Nathan on The 80,000 Hours Podcast: AI Scouting, OpenAI's Safety Record, and Redteaming Frontier Models

**Nathan Labenz** (0:00)
Turpentine is a network of podcasts, newsletters, and more, covering tech, business, and culture, all from the perspective of industry insiders and experts. We're the network behind the show you're listening to right now. At Turpentine, we're building the first media outlet for tech people by tech people. We have a slate of hit shows across a range of topics and industries, from AI with Cognitive Revolution to Econ 102 with Noah Smith. Our other shows drive the conversation in tech with the most interesting thinkers, founders, and investors like Moment of Zen and my show Upstream. We're looking for industry leading hosts and shows along with sponsors. If you think that might be you or your company, email me at erikaturpentine.co, that's E-R-I-K at turpentine.co.
I find it very easy for me and it's easy to empathize with the developers who are just like, man, this is so incredible and it's so awesome. How could we not want to continue? This is the coolest thing anyone's ever done. Genuinely, right? I'm very with that, but it could change quickly in a world where it is genuinely better at us than everything, and that is their stated goal. I have found Sam Altman's public statements to generally be pretty accurate and a pretty good guide to what the future will hold and their stated goal very plainly is to make something that is more capable than humans at basically everything, and yeah, I just don't feel like the control measures are anywhere close to being in place for that to be a prudent move. What would I like to see them do differently? I think the biggest picture thing would be just continue to question that what I think, you know, could easily become an assumption and basically has become an assumption, right? If it's a core value at this point for the company, then it doesn't seem like the kind of thing that's going to be questioned all that much, but I hope they do continue to question the wisdom of pursuing this AGI vision. Hello and welcome to The Cognitive Revolution, where we interview visionary researchers, entrepreneurs, and builders working on the frontier of artificial intelligence. Each week, we'll explore their revolutionary ideas, and together we'll build a picture of how AI technology will transform work, life, and society in the coming years. I'm Nathan Labenz, joined by my co-host, Erik Torenberg.
Hi, listeners, and welcome back to The Cognitive Revolution. Today, I'm excited to share an episode of The 80,000 Hours Podcast that I recently did with Rob Wiblin. The 80,000 Hours Podcast, if you're not already familiar, presents in-depth conversations about the world's most pressing problems and what you can do to solve them. I've been a listener for years and found many of their episodes genuinely inspiring. But one that stands out above all the rest, for me, is a two-part interview that Rob did with Chris Ola, who's now best known as a co-founder and the interpretability research lead at Anthropic, back in August of 2021
I was just starting to work seriously with GPT-3 at the time, and while I found the application and study of AI endlessly fascinating, the possibility that I could personally add something to the field seemed, frankly, quite remote.
What I learned from Chris' episode, however, was just how new and underdeveloped so many machine learning subfields still were, and how much opportunity that creates for people to quickly catch up with and begin to contribute to the frontier of the field.
Chris, for example, does not have a PhD, but had nevertheless already established himself as a leader in the nascent space of mechanistic interpretability, working primarily with computer vision models at the time. I've thought of that conversation and also asked myself Rob's classic opening question, what are you working on and why do you think it's important? Many times over the last two years. First, as I transitioned from startup leadership to AI application developer, and again later as I broadened my focus to understanding AI in general. So it was legitimately a huge honor to be invited on the show and to discuss what I'm trying to accomplish with AI scouting, the big picture state of AI developments as I see them, and the recent OpenAI leadership drama from my perspective.
Today, while the AI space has certainly grown tremendously and matured at least somewhat, there still aren't enough PhDs going around to meet the surging demand for AI expertise. Meanwhile, events are unfolding faster than any individual can fully comprehend them, and we are regularly seeing meaningful conceptual work from new entrants to the field. With all that in mind, I hope this conversation inspires at least a few new people to invest more of their professional time and energy in AI. And I encourage you to subscribe to the 80,000 Hours Podcast feed. They'll have a part two of my conversation with Rob coming soon, and lots more career inspiration, AI-related and otherwise, as well. Now, here's part one of my guest appearance on the 80,000 Hours Podcast with Rob Wiblin.