Learning from Past AI Over-Optimism and Rapid Compute Growth

Historically, AI experts overestimated how quickly machines could achieve general intelligence. Predictions from the 1960s to early 2000s often expected near-term breakthroughs that didn’t materialize. However, AI compute and model capabilities have grown exponentially, with training performance improving 4–5x per year in recent years. This growth suggests that while past optimism was misplaced, current AI progress is on a trajectory that could make AGI achievable sooner than expected.