Ai Dreams Forum
Artificial Intelligence => General AI Discussion => Topic started by: Bill DeWitt on November 18, 2006, 08:11:34 pm
-
2) Adaptivity - The Agent is capable of adapting it's behavior, processes, and components - to changes in it's environment in order to enhance the Agent's chances of further existence (i.e. survival).
Lamarckism is thoroughly debunked elsewhere, so I will assume you mean behavior only. Either that, or you mistake evolution to be goal driven. If the latter, suffice it to say that random changes in the allele are not ?in order to? do anything, they simply are.
Now ?changing behavior to adapt to changes in the environment? is almost certainly a test for intelligence. Unfortunately we are talking about a test for Machine Life. Again, I return to the simpler forms of life, starting with the virus.
It can?t change its behavior in any way (Nor, for thoroughness, its processes or components). All these changes are either done through the failure of autopoiesis or through a failure of allopoiesis. This leaves us in the unenviable position that if we accept your #1, then we must abandon your #2, and if we accept your #2, we must discard your #1.
Be that as it may, let?s move up a step in the ladder of life and consider a bacterium.
Little Joe, the bacterium, is floating in some goo. Suddenly, the PH changes to a deadly level. Little Joe wishes he could change the hydrogen bonds of his integument to adapt, but he cannot change his processes. He wishes he could grow cilia and wriggle away, but he cannot change his components. He wishes he could do something, anything, to escape, but he has no behavior to change. He has three functions, Eat, Grow, Reproduce, and no way to change any of those.
Fortunately, earlier, one of his functions, Reproduction, didn?t work quite right. His offspring was created with stronger hydrogen bonds on his skin. So while Little Joe died because he could not adapt, he was, nonetheless, alive.
So we have at least two forms of life that do not directly exhibit adaptation. The first can only be shown to change through a failure of your test #1, and the second only can be shown to change due to a failure of your #3.