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.
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.










