A talk by AI researcher François Chollet. He discusses the ARC-AGI benchmark, which was created in 2019, and the new ARC-AGI-2 benchmark, which was announced this year.
He also briefly discusses the upcoming benchmark ARC-AGI-3, which is planned for a release in 2026.
Scoring high on these benchmarks does not show that a system is an AGI. Rather, these benchmarks are intended to measure incremental progress towards AGI.
Video description:
ARC-AGI-2 was launched on March 24, 2025. This second edition in the ARC-AGI series raises the bar for difficulty for AI while maintaining the same relative ease for humans.
It represents a compass pointing towards useful research direction, a playground to test few-shot reasoning architectures, a tool to accelerate progress towards AGI.
Video URL: https://youtu.be/TWHezX43I-4