Do any of the big AI companies have a model that are good at tasks that require learning?
For example, every day people teach teenagers how to drive and with only dozens of hours of practice, they are on the road.
is this not essentially what ARC-AGI-3 is? i agree that in-context/continual learning is somewhere the models are still mostly weak at