Mat and Buck Excellent Adventure in Computer Learning

Today I attended a presentation by Matthew Alger and Buck Shlegeris, students at the Australian National University, on "Reward-modulated inference". The topic sounds dry and technical, but Matthew and Buck's two person hyperactive performance was very entertaining, in the style of "Bill & Ted's Excellent Adventure". While not qualified to comment on the technical detail, it stuck me that there are further insights for human learning which can be gained from machine learning. These include finding the right combination of supervised and unsupervised learning, rewards for reinforced learning, the value of preprocessing, deep learning in neural networks and advanced in extreme learning.