Time To Start Panicking About AI?

In this episode, Henry and I finally do something we probably should have done in the first episode: introduce ourselves. We talk about our backgrounds in philosophy, how we became interested in psychology and cognitive science, and what drew us to thinking about AI. From there, we dig into the current state of AI capabilities, especially “agentic” AI (e.g., Claude Code), the politics of AI (including the Trump administration's recent conflict with Anthropic), and whether the growing public hostility to AI is well-founded or misdirected. We wrap up with a big question: is it time to start panicking about AI? Henry says the time to panic was five years ago. I argue that for panic or any other emotion to be productive, it must be anchored in an accurate, evidence-based understanding of what is happening, which is missing from lots of the current discourse about AI. Links * Dan Williams, The Mind as a Predictive Modelling Engine: Generative Models, Structural Similarity, and Mental Representation (PhD thesis, University of Cambridge, 2018). * Dan Williams, “Socially Adaptive Belief” (2021)* Henry Shevlin, “Three Frameworks for AI Mentality” (2026) * Henry Shevlin, “A Lack of Understanding: Storytelling for Robots” (2019) — Litro Magazine. * Lake et al, “Building Machines That Learn and Think Like People” (2017) * Matt Shumer, “Something Big Is Happening” (2026)* Leopold Aschenbrenner, Situational Awareness: The Decade Ahead (2024) * Joseph Heath, “Highbrow Climate Misinformation” (2025) * Dean Ball* Ethan Mollick * Leopold Aschenbrenner Transcript(Note that this transcript is AI-edited and may contain minor mistakes).Introducing OurselvesDan: Welcome back. I’m Dan Williams, and I’m back with Henry Shevlin. Today we’re going to be discussing some questions about the nature of AI as it’s developed over the past couple of months. We’re also going to be talking about the politics of AI and probably some questions about AI and public opinion — some of the backlash that appears to be brewing among certain segments of the public when it comes to AI.But to kick things off, we’re going to do something we probably should have done in the first episode but haven’t actually done yet, which is to introduce ourselves. So Henry, to begin with — who are you?Henry: So many different descriptors I could choose from. I think I’ll start with philosopher of cognitive science. I’m also a father, husband, son, D&D player, big video gamer, runner, cyclist — all that good stuff. But let me talk a little more about the philosopher of cognitive science side.I’m the associate director at the Leverhulme Centre for the Future of Intelligence, Cambridge’s main AI ethics, theory, policy, and law research centre. Basically, everything except building the models. We do practical benchmarking work on capabilities, legal reviews, sociology and critical theory of AI — it’s a really big interdisciplinary centre. I’ve been there now going on nine years. I joined early 2017, all the way back when state-of-the-art AI was stuff like AlphaGo. We were created just as that story was brewing. In 2016, AlphaGo won a very surprising victory against Lee Sedol in the game of Go, which was seen by many as an almost impossible challenge for AI because of its combinatorial complexity.It’s been amazing working in this role — having these front row seats to what I think is a unique period, not just in the history of AI, but in the history of human civilisation. In the last nine years, it really was like having a front seat in Lancashire during the Industrial Revolution, watching the development of various industrial applications.Dan: Yeah.Henry: Before we get more into AI, maybe a little more background. I’m from the UK, originally from Staffordshire. I was a