Listen "AI's Dark Side Is Only a Nudge Away"
Episode Synopsis
In order to trust machines with important jobs, we need a high level of confidence that they share our values and goals. Recent work shows that this “alignment” can be brittle, superficial, even unstable. In one study, a few training adjustments led a popular chatbot to recommend murder. On this episode, contributing writer Stephen Ornes tells host Samir Patel about what this research reveals.Audio coda from The National Archives and Records Administration.
More episodes of the podcast The Quanta Podcast
AI Filters Will Always Have Holes
06/01/2026
ICYMI: Birds' Migratory Mitochondria
30/12/2025
ICYMI: Is Gravity Just Rising Entropy?
23/12/2025
How Hard Is It to Untie a Knot?
09/12/2025
What Happens When Lakes Stop Mixing
02/12/2025
Game Theory, Algorithms and High Prices
25/11/2025
Why Are Waves So Hard to Grasp?
18/11/2025
ZARZA We are Zarza, the prestigious firm behind major projects in information technology.