International AI Safety Report
22 predictions spanning 2026-2030
2026
2027
2028
2030
- AI systems complete days-long software projects
- AI systems complete month-long professional tasks
- AI systems do remote-worker jobs at human level
- AI-guided robots handle complex real-world tasks
- AI capabilities largely plateau
- AI improves across specialized domains
- AI systems become useful digital assistants
- AI uses 7.4% of US electricity
- Chip supply supports 100,000x GPT-4 compute
- Energy bottlenecks do not stop AI scaling
- Frontier models train with 3,000x more compute
- Largest training runs need 4-16 GW
- Largest-model training compute grows 125-fold
- Models need 10-100x less compute
- AI models reach 75% on undergraduate FrontierMath
- Autonomous vehicle deployment remains limited