In addition to coursework and reading, I’ve continued to think about checkers—because it seems both challenging and tractable, and I have the basic domain knowledge. These supports two main short term goals for the next 6 months or so. The first goal is to learn and understand the next level of AI. This includes deep learning in general, convolutional NNs and reinforcement learning (as opposed to supervised learning). The other goal is to have a successful project where I’m able to build something new (to me) using AI.

Early on in thinking through the checkers problem, I had a belief that I would be able to build such a thing with the knowledge that I had. However, the outside reading that I’ve done, combined with the issues I’ve been thinking through concerning checkers have made me realize that right now, I just don’t yet have the right tools in the toolshed to tackle the problem. That said, as I continue through my deep learning coursework, I will continue to think about the approach to checkers and where what I’m learning fits with the particulars of that challenge.

I suppose one of my main challenges will be avoiding spoilers. I’m sure someone somewhere has done this already, and I want to avoid looking to ready-made solutions to the problem as I design my own solution. Of course after the fact I will probably look for how other solutions differ from my own and see how perhaps I could improve what I’ve done.