Fin is a Research Fellow at Forethought, an organisation that researches how we might thrive under extremely fast AI progress. As part of his work, he also hosts the Forethought podcast ForeCast, where you can hear more about the ideas we’ll discuss today. Previously, he co-hosted the Hear This Idea podcast, worked at Longview Philanthropy, the Future of Humanity Institute, and much more. You can find out Fin’s whole professional history, along with just about everything he’s read and listened to via his handsomely designed website.

We talk to Fin mostly about three things:

* The Better Futures series — where he argues, with Will MacAskill, that it may be as good, or better, to work on increasing the value of the future, rather than increasing the odds humanity survives.

* The intelligence explosion, the idea that rapidly improving AI could accelerate technological and social progress.

* And finally, Fin’s recent essay, no ghost in the machine, arguing for a non-realist view of consciousness, which I (Toby) would highly recommend reading.

Subscribe to get the transcript, with links (and a graph!) in the next post.

Steven Adler is a former safety researcher at OpenAI, the frontier lab behind ChatGPT. Steven worked there for four years on safety and policy, including leading the Dangerous Capability Evaluations team. He now writes for his Substack, Clear‑Eyed AI, about the development of very capable systems and how to make AI go better.

In this episode, Steven walks through the risks surrounding advanced artificial intelligence: (1) concerning behaviours we already see in current models; (2) plausible future risks; and (3) concrete ways people, technical or not, can work on the problem. I really enjoyed this conversation and I’m excited for everyone to listen!