AI Generates Uncanny Player Movement in Games

Given my love for video games and the nature of this class, I present to you the lovechild of this process. At University of Edinburgh, student Daniel Holden, professor Taku Komura, and research engineer Jun Saito created "a novel neural network architecture called a Phase-Functioned Neural Network" in which they were able to generate the movement of a video game model that acts and performs just as if a user was manually operating the character. The comparison of a standard NN, recurrent NN, and Autoregressive Gausian Process to their network is unbelievable. The former processes are rubbish and have a clear flaw in their movement, be it clunky or stiff, but the Phase-Functioned NN generates next level movement that has to be seen.