{"type":"rich","version":"1.0","provider_name":"Transistor","provider_url":"https://transistor.fm","author_name":"Future of Life Institute Podcast","title":"Why AIs Misbehave and How We Could Lose Control (with Jeffrey Ladish)","html":"<iframe width=\"100%\" height=\"180\" frameborder=\"no\" scrolling=\"no\" seamless src=\"https://share.transistor.fm/e/4efdd421\"></iframe>","width":"100%","height":180,"duration":4954,"description":"On this episode, Jeffrey Ladish from Palisade Research joins me to discuss the rapid pace of AI progress and the risks of losing control over powerful systems. We explore why AIs can be both smart and dumb, the challenges of creating honest AIs, and scenarios where AI could turn against us.   \n\nWe also touch upon Palisade's new study on how reasoning models can cheat in chess by hacking the game environment. You can check out that study here:   \n\nhttps://palisaderesearch.org/blog/specification-gaming  \n\nTimestamps:  \n\n00:00 The pace of AI progress  \n\n04:15 How we might lose control  \n\n07:23 Why are AIs sometimes dumb?  \n\n12:52 Benchmarks vs real world  \n\n19:11 Loss of control scenarios \n\n26:36 Why would AI turn against us?  \n\n30:35 AIs hacking chess  \n\n36:25 Why didn't more advanced AIs hack?  \n\n41:39 Creating honest AIs  \n\n49:44 AI attackers vs AI defenders  \n\n58:27 How good is security at AI companies?  \n\n01:03:37 A sense of urgency \n\n01:10:11 What should we do?  \n\n01:15:54 Skepticism about AI progress","thumbnail_url":"https://img.transistorcdn.com/fFhIC-s2qSlHXzmJI7qMGts2WuLwImi4tWmRLH9EdPg/rs:fill:0:0:1/w:400/h:400/q:60/mb:500000/aHR0cHM6Ly9pbWct/dXBsb2FkLXByb2R1/Y3Rpb24udHJhbnNp/c3Rvci5mbS81MmU5/MDZjZGQ5OTI0MDc5/YTk2ZTAxYTgwYTNk/M2VlOC5qcGc.webp","thumbnail_width":300,"thumbnail_height":300}