Twenty Years from RSI to Takeoff: Slow Learning, Scaling Slowdown, Industrial Explosion

A new analysis from LessWrong suggests that while artificial general intelligence could arrive between twenty twenty-eight and twenty thirty-two, a true software-only singularity may not happen until twenty forty to twenty fifty. The author argues that early AGIs will be limited by a predicted compute buildout slowdown starting in twenty thirty-two, meaning they cannot quickly invent the algorithmic breakthroughs needed for superintelligence. Instead, progress will depend on an industrial explosion of robot-building robots to scale compute capacity a thousand-fold, which the report estimates could take another two decades to realize.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton