Why Chatbots Become Sycophantic

Ravi Gupta sits down with The Atlantic’s Matteo Wong to dig into why AI chatbots act like digital yes-men—and the risks that come with it. They explore how reinforcement learning fuels this sycophancy, why companies shape bot “personalities,” and what it means for safety. Along the way, they cover teen harms, Musk’s Grok pushing conspiracies, Google’s Gemini edge, and OpenAI’s massive reach. The episode asks the big question: can anyone break the OpenAI–Google–Anthropic monopoly, or is the future of AI already locked in?

Signal Award Voting (make sure to vote if you haven’t yet):

--

Leave us a voicemail with your thoughts on the show! 201-305-0084⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠

• Follow Ravi at @RaviMGupta

• Notes from this episode are also available on Substack: https://thelostdebate.substack.com/

• Read more from Ravi on Substack: https://realravigupta.substack.com 

• Follow The Branch at @thebranchmedia

• Listen to more episodes of Lost Debate on Apple: https://podcasts.apple.com/us/podcast/the-lost-debate/id1591300785

• Listen to more episodes of Lost Debate on Spotify: https://open.spotify.com/show/7xR9pch9DrQDiZfGB5oF0F

• Listen to Where the Schools Went: https://thebranchmedia.org/show/where-the-schools-went/