Wednesday, October 3, 2012

Algorithms That Learn: The Rise of Artificial Intelligence

▲ ▲ ▲

A system is intelligent if it is more useful to talk about in terms of goals than in terms in mechanisms. ~ Richard Sutton

The above quote that an intelligent system has goals is skirting the edges of the definition and discussion of consciousness. Reaching the point when software, an algorithm or even a network, "wakes up" and has consciousness has been the subject and speculation of many a technician and philosopher. This is the ultimate line to be crossed to achieve artificial or transhuman intelligence.

How do you reach this line to be crossed? Lukasz Kaiser begins with describing the existing deterministic "agents" as receiving an input stream, then a program that produces an output stream. This is the mechanistic view. Another view is to think of developing software in terms of the program having goals, as noted in the quote above. Kaiser then presents a higher-level model to achieve goal-oriented results, an algorithm that learns by observing.

Lukasz Kaiser: Playing General Structure Rewriting Games Dr. Kaiser’s research presentation at the Third Conference on Artificial General Intelligence. This is a joint work with Lukasz Stafiniak.

It’s a Bright Future If You Are an Algorithm, the New Evolutionary Force

Proposed Sign for "Dangerous Artificial Intelligence"

▲ ▲ ▲

Seeking Alpha