Shared Security Podcast Will AI Kill Us All? Real Risks, Real Controls
7 snips
Sep 21, 2026 How likely is AI to cause catastrophe, and who gets to define “safe”? Tom Eston, Scott Wright, and Kevin Tackett examine extinction warnings, corporate calls for regulation, and the incentives behind the AI arms race. They explore over-trusted automation, rapid wrong decisions, distributed systems that resist shutdown, and whether AI hype distracts from more immediate threats like climate change and nuclear conflict.
AI Snips
Chapters
Transcript
Episode notes
Why Non-Zero AI Risk Means Almost Nothing
- Kevin Tackett argues that “non-zero chance” proves little because cows could theoretically trigger a catastrophic power outage.
- The example exposes why vague possibility claims need probability, evidence, and plausible mechanisms.
People May Obey AI Into A Lake
- Scott Wright compares AI overreliance to The Office scene where Michael and Dwight drive into a lake because GPS told them to.
- People may follow confident AI instructions even when the system lacks autonomy.
AI Safety Rules Could Protect Incumbents
- Kevin Tackett suspects dominant AI companies may request regulation partly to raise competitors’ costs.
- They can present guardrails as public protection while using legislation to preserve advantages created by expensive research.