Ben Kamphaus
Cognitect
@benkamphaus
Ben Kamphaus is a developer at Cognitect. He got his start with machine learning and computer vision over a decade ago when he had a crazy idea for finding archaeological sites in satellite imagery. In the past few years he’s focused on understanding, implementing, deploying, and scaling deep learning models. He relies heavily on Clojure and Datomic to stay sane when integrating those models into larger systems. He is the primary organizer for the Boulder Deep Learning and Data Science meetups and enjoys hiking, climbing, and judo in his free time.
Foundations for Artificial Minds
Traditional programming involves constructing systems which are perfectly logical. Human beings by contrast are frequently illogical, and for very pragmatic reasons. We have to accomplish things for which perfectly logical solutions are either not computable, tractable, or feasible. Improving and perfecting upon the human capacity for logic is a fairly trivial task when constructing AI systems, but mapping from the vague and contingent components of human knowledge, perception, and sense-making necessitates empirical, inductive approaches like machine learning. In this talk, I’ll work from a motivating hard problem in AI, discuss the mixture of traditional symbolist and recent machine learning driven approaches to the problem, and highlight how Clojure’s mindset, mix of powerful tools, and pragmatism are the perfect starting point for tackling it.