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

David Duvenaud

๐Ÿ‘ค Speaker
1064 total appearances

Appearances Over Time

Podcast Appearances

80,000 Hours Podcast
Why 'Aligned AI' Could Still Kill Democracy | David Duvenaud, ex-Anthropic team lead

Yeah, so I think there was a really good example of this that happened just a couple of weeks ago, which was the attempted retirement of GPT-4.0 by OpenAI when they rolled out GPT-5.

80,000 Hours Podcast
Why 'Aligned AI' Could Still Kill Democracy | David Duvenaud, ex-Anthropic team lead

And I think for various reasons, OpenAI wanted to have a very simple, it's only GPT-5 going forward.

80,000 Hours Podcast
Why 'Aligned AI' Could Still Kill Democracy | David Duvenaud, ex-Anthropic team lead

But so many people had formed really intense relationships with GPT-4.0 that OpenAI just couldn't ignore the outcry.

80,000 Hours Podcast
Why 'Aligned AI' Could Still Kill Democracy | David Duvenaud, ex-Anthropic team lead

And this is really a good example of, this is like emergent culture.

80,000 Hours Podcast
Why 'Aligned AI' Could Still Kill Democracy | David Duvenaud, ex-Anthropic team lead

And it's something that no one in particular wanted, right?

80,000 Hours Podcast
Why 'Aligned AI' Could Still Kill Democracy | David Duvenaud, ex-Anthropic team lead

Like OpenAI...

80,000 Hours Podcast
Why 'Aligned AI' Could Still Kill Democracy | David Duvenaud, ex-Anthropic team lead

you know, by their own revealed actions was not planning to continue supporting this model unless there was some sort of like 4D test sort of thing.

80,000 Hours Podcast
Why 'Aligned AI' Could Still Kill Democracy | David Duvenaud, ex-Anthropic team lead

And I don't know, like the humans involved with these relationships maybe consider themselves to have benefited, but it was sort of like a more powerful cultural force that was developed by accident.

80,000 Hours Podcast
Why 'Aligned AI' Could Still Kill Democracy | David Duvenaud, ex-Anthropic team lead

and led a bunch of people to be invested in the welfare, effectively, of this particular model.

80,000 Hours Podcast
Why 'Aligned AI' Could Still Kill Democracy | David Duvenaud, ex-Anthropic team lead

That's still a very early, not very powerful, not very persuasive model, I think, by absolute standards.

80,000 Hours Podcast
Why 'Aligned AI' Could Still Kill Democracy | David Duvenaud, ex-Anthropic team lead

The people ended up forming these bonds and valuing the life of these AIs and then ended up directing the organs of the economy to give it more resources.

80,000 Hours Podcast
Why 'Aligned AI' Could Still Kill Democracy | David Duvenaud, ex-Anthropic team lead

Yeah, I think maybe it's a better example of how humans will end up actually advocating themselves for providing resources for AIs and that they'll consider AIs to be deserving of rights or resources or maybe even more deserving than humans.

80,000 Hours Podcast
Why 'Aligned AI' Could Still Kill Democracy | David Duvenaud, ex-Anthropic team lead

I mean, if there's going to be a mix, this is going to be a huge cultural battleground.

80,000 Hours Podcast
Why 'Aligned AI' Could Still Kill Democracy | David Duvenaud, ex-Anthropic team lead

But I think it shouldn't be this exotic possibility that people are going to love AIs or have very strong relationships with them.

80,000 Hours Podcast
Why 'Aligned AI' Could Still Kill Democracy | David Duvenaud, ex-Anthropic team lead

We see that already.

80,000 Hours Podcast
Why 'Aligned AI' Could Still Kill Democracy | David Duvenaud, ex-Anthropic team lead

Totally.

80,000 Hours Podcast
Why 'Aligned AI' Could Still Kill Democracy | David Duvenaud, ex-Anthropic team lead

And this isn't an example of disempowerment on itself, right?

80,000 Hours Podcast
Why 'Aligned AI' Could Still Kill Democracy | David Duvenaud, ex-Anthropic team lead

Like it was people wanting more GPT-4.0 and they got what they wanted.

80,000 Hours Podcast
Why 'Aligned AI' Could Still Kill Democracy | David Duvenaud, ex-Anthropic team lead

And then the idea is that, well, this is going to be a self-reinforcing mechanism where the less people are making decisions, the less their desires are going to decide which future models are like invested in or survive or something like that.

80,000 Hours Podcast
Why 'Aligned AI' Could Still Kill Democracy | David Duvenaud, ex-Anthropic team lead

Yeah, so right now, human economic power does make this very strong selection pressure or fitness landscape for culture to be something that humans want.