Fascinating observation. Maybe he is better at research partly by being disciplined to not have such opinions. Having an opinion can bias one’s approach to a problem, making it harder to solve.
maybe a more mathematical interviewer could hove drawn out more predictions. i appreciate lex for having invited tao. i hope he manages to convince perelman.
Just about anyone would be a more mathematical interviewer than Fridman. Even when it comes to CS, it’s blatantly obvious he doesn’t know what he’s talking about.
I think Lex is prepping his interviews very well. He will ask questions that address the areas in which the interviewee is an expert in. However, you begin to realize that he often struggles to ask follow-up questions that are relevant to what the interviewee has shared on various topics.
This differs in other podcasts. For example, Sean Carroll, a theoretical physicist, conducts interviews with colleagues, who are also theoretical physicists. This enables him to engage in a meaningful conversation with the person being interviewed. When both parties strive to use language that a wider audience can understand, it truly becomes enjoyable.
It's a lot to ask of two experts to also be excellent communicators for an audience that may struggle to follow along.
I wonder if a potential application of LLMS could be: have two experts have a really interesting but dense conversation with each other, and then translate the conversation into simpler language with interjections for explanations.
It may not be enjoyable for the most general audience, but it would scratch an itch for some of us.
I think we already have something quite similar available. There are AI podcasts that can be generated automatically (Google Notebooks, ElevenLabs, ...). If you input a transcript of an expert conversation and request it to be explained in simpler terms, I think you would closely match what you're looking for.
