The "why" of modern machine learning algorithms, and especially the present generation of deep neural network derivatives, is spoken of in news, culture, and media alternately as an unfathomable black box incomprehensible to mortal minds on the one hand, and, as an utterly simple mechanical triviality absolutely obviously devoid of any manner of there there on the other.

Both of these extreme simplifications are wrong.  The reality is far more complicated, more interesting, and more difficult to grapple with both technically and in terms of broader societal impact.  Grapple we must.  Sane progress, sound policy, and future fairness and safety depend critically on nuanced technical literacy winning out over motivated grandstanding, trite minimization, frenzied overstatement, and handwaved halftruths. 

My teaching agenda centers on grounding students' understandings of the new tools now permeating our world in the bedrock mathematics and numerical methods from which they emerge, together tracing the continuous thread from zeros and ones to today's artificial intelligence.  This fall term I'm teaching Math in Machine Learning, and in the spring I will teach Generative AI and LLMs.