This is a ridiculously reductive take on DeepMind... they produce a TON of great research, and have helped advance the state of RL considerably (speaking as an AI researcher here, having read many of their papers). Google should be commended for investing in largely basic AI research, even if once in a while they make a PR splash out of it, IMO.