AI: Artificial or Actual Intelligence?
Sunday, January 24th, 2010Artificial intelligence is a system that makes decisions on its own and learns from some kind of input. The goal of AI is to make our lives easier by enabling a computer to be able to perform not just menial tasks faster and better but to go a step further and to have the computer be able to perform more advanced tasks faster and better.
Some uses have been to take an image and to have the program produce a 3D model of that image, determine the most efficient layout of pipes supplying fuel and many other tasks that would take people several years of experience and testing to figure out. While this is all useful, I find it difficult to define this as even artificial intelligence. For an artificial intelligence, it would require an intelligence. So far all we have been able to accomplish is making things learn a defined system quickly, not to learn and adjust to new systems.
I’m being highly critical of a very useful field of study, so what gives me the right and what ideas to I have to improve the system? The current AI programs are good tools for what would be an actually intelligent program. The learning algorithms would be useful if the program could use them by choice in order to come up with judgments. What would make a program intelligent is to give it a method for judging not just the intended result, but all the results that could come from the actions and learning new systems on its own. If humans only concerned themselves with whether their action was good or bad based on the intended result, then we would miss out on so much innovation that we would still be carving wheels out of stone (but very nice wheels). One would never consider making wheels out of wood because it would have required the imagination of taking the accidental effect of being able to bend wood and then thinking that one could bend that wood into the shape of a wheel. This is the same for AI, right now we are telling the program to quickly make a really nice stone wheel, when we could be trying to get it to consider what would make a better wheel.
Now this is a very complicated task, which is probably why it has never been accomplished. Another problem with this is that we may not even know if we have accomplished it if we were trying for it. The AI may decide that it doesn’t want to respond or that it only wants to do what it’s told. Another reason could be that we would understand the mechanics for its decisions and therefore always classify the choices as pre-programmed responses. Of course as the understanding of our own consciousness may help in finding a better method for an artificial intelligence to appear more intelligent.
I think that it is very possible to create an actual intelligence, it’s just that the applications and usefulness of one are too far away from what is wanted to put too much effort into it.
