OpenAI's o3 hits 87.5% on ARC-AGI, reigniting AGI debate
A new Reddit thread questions if AGI is finally here after o3's record score.
Deep Dive
A Reddit post was submitted by user beasthunterr69.
Key Points
- OpenAI's o3 scored 87.5% on ARC-AGI, beating human average of 75%, but costs $20 per task in compute.
- Reddit post by beasthunterr69 has 12k upvotes and 2,500 comments, reflecting intense AGI debate.
- François Chollet called o3 'a serious step forward' but not AGI, citing lack of sample efficiency.
Why It Matters
If 87.5% on a reasoning benchmark is not AGI, how close are we? This debate shapes investment and safety priorities.