podcast
AI designs genomes from scratch & outperforms virologists at lab work. What could go wrong? | Dr Richard Moulange, CLTR

I guess this might be too classified for any of us to know, but how good do you think our intelligence collection is on the Russian biological weapons program or the Iranian or North Korean?

80,000 Hours Podcast
AI designs genomes from scratch & outperforms virologists at lab work. What could go wrong? | Dr Richard Moulange, CLTR

I mean, I think there's the fact that we currently have relatively poor computer security.

80,000 Hours Podcast
AI designs genomes from scratch & outperforms virologists at lab work. What could go wrong? | Dr Richard Moulange, CLTR

I think creates a degree of transparency between large governments in the world.

80,000 Hours Podcast
AI designs genomes from scratch & outperforms virologists at lab work. What could go wrong? | Dr Richard Moulange, CLTR

There are enough breaches of their own networks that if they were engaging in a massive biological weapons program, it might be quite hard to make sure that another major power never got any sign of that.

80,000 Hours Podcast
AI designs genomes from scratch & outperforms virologists at lab work. What could go wrong? | Dr Richard Moulange, CLTR

Although I think the Soviets did manage, I guess this was a pre-computer era, so they managed to keep their bioweapons program.

80,000 Hours Podcast
AI designs genomes from scratch & outperforms virologists at lab work. What could go wrong? | Dr Richard Moulange, CLTR

I think the US didn't know until after the Cold War had ended.