On Monday 4 May 2026 I had the privilege of giving a talk to the Edinburgh Humanists on AI and society. The brief was a 30-40 minute talk followed by Q&A in a one-hour slot.
The title was Artificial Intelligence: From Hype and Fear to Hope and Possibility. The slides are here, and all the materials — speaker notes, image credits, and the iterative drafting history — live in a public GitHub repo.
The blurb
Since ChatGPT’s release in late 2022, billions of people have met chatbots. After the first wave of hype, disillusionment set in: hallucinations, sycophancy, confidently-wrong answers. The public mood cooled — even as the share prices climbed.
This talk argues two things. First: AI is older and stranger than ChatGPT, and its quieter builders — DeepMind, Anthropic — have always been more interesting than its loudest names. Second: underlying capabilities have kept improving through the disillusionment, and a new arrangement called agentic AI is starting to deliver what people first hoped — and feared — that ChatGPT would.
What follows from that is harder. Drawing on Polanyi, Kahneman and Iain M. Banks, the talk closes on what cheap cognition does to work, inequality, science and our sense of ourselves — and on a quieter, more hopeful AI imaginary that the public conversation has neglected.
Links
- Slides
- Talk repo (speaker notes, drafts, image credits)
- Digital Dogs in Hats — the longer companion piece
- Edinburgh Humanists