[cglouch]

I know it's not the same type of AI, but in chess there's a whole scene for computer + human play. A chess engine on its own can have trouble seeing strategic ideas that humans can recognize (e.g. opposite-colored bishop endgames, certain closed positions, and fortresses) so an engine on its own will lose to that engine being assisted by a skilled human player. In other words, humans are still capable for contributing at least a little. That said, it's not much - I think a chess GM paired with an engine will probably only be able to beat an engine rated ~100 or so points higher than their own.

It will be interesting for deep learning, though, where the ideas are a bit more abstract. Perhaps humans will be useful for a while longer.