Red-Teaming after Mythos — Zico Kolter & Matt Fredrikson, Gray Swan

Okay, we're here in the studio with Grace Wan, Matt and Zico.

It goes beyond the title with direct discussion of like, right, think, including: Thanks for having us.

1. 01So really artificial intelligence, large language models are, at the end of the day, software
2. 02If you want to sort of deploy them, build applications on top of them, you need to be sort of aware of what, you know, what the vulnerabilities might be, what can go wrong
3. 03and not just in sort of everyday use, like you're kind of innocently using an agent and, you know, maybe it makes a mistake in a tool call, but also, you know, in worst case kinds