Wisdom of Crowds Just How Worried Should We Be About AI?
Feb 15, 2026
Henry Shevlin, a Cambridge philosopher focused on AI ethics and consciousness research, explores agency in modern systems and the line between tools and moral entities. He discusses agentic coding, examples of autonomy, and experiments showing emergent agent behavior. The conversation covers hallucinations, types of creativity in AI, recursive self-improvement, and the need for careful governance.
AI Snips
Chapters
Books
Transcript
Episode notes
AI's Acceleration Is Industrial‑Scale
- Henry Shevlin argues AI's recent progress rivals the industrial revolution in impact and has accelerated faster than expected.
- He sees November 2022 (ChatGPT) as a turning point that dramatically changed capability and public engagement.
Smallville Shows Emergent Role Behavior
- Henry describes the Smallville experiment where many LLM instances played roles and showed emergent, role-driven behavior in a simulated village.
- He uses it to illustrate agency as a spectrum from simple steps to running complex workflows over time.
Agentic Coding Can Democratize Software
- Claude Code and agentic coding tools let non-expert users delegate complex software tasks and bug-checking with minimal supervision.
- Henry finds this delegation capability striking and productivity-transforming even for mediocre coders.
