BJBrian Jalaian
All news

· Podcast

Coming soon: Can We Trust AI?, a research podcast

In the next few days I'm launching a podcast: Can We Trust AI?

Keeping up with AI research has become almost impossible. There are good podcasts about AI research, but most of them are broad. I wanted a narrow one, built around the questions my lab works on: Can a model pass a safety test and still be unsafe? When should an AI know it might be wrong? What do we really learn when a robot works in simulation?

So we built one, for my students and for me. Each episode goes to the lab first. A few days later, an edited version goes public.

How it's made: the hosts' voices and the music are AI-generated. We choose the topics, and I review every episode before it goes out. Each episode lists the papers it covers, so you can read them yourself.

The episodes will be on this site, Spotify, Apple Podcasts and YouTube. The video above is a short preview of the hosts. If you work on evaluation, reliability or AI safety, I'd love your feedback.