The Most Dangerous AI Hallucinations Are Hard to Spot | Dream Machines @KQEDnews
The Most Dangerous AI Hallucinations Are Hard to Spot | Dream Machines  @KQEDnews
Uploaded August 2026 | Updated November 2026, 4 days ago
AI hallucinations used to be obvious: a fake paper, a made-up source, a fact that was clearly wrong. But as AI systems improve, Jungwon Byun says their mistakes are becoming much harder to detect.

Alexis Madrigal and Robin Sloan talk with Byun about why an AI can cite a real research paper and still misrepresent what it says—and why subtle errors may be more dangerous than the wild hallucinations of earlier models.

🎙 Hosts: Alexis Madrigal, Robin Sloan
🎙 Guest: Jungwon Byun
📌 Show: Dream Machines
👍 Subscribe for more conversations about AI, technology and the future.
The Most Dangerous AI Hallucinations Are Hard to Spot | Dream MachinesThe Truth About John Henry, The Steel-Driving ManKatie Porters Plans To Lower Costs And Regulate Silicon Valley | Town HallJane Kim Wants to Make Nixon’s Healthcare Idea a Reality | Political BreakdownResidents Celebrate Día de los Muertos in San Francisco’s Mission DistrictChad Bianco on Enforcing Laws He Disagrees With | Political BreakdownCan Democrats Flip the Central Valley? | Political BreakdownThe California Governor’s Race Ultimate Debrief | Political BreakdownWhy Swalwell Victim Hesitated to Come Forward | Political BreakdownFans Cheer On the Valkyries  For Game 6Election Results You Need to KnowWhy Californians Struggle to Get Insurance Help | Political Breakdown
KQED News |

The Most Dangerous AI Hallucinations Are Hard to Spot | Dream Machines

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER