The unconscious parts of intelligence ft. Joe Cole - Tacit Intelligence

Joe Cole, Co-Founder and CEO of Tacit Intelligence, joins The Lab to explore what elite sport can teach us about artificial intelligence. Joe explains how Tacit captures the hidden judgment of human experts, trains specialised AI models and evaluates them for high-stakes environments. The conversation covers reinforcement learning, sleep, AI safety, tacit knowledge, digital twins, model memory and the future of human expertise. Learn more about Tacit Intelligence: https://www.tacitintelligence.co/ Check us out on Spotify and Apple: https://open.spotify.com/show/033KAkE... https://podcasts.apple.com/us/podcast... CHAPTERS: 00:00 Preview 00:22 Episode introduction 00:49 Joe Cole and elite sport 02:38 Applying performance science to AI 05:16 How learning really happens 08:08 Sleep, recovery and performance 10:46 A human-centred approach to improvement 17:55 Building safer AI systems 22:08 Why LLM outputs can vary dramatically 24:38 Capturing tacit knowledge 28:14 Specialist models versus frontier models 31:33 How far can AI continue improving? 36:19 Could AI models dream? 46:17 How Tacit trains and evaluates models 51:03 Should AI become superhuman? 57:26 Business applications and expert abundance 1:09:24 AI mentors and the context problem 1:14:34 The valuable data companies aren’t capturing 1:19:47 Turning AI history into transferable expertise 1:23:09 Agents, skills, linting and LLM judges 1:26:38 Can AI memory hold you back? 1:34:41 Why negative prompts can backfire 1:39:59 Closing thoughts