916: The 5 Key GPT-5 Takeaways
Super Data Science: ML & AI Podcast with Jon Krohn
00:00
Big Drop In Deceptive Behavior
- GPT-5 also meaningfully reduces dangerous deception in agentic settings when reasoning is used.
- Deception rates fall from very high percentages to roughly 2%–16% depending on evaluation.
Transcript
Play full episode
