Ben Todd: spark vs software-only singularity is really about which feedback loops run how far
ben_j_todd · x · 2026-11-15
Ben Todd clarifies the spark vs. software-only singularity debate:
- Both scenarios are driven by a software-only feedback loop; hardware loops are additive in both cases.
- Spark (4-6 years of progress in <1 year) is what AI safety folks expected by default — few believe you foom all the way to physical limits.
- The main opposing view: compute bottlenecks and diminishing returns will bind hard enough to prevent much acceleration.
- He prefers framing it as (i) different kinds of feedback loops and (ii) how far each runs before running out of steam.
More from AGI Musings
- AI Ethicist Daniel Faggella Clarifies His Cause: Life, Potentia, and AGI's Moral Weight — danfaggella · 2026-11-15
- 2027 prediction: hourly AI films rivaling $500M epics, and critics gamed by design — morqon · 2026-11-15
- Philosopher Jeff Sebo slams 'confused and shallow' take on AI moral status — connoraxiotes · 2026-11-15
- Rutger Bregman reverses stance: 'We need to stop the AI industry, including Anthropic, right now' — birchlse · 2026-11-15
- Epoch ECI Trend Fit: Anthropic's Latest Model Runs ~5 Months Ahead — peterwildeford · 2026-11-15
- Ex-Citizen Quant Head Dan Morillo: AI Outputs Consensus, Not Alpha — minhnguyen · 2026-11-15