Evaluators and safety organisations · Article · Epoch AI
Can AI automate AI R&D yet?
Read at epoch.ai ↗Read at sourceAll rights reserved
Outline · 132 paragraphs
- Introduction¶1
- Methodology¶4
- Task setup¶13
- Environment¶19
- Results¶26
- AI did not discover anything comparable to the original innovation¶27
- Agents made misleading claims about their work¶32
- Could scaling up spending improve AI results?¶36
- Even AI models that had seen the original paper struggled to match it¶40
- AI struggles at end-to-end AI algorithms R&D… for now¶45
- Appendix: evaluation details¶49
- Scoring¶50
- Prompting¶60