quote:
A AI capable of doing it's own "artificial" selection based on recongnizing something smarter would qualify.
That is a great idea, if you ask me. Researching genetic algorithms I came across
this article. The investigators allowed random connections between different electrical switches, transistors, etc. The selection pressure was towards evolving a circuit that produced an oscillating signal. Therefore, each time this ability was improved part of the circuit was preserved and new combinations were added. By the end the scientists had a fully functioning oscillating circuit. However, the most stunning part was that the circuit actually relied on a form of parasitism. Instead of actually producing the oscillating current, the circuit used a long run of copper wire in the circuit board to recieve oscillating radio frequencies from surrounding computers. Voila, the radio was accidently reinvented throught the process of variation and selection. Quite a stunning example of how evolutionary algorithms can come up with strategies that seem counter-intuitive but still work, just as we see in nature.
As to developing AI, the same process could probably be used. The exciting part is that evolutionary algorithms usually come up with solutions that humans would not have thought of, which makes this field a very rich resource for designing circuits and software. However, without the correct selective pressures we may end up with results that do not meet our starting criteria, as was seen in the example above. I think it is very possible that AI could be evolved, but it may come down to fine tuning the selective pressures and the source of variation.this article/url. The investigators allowed random connections between different electrical switches, transistors, etc. The selection pressure was towards evolving a circuit that produced an oscillating signal. Therefore, each time this ability was improved part of the circuit was preserved and new combinations were added. By the end the scientists had a fully functioning oscillating circuit. However, the most stunning part was that the circuit actually relied on a form of parasitism. Instead of actually producing the oscillating current, the circuit used a long run of copper wire in the circuit board to recieve oscillating radio frequencies from surrounding computers. Voila, the radio was accidently reinvented throught the process of variation and selection. Quite a stunning example of how evolutionary algorithms can come up with strategies that seem counter-intuitive but still work, just as we see in nature.
This message has been edited by Loudmouth, 08-13-2004 03:15 PM
This message has been edited by Loudmouth, 08-13-2004 03:16 PM
™ Version 4.2