1 of 1
Story summary
- Jared Kaplan, Anthropic's chief scientist, warns that by 2030 humanity will face a decision on whether AI should train itself, risking an intelligence explosion.
- Kaplan cautions that recursive self-improvement could let AIs learn autonomously, raising questions about their impact.
- Kaplan notes some experts argue current AI technologies may not evolve into the powerful systems he fears.
- Yet Kaplan concedes AI capabilities could stagnate, suggesting the best AI may already exist.
