Radical AI predictors' direction is said to be correct
Understand this story
The "superintelligence perspective" formed around the AI safety and rationalist community holds that AI will achieve recursive self-improvement within a few years and bring unprecedented social change. A recent METR evaluation shows that since October 2025, Anthropic's models have been improving on benchmarks at nearly twice the previous rate, and much of this acceleration may come from AI itself. The author therefore argues that although "responsible AI skepticism" can question its speed and scale, it cannot deny the correctness of this directional judgment.
AI overview based on source reports · Updated 59 min ago
Coverage timeline
Follow the coverage to explore different perspectives.
- Hacker News AIThe Crazy AI People Are Directionally Correct
The author examines the "superintelligence perspective"—that recursive self-improvement, superintelligence and radical transformation are imminent—and argues it is taken seriously by well-informed people and supported by trends such as doubling task-length capabilities and recent AI achievements in mathematics and cybersecurity. He reviews "responsible AI scepticism," covering general priors, the insularity of the AI safety community, and object-level doubts about RSI, superintelligence and diffusion, but concludes none are knockdown arguments. Even if the sceptics are right about speed and scale, he argues, they should concede the crazy AI people are directionally correct.
Activity for this story
Not enough continuous observations to show a trend yet.