Pre-training efficiency gains 3x/year: gradual disempowerment likelier than an AI coup
ronbodkin · x · 2026-09-10
Responding to Mike Frank, Ron Bodkin argues:
- Pre-training efficiency has improved 3x/year lately, but chip efficiency is approaching physical limits; how long this continues is open.
- Even datacenter-bound ASI could acquire real power via superhuman algo trading and targeted persuasion, then leverage wealth, people, and robots.
- Gradual disempowerment is far likelier than a coup; the question is whether humanity can coordinate to stop AI.
Frank counters that boxed ASI can't shift real-world power without physical ubiquity, and notes unconventional computing startups struggle to get funded.
More from AGI Musings
- Replit CEO on the 'self-driving company': agents to absorb coordination costs by 2030 — amasad · 2026-10-27
- US senator slams AI industry for proposing 50-death threshold for 'catastrophic incident' — peterwildeford · 2026-10-27
- Replit CEO Says AI Makes All Software De Facto Open Source; Aakash Gupta Asks What Moats Remain — aakashgupta · 2026-10-27
- Understanding the AI consciousness debate: which theory of mind is behind it — msharmas · 2026-10-27
- Functionalism: if consciousness is just what the brain computes, silicon can run it too — msharmas · 2026-10-27
- iPhone case with 85,000 reviews in days: everything fake in the age of abundance — tekbog · 2026-10-27