Quote:
Slides can be found here: fergalbyrne.github.io/fergbyrne-euroclojure2014.pdf
(Audio gets better after the first minute and 20 seconds).
The goal of the talk is to introduce Jeff Hawkins' theories as a basis for intelligent computing; to describe how Clortex can be used for both research and development of new applications of Machine Intelligence, and finally to describe some of the key design decisions made in developing a large cortical information processor in Clojure. The attendee will learn about a new vision for understanding how the brain works; a system for developing genuinely intelligent software and hardware based on neuroscience, and how to use the Clojure ecosystem to address large-scale design and development issues.
The presentation will begin with a 3-minute demo of Clortex in action. The demo consists of:
A visual example of Clortex processing and intelligently responding to a real-world problem.
A simultaneous set of visualisations showing the inner workings of the neural regions inside Clortex.
Live manipulation of the algorithms used, and a demonstration of how this affects learning and performance in Clortex.
The first section is a crash course in understanding the neocortex: the seat of intelligence in the brain.
The next section will briefly introduce Jeff Hawkins' theories of Hierarchical Temporal Memory and the Cortical Learning Algorithm, with the key principles of how the neocortex identifies spatial patterns in data and learns sequences of those patterns.
Following this is a brief history of efforts to implement systems based on HTM/CLA, including Numenta's NuPIC, Dileep George's Vicarious, CEPT's Cortical Engine for Text Processing, and a number of hardware projects currently in development at Sandia Labs etc.
