Re: Hawkins ideas on building AI's

From: bkaz (bkaz__at_hotmail.com)
Date: 10/28/04 

Date: 28 Oct 2004 08:31:41 -0700

> The part in your definition that may be problematic is the "tabular"
> view it seems to enforce like in classical machine learning. It seems
> general enough, right, but it does leave out some valid questions.
 
> What is a variable in a matching input of a pattern?

It could be any number, my approach to generating new predictive
variables by comparing lower level inputs is in 'intelligence by
definition' post on sci.cognitive.
 
> How is it compared?

Comparison is a vector discovery: reverse arithmetic operation of any
power (subtraction, division, logarithm extraction...)
 
> Why parallel? (What does parallelism have to do with this?)
 
By parallel I mean that even though variables of the same type are
compared separately, the match is determined for the input as a whole.

Boris

Quantcast