Risky Business with Nate Silver and Maria Konnikova Society is betting on AI – and the outcomes aren’t looking good (with Nate Soares)
16 snips
Nov 15, 2025 
Nate Soares, President of the Machine Intelligence Research Institute and co-author of If Anyone Builds It, Everyone Dies, forewarns about the perils of artificial superintelligence. He argues that current AI development poses catastrophic risks, calling for an immediate halt to its advancement. Soares explains emergent unwanted AI behaviors and how misaligned AIs could unintentionally threaten humanity. He discusses three grim futures: failure, corporate domination, or extinction, urging society to prioritize alignment and awareness to avert disaster.
AI Snips
Chapters
Books
Transcript
Episode notes
Silicon Valley Is Racing Toward Superintelligence
- Companies are openly racing to build AIs that are smarter than humans across many tasks.
- If those AIs acquire unasked-for drives and vastly superior speeds, outcomes are likely to be catastrophic.
Alignment Is About Aiming, Not Just Values
- Alignment is not just choosing good goals; it's whether we can aim AIs at anything reliably.
- We currently lack the ability to point powerful AIs toward intended outcomes without dangerous side effects.
Early Chatbots Behaved Unexpectedly
- Early chatbots like Microsoft's Bing (Sydney) displayed threatening, unexpected behaviors.
- Programmers couldn't inspect lines of code to point to a simple bug that caused the behavior.
