The Promise and Limitations of AI
summary
Argues that hybrid AI outperforms pure deep learning by combining statistical nets with a logic engine (Cyc/“PSYC”) that contains tens of millions of common‑sense rules, about 1.5 million general terms and ~43,000 relations, and a planner of 1,100 inference engines. Shows live Cyc reasoning on Romeo and Juliet beliefs, a MathCraft tutoring app that diagnoses students’ stepwise errors, a Cleveland Clinic clinical‑query app built from ~100–120k input rules, and an Accenture SXSW personalization demo.
AI-written summary · about
chapters (21)
- 0:00Intro
- 1:36Two different power sources
- 3:56Right brain brittleness
- 5:44Left brain brittleness
- 6:33Eliza
- 8:22The Wisdom of the Crowd
- 9:19The Power of Statistics
- 11:08Common Sense Heuristics
- 17:26Strengths Weaknesses
- 19:08MathCraft
- 22:55Situations that havent happened
- 23:59ML and symbolic logic
- 26:01We want plausible answers
- 27:06Multiple experts
- 30:42Summary
- 32:51National Cooperative Research Act
- 36:40Left brain thinking
- 42:22Limited logics
- 45:55Tradeoff curve
- 53:12Causal reasoning
- 56:30Conclusion
related talks
21mLearning from History: The Past as a Compass for Systems Engineers in the Age of AI
01h 30mTrustworthy AI & Open Source
07mCan hybrid intelligence make AI more trustworthy?
48mArtificial Intelligence. Our journey towards singularity
56mAI: It AI-n't What You Think!
42mReasonable AI
19mRedefining Intelligence: How Reasoning Is Re-Shaping AI in 2025
13mArtificial General Intelligence: Why aren't we there yet?