Investor revisits his 2019 AI predictions: mostly right, divergent futures ahead
sebkrier · x · 2026-10-09
Krishnan reviews his 2019 predictions, arguing most held up well: 1) Scale works for general intelligence, one architecture eats everything; 2) Scale learns any pattern inherent in the data, including patterns of learning; 3) AI training grows but hits Moravec-style paradoxes; 4) AI Zeno's-paradoxes its way through most jobs without replacing them all; 5) Slow capability takeoff since linear gains need exponential investment; 6) Anything AlphaZero-able sees superhuman output but doesn't generalise; 7) Kurzweil-style neuron-compute scaling helps reach superintelligent digital minds. His key point: all of these predicted today's world, yet they imply different — even competing — futures.
More from AGI Musings
- Are the weights conscious? Anthropic's stance sparks playful probing of AI consciousness criteria — akbirthko · 2026-10-09
- Pretrain Falls to 7% of Compute as Yudkowsky's Doom Scenario Regains Ground — DKokotajlo · 2026-10-09
- Community Effort Boosts OpenAI Math Proof Result 500,000-Fold, Tightening κ to 2⁻¹⁵ — eldonredwards · 2026-10-09
- Meta's Alexandr Wang: Execution, Not IQ, Is the Real Moat at the Top — arthurcolle · 2026-10-09
- Gradio founder: autonomous agents are breaking the incentives to share science openly — Gradio · 2026-10-09
- DeepMind essay: what 15M Gemini interactions reveal about AI-accelerated science — JMateosGarcia · 2026-10-09