Ben Goertzel uses d-calculus to argue self-improving AIs can converge on beneficial superintelligence
bengoertzel · x · 2026-10-04
- AGI pioneer Ben Goertzel published a long, math-heavy article, "Finding the Beneficial-Superintelligence Attractor," the sequel to his piece introducing d-calculus.
- d-calculus is an early-stage mathematical framework for minds that fork, merge, learn and rewrite themselves. Goertzel applies it to his central concern: whether and under what conditions AIs that modify and improve themselves will retain their goals as they evolve.
- His claim: under the right conditions, self-improving AI can fall into the basin of attraction of a "beneficial superintelligence" strange attractor — staying aligned as capability grows. He argues this matters urgently as recursive self-improvement (RSI) becomes a practical engineering reality.
More from AGI Musings
- Professor refuses to read AI-generated cold emails from prospective students — anshulkundaje · 2026-10-04
- The AI timescale lasagna: unpacking the timeline tensions as AI deploys into the economy — _AustinCalvert_ · 2026-10-04
- Aphasia Patients Still Do Algebra: Evidence That Language Isn't Thinking — ctjlewis · 2026-10-04
- Cameron puts 90% odds on AI becoming fully creative without human examples — c_valenzuelab · 2026-10-04
- Intelligence lives in the geometry of information, not the weights, argues thread — yunta_tsai · 2026-10-04
- The Self-Fulfilling Prophecy of AI Control — tingshuo · 2026-10-04