Future of Life Institute Podcast Why AI Is Not a Normal Technology (with Peter Wildeford)
100 snips
Apr 29, 2026 
Peter Wildeford, Head of Policy at the AI Policy Network and leading AI forecaster, explains why AI is neither a bubble nor a normal technology. He discusses forecasting AI progress, economic and employment timing, adoption gaps and power users, rising cyber and military risks, export controls, and the evolving role of prediction markets.
AI Snips
Chapters
Transcript
Episode notes
Expect Noticeable Employment Effects Within A Year
- Wildeford expects employment effects to remain hard to detect now but to become noticeable within a year, potentially adding around one percentage point of unemployment.
- His diffusion model predicts small current signal that could grow rapidly as capabilities spread.
Adoption Lag Masks AI's Economic Impact
- Current economic indicators lag because powerful models aren't yet widely used at full capability in workplaces.
- Wildeford notes many users stick to basic tasks (proofreading), while top adopters using cloud code are in the top 0.1% and drive outsized productivity gains.
Integrate AI Into Defense While Prioritizing Control
- Governments should both integrate AI into militaries and develop secure, controllable deployments to retain advantage over adversaries.
- Wildeford urges the US to lead on control, understanding, and hardening against theft or adversarial tampering of frontier models.