Penny University Lightning Talks, May 28, 2020
These slides built using Jupyter Notebook and RISE. Best viewed on desktop browser.
As a compuational physicist, my interest in machine learning (ML) was only as it relates to regression, i.e. curve fitting.
Because that's what we do in comp. phys.
But ML tutorials & discussions I found always centered on classification problems, i.e. sorting things into categories:
As part of an interdisciplinary program: Bridging the Two Cultures of Science and the Humanities II.
For two summers, I was mixed in historians, philsophers, biologists, theologians, English teachers, & more.
...and I noticed that many of the discussions were about (what a ML researcher might call) classification problems:
And I started to think: how would some (perhaps unwise) 'Silicon Valley person' try to automate the sorts of decision-making I was hearing -- because someone probably already is.
...at the Royal Society: "You and AI – Machine learning, bias and implications for inequality"
"Systems of classification are themselves instruments of power," and have aided oppressive governments for a long time.
..and I came to understand that classification problems are central to how humans function as a society and as individuals.
Talking to librarians, doctors, biologists, historians, philosophers,... to better understand how humans do classification, and what happens when machines do it.
And hired philosophy students to help me research the history of classification: Plato, Aristotle, Aquinas, Duns Scotus, Ockham, Kant, Linneus, Darwin, Nietzsche, James, Wittgenstein, Mai, Durkheim, Berlin, Rosch...to name a few!
And am writing a book! Taking sabbatical (off) this fall. The eBook will have interactive graphical examples such as these.
Here are slides to a longer talk I gave about this at Wheaton University, with lots of interactive Javascript examples: https://hedges.belmont.edu/~shawley/classy/