Comment by andai

Does ARC measure "one shot learning"? I heard that the major unsolved problem in ML was developing systems that are good at dealing with novel problems.