The ARC Prize, which launched in 2024 as a bold $1 million competition to advance artificial general intelligence research, is evolving into a formal nonprofit organization. The newly established ARC Prize Foundation, co-founded by influential AI researcher François Chollet, aims to continue its mission of developing rigorous benchmarks for measuring AGI.
Key Points:
- ARC Prize began as a response to perceived stagnation in AGI progress
- The initiative's first year attracted over 1,400 teams and 17,000 submissions
- The foundation will transition to a 501(c)(3) nonprofit and begin fundraising this month.
When ARC Prize launched in June last year, it posed a provocative challenge to the AI community. While companies were racing to build larger language models, Chollet and Zapier co-founder Mike Knoop argued that true AGI required something fundamentally different: the ability to efficiently acquire new skills and solve novel problems, not just memorize patterns from training data.
The organization's transition to nonprofit status comes after a remarkable first year. The initiative garnered attention from major AI labs, hosted a university tour across 14 top AI programs, and built a vibrant community of researchers and enthusiasts. The ARC-AGI benchmark proved particularly revealing – while OpenAI's unreleased o3 model achieved a qualifying score, it required extensive computing power and still falls short of human-level performance on tasks that children can master quickly.
Chris McKay
"We're growing into a proper nonprofit foundation to act as a useful north star toward artificial general intelligence," Chollet wrote in the announcement. The foundation will be led by Greg Kamradt, former Salesforce engineering director, who joins Chollet and Knoop on the board.
The organization plans to launch ARC-AGI-2 benchmark in early 2025, alongside a new competition, and is already working on ARC-AGI-3. Early testing suggests this revised version (ARC-AGI-2) will prove even more challenging for advanced AI models, potentially reducing their success rates to below 30% on tasks that humans can solve with 95% accuracy.
The foundation aims to expand its impact through partnerships with leading AI labs and academic institutions. OpenAI CEO Sam Altman has already expressed interest in collaborating on future benchmarks. As it transitions to nonprofit status, the organization will begin fundraising later this month, accepting tax-deductible donations to support its expanded mission.
This evolution from competition to foundation represents a significant milestone in the field's ongoing effort to develop clear, standardized methods for evaluating AI capabilities. In an industry often driven by corporate interests and closed research, the ARC Prize Foundation stands as an independent voice pushing for open collaboration and rigorous testing in the pursuit of artificial general intelligence.