Researchers developing Alan Turing artificial intelligence 'gooware'

  • Tweet this

Alan Turing was born in 1912 and hearlded as the father of modern computing

When Alan Turing first proposed his idea of “unorganised machines” in 1948, he was attempting to represent a new approach to artificial intelligence.

The infant human cortex, he reasoned, was an example of an unorganised machine, with its seeming random initial construction proving capable of being trained to perform particular tasks.

Now researchers at think they've made progress in building their own “gooware” computer, based on Turing's unorganised machines.

In Turing's seminal paper, he argued the simplest form of unorganised machine would be a randomly connected network of NAND logic gates – which he dubbed A-type machines.

Andrew Adamatzky and his colleagues the University of the West of England's Unconventional Computing lab have long been interested in chemical computers – mostly based on an unusual group of chemical reactions known as a Belousov–Zhabotinsky (BZ) reaction.

What's interesting about these reactions is that they don't reach a stable equilibrium point, and subtle stimuli can produce patterns in an otherwise tranquil mixture. The waves of chemical changes in the BZ reaction are used as the basis of transferring information.

Adamatzky and his team have shown previously that it's possible to build a NAND gate using a BZ reaction.

Unlike traditional processors, that have intricate nano-scale designs and bits behave in a regimented fashion, a BZ processor can potentially move information in any direction. It could be capable of handling far more data than a traditional computer.

Now, Adamatsky's team have shown they can use their BZ NAND gate as the basis of an A-type unorganised machine.

“It was then shown how a number of well-known benchmark logic circuits can be designed from A-type unorganised machines using an approach inspired by a comment from Turing on cultural search,” the group wrote.

At this stage, the work is still some way off producing a new form of artificial intelligence, but the researchers believe it has serious potential in the pharmaceutical industry, as the basis of so-called smart drugs, which only deliver their cocktail of chemicals when the conditions are right.

07 Dec 2012

What do you think?
blog comments powered by Disqus
To send to more than one email address, simply separate each address with a comma.