DeepMind essay maps four paths to superintelligence: effective compute grows 10x per year

udmrzn · x · 2026-11-20

A new DeepMind Institute essay by Tim Genewein, announced by Shane Legg, argues AGI is 'coming within reach' and lays out four paths to ASI: bigger models and data, new paradigms, millions of coordinated agents, and recursive self-improvement. Effective compute stock grows at least 10x/year (hardware 1.5x, algorithmic efficiency 3–6x, investment 2.5x); a workload costing $1B in compute today would cost $1,000 in ten years, with Epoch finding fixed-benchmark prices falling 13x/year since 2023. The essay also maps a continuum from sub-AGI to true ASI and argues restricting one input like training compute won't stop progress.

Related event: DeepMind Publishes "What Comes After AGI?": Four Paths to Superintelligence(5 posts)→

Original post →

More from AGI Musings

AGI Musings channel →