Great post by Melanie on how being able to form new concepts & abstractions is the #1 unsolved problem in AI, and why ARC is a useful stepping stone to get there. ARC is the kind of meter that progress towards generality in AI should be measured against.