Greg BakerGuest
Greta ThomasHost
Claire HattonHost
Well, as you know, AI changed very rapidly in the 2010s, particularly in the 2020s.



GPT-3 was around before GPT-3.5, and a number of my colleagues were using it very extensively.

I'm not sure if it's a site you're aware of, but if you want to go and look up any piece of trivia about any pop culture, that's the place to go.

Because it's fan-generated content, it's difficult to know what everything's about.

And so we wanted to map, this is about a Star Trek character in the original series, or this is about a Star Wars character or a Star Wars planet.

So I wrote this very, very elegant piece of artificial intelligence code that clearly explained its arguments, how it was reaching its conclusions, did very careful analysis of text.

And of course, his worked better than mine, except for the problem of, well, why did it say that? And that meant we've thrown away the ability to understand and take responsibility for an outcome.

And we've just delegated, well, we just abrogated all responsibility and handed it over to something else.

And that's because of that sort of the black box nature of generative AI in particular, I guess, where you really don't know what sources it's using and how it comes up.

And similarly, you can ask it one question repeatedly and you'll get a different answer each time in terms of construction.

Well, as you know, AI changed very rapidly in the 2010s, particularly in the 2020s.



GPT-3 was around before GPT-3.5, and a number of my colleagues were using it very extensively.

I'm not sure if it's a site you're aware of, but if you want to go and look up any piece of trivia about any pop culture, that's the place to go.

Because it's fan-generated content, it's difficult to know what everything's about.

And so we wanted to map, this is about a Star Trek character in the original series, or this is about a Star Wars character or a Star Wars planet.

So I wrote this very, very elegant piece of artificial intelligence code that clearly explained its arguments, how it was reaching its conclusions, did very careful analysis of text.

And of course, his worked better than mine, except for the problem of, well, why did it say that? And that meant we've thrown away the ability to understand and take responsibility for an outcome.

And we've just delegated, well, we just abrogated all responsibility and handed it over to something else.

And that's because of that sort of the black box nature of generative AI in particular, I guess, where you really don't know what sources it's using and how it comes up.

And similarly, you can ask it one question repeatedly and you'll get a different answer each time in terms of construction.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.