One of the most promising areas in AI is Reinforcement Learning. We’ve seen it beat top humans in Go, League of Legends and Starcraft. We’ve also seen it learn to play hide and seek. This kind of feats was impossible to achieve using former learning styles (supervised or unsupervised). Although we have not seen a lot of business applications for Reinforcement Learning, we’ve seen very interesting robots and some recommendation engines but I think there is a lot of space to explore this intersection.
A lot of research is being done on the agent side of Reinforcement Learning. We’ve seen techniques that allow agents to solve problems faster. We are getting better at beating humans in Go but that won’t necessarily help us apply this kind of technique in the business world.
In order to apply Reinforcement Learning in the business world, we need an environment builder. From my perspective, we have great agents and there is a lot of momentum on that front. We are lacking a way someone can build an environment that represents a business problem. The holy grail would be that a business user can build that environment by himself (with drag and drop), we can start with baby steps even a little framework with python scripts could work.
Once we have a way to build these environments, I think we’ll see a lot of different applications. We could see new stuff on forecasting, navigating complex sites and invoicing to name a few.
If this is something you are working on or seems interesting to you I would love to hear about it on twitter.