This is a very pertinent reply, very much in-line with our views!

DL most certainly does generalise for specific tasks, today. One example comes from DeepMind who published a model that generalises to Atari games. Saying that a bee "generalises" whilst DL doesn't, it's a far fetched statement.