Evidence receipt / belief
Published · transcript-backedAravind Srinivas: belief
19 Jun 2024 Lex Fridman Podcast #434 – Aravind Srinivas: Perplexity CEO on Future of AI, Search & the Internet
“I think another dimension is, we are just deeply curious as a species, and I think we have… Some work in AI’s, have explored this curiosity driven exploration.”
Source trail
Everything needed to verify it.
- Speaker
- Aravind Srinivas
- Attribution
- Verified speaker
- Claim type
- belief
- Recorded
- 19 Jun 2024
- Publisher
- Lex Fridman Podcast
Transcript context
…Yeah, that’s one of the missions of the company is to cater to human curiosity. And it surfaces this fundamental question is like, where does that curiosity come from? Exactly. It’s not well understood. And I also think it’s what makes us really special. I know you talk a lot about this. What makes human special is love, natural beauty to how we live and things like that. I think another dimension is, we are just deeply curious as a species, and I think we have… Some work in AI’s, have explored this curiosity driven exploration. A Berkeley professor, Alyosha Efros’ written some papers on this where in our rail, what happens if you just don’t have any reward signal? And agent just explores based on prediction errors. He showed that you can even complete a whole Mario game or a level, by literally just being curious. Because games are designed that way by the designer to keep leading you to new things. But that’s just works at the game level and nothing has been done to really mimic real human curiosity. So I feel like even in a world where you call that an AGI, if you feel like you can have a conversation with an AI scientist at the level of Feynman, even in such a world, I don’t think there’s any indication to me that we can mimic Feynman’s curiosity. We could mimic Feynman’s ability to thoroughly research something, and come up with non-trivial answers to something. But can we mimic his natural curiosity about just his period of just being naturally curious about so many different things? And endeavoring to try to understand the right question, or seek explanations for the right question? It’s not clear to me yet. It feels like the process the Perplexity is doing where you ask a question and you answer it and then you go on to the next related question, and this chain of questions. That feels like that could be instilled into AI just constantly searching-…
Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.