The problem is that reinforcement learning is far from solved and doesn't work all that well yet, so these toy problems are probably what researchers will stick

The problem is that reinforcement learning is far from solved and doesn't work all that well yet, so these toy problems are probably what researchers will stick with for some time to come.