TFS#246 - Chris Szegedy on Co-Founding xAI with Elon Musk & the Future of Truth-Seeking AI

Christian Szegedy is a renowned AI researcher and entrepreneur recognised for his contributions to deep learning. He spent over 12 years at Google, advancing large-scale AI systems in computer vision, deep learning, and formal reasoning, and became recognized for his pioneering contributions to adversarial machine learning.Later, he co-founded xAI with Elon Musk and is now Chief Scientist at Morph Labs and founder of Math Incorporated. He focuses on verified superintelligence, aiming to bring mathematical rigor and formal verification to AI for greater reliability, safety, and trust.https://www.linkedin.com/in/christian-szegedy-bb284816https://x.com/ChrSzegedyhttps://www.math.inc/CHAPTERS:0:00 – Introduction1:20 – Meet Christian Szegedy2:32 – Why Chris left xAI and the idea of “verified superintelligence”3:07 – Using AI to formally verify mathematical proofs5:41 – Why Chris spun out of Morph Labs to start Math Incorporated6:48 – How new Math Incorporated is and what stage the company is at8:09 – How verified AI could impact everyday AI use10:07 – Multi-step AI verification workflow11:56 – How ChatGPT decides what a “good” response is12:42 – Was Chris always this technical and math-focused?13:40 – Chris' family background and his mathematically gifted brothers14:16 – Chris talks about his son, who loves mathematics at a young age15:00 – How he teaches his son about math and coding without overusing AI15:57 – What happens when humans cognitively offload everything to AI18:03 – Will AI eliminate jobs and lead to universal basic income?20:20 – Career advice for people in their 30s in an AI-driven economy22:55 – Are people becoming allergic to AI-generated content?24:22 – Adversarial AI: How to verify whether content is real or AI-generated25:39 – How the Pope used AI to generate a tweet about AI26:45 – Why Chris is deeply passionate about formal verification27:29 – Formal verification in simple terms32:21 – Is AI smart enough today to reason from axioms?34:38 – Why formal verification exists but isn’t widely adopted36:43 – The biggest bottlenecks slowing automated verification38:48 – What’s limiting AI from verifying math papers instantly?40:40 – Current team at Math Incorporated41:31 – Chris' hiring philosophy and working with young talent43:24 – Co-founding xAI with Elon Musk45:50 – Why meetings with Elon Musk were so long48:06 – How technically deep Elon Musk really is49:22 – Key lessons Chris learned working closely with Elon50:59 – Elon Musk's goal for xAI51:32 – Should people still learn to code in the age of AI?52:45 – The best programming languages to start with today53:55 – Learning just enough code to avoid being “blind”55:35 – Using AI to automate podcast clip distribution56:18 – How Chris personally uses AI on a day-to-day basis58:11 – Where Grok AI scrapes their data59:09 – What most people misunderstand about the next 5 yrs of AI1:00:35 – AI integration into the real world and robotics1:03:06 – Formal vs. informal AI and truth-seeking systems1:04:08 – Can formal AI help prevent unsafe or deceptive AI?1:05:58 – Chris' thoughts on Elon Musk’s goal for the most truth-seeking AI1:06:33 – Chris shares some specifications he is pushing right now1:07:18 – The limits of formalizing concepts like “cat detection”1:11:17 – What is chip and chip verification?1:13:39 – What was the first chip ever made?1:16:02 – How logic gates shrank from calculators to iPhones1:17:23 – Chris shares where the world’s most advanced chips are made1:19:38 – Chris talks about living in America vs. Europe and returning to Hungary1:20:20 – Why Chris hasn’t started a company with his brothers1:21:22 – Will AI be winner-takes-all or stay competitive?1:22:23 – How formal AI competes with informal reasoning models1:22:58 – Chris talks about DeepMind by Google1:24:30 – Chris' recent life discoveries1:25:10 – Chris’ personal goals for the next 6 months1:25:35 – Connect with Chris1:26:18 – Outro