 So, in my perspective, PyTorch is just wonderful. It's been taken up by the academic community extremely rapidly. This is what you can see on the left side here. PyTorch came, started becoming popular in late 2017 and by mid 2019, it had reached the majority of the scientific community. Before that, a lot of people were using TensorFlow and the transition was extremely fast. Here's a fun comment by André Capathy, one of the leaders in that field about it. I've been using PyTorch a few months now and I've never felt better, I've more energy, my skin is clearer, my eyesight has improved. But this is just an example of how people get excited about it. It's a very useful environment in which to build deep-blowning systems. So, now, whatever we're going to implement, in this case, AlphaZero game playing, the key for any successful deep-blowning project is having a good intuition of what really that problem is that we will want to solve. So why would we want to use deep-blowning in this context? Well, there's lots of different ways how we could use it. Maybe we could say, let's choose the best action as a function of the bot. And we could say, let's use machine-blowning, deep-blowning for this estimation, the best action as a function of what the bot is. This is not how AlphaZero actually works. So how would you use deep-blowning to solve such a problem? So now your task is the following. Play another student and while you play with them, really train, ask yourself, what are you doing and why are you doing it? And describe how you played. While you were doing that, were you optimizing something? How could you imagine making a computer play well in this context?