I used to work in RL. While it is an amazing area to research in, I don't see many other use cases for it other than training robots to do a specific thing.<p>What other use cases are out there and which companies are working on it?
I saw someone from DeepMind mention using RL for Recommender Systems at Google at their ICML 2016 presentation, but they were very tight lipped when I asked for more details.<p>Bandit algorithms are pretty popular in the ads world, and there is probably some use in extending that to the full RL setup.