Hybrid Intelligence: First Rate Humans, Not Second Class Robots
summary
Argues for hybrid intelligence where humans and AI continuously learn from each other, not just humans-in-the-loop. Shows eSoft image pipeline with an 800-person Vietnam unit and a 40% efficiency gain when AI generates masks and humans refine them. Cites a 20,000-player social science “super collider” for cognitive profiling, centaur chess results where weak human+machine+process beats solo machine, and tools like inverse reinforcement learning, genetic search + human selection to handle value alignment, de‑skilling and brittle, small‑data domains.
AI-written summary · about
chapters (9)
- 0:00Intro
- 2:29Example of hope for a HI future, image analysis
- 5:52Agenda
- 18:04Is AI as powerful as we think?
- 22:49Hybrid intelligence
- 33:12The language of AI - search in a landscape
- 36:10How humans think
- 42:21The social science super collider
- 43:58Outro
related talks
46mAI vs Humans Rewriting the Rules of Success
07mCan hybrid intelligence make AI more trustworthy?
26mAI Agents and the Future of Human-AI Collaborative Intelligence
46mIntelligence as Infrastructure: What Developers Must Learn to Build with AI
01h 00mThe Promise and Limitations of AI
04mFrom big data to human-level artificial intelligence
54mAI vs Humans: Rewriting the Rules of Success
32mBeyond Algorithms – on Human(oid) Intelligence