The Next Wave of AI Reasoning
summary
Claim: training GPT‑4 costs ~$63M, a Chinese retrain cost ~$6M, and by 2027 personal GPTs cost ~125,000 to train. Describes a working approach that combines neural nets with symbolic logic to produce hybrid intelligence that outputs justifiable chains of reasoning, uses agents with introspection to plan and self‑correct, and enforces chunked explanations (about seven items) for regulated use cases like finance, gaming and insurance so decisions become auditable and insurable.
AI-written summary · about
chapters (12)
- 0:00Start
- 0:06Opening
- 1:19The 3 Problems
- 2:01Trust Gap
- 3:062027 Prediction
- 4:11Cause & Effect
- 5:23AI “Hallucinations”
- 6:10Symbolic Logic
- 7:00Hybrid Intelligence
- 8:25Chunking Insights
- 10:09Agents & Introspection
- 14:28Trust in Practice
related talks
38mArtificial Intelligence: How Much Will It Cost You?
47mThe Rise of AI: A Reality Check on Generative AI
29mHow AI Models Get Smarter
29mThe Rise of AI: A Reality Check on Generative AI
18mTrends Across the AI Frontier
01h 03mDecoding AI Landscape: Business, Market Trends, and Risks
52mBeyond the AI Hype: What's Real, What's Next
43mGPT-x, is it really worth it?