radiopeter

The subject of this post is part of the Dune mythos that the great Frank Herbert created. It’s something I have believed with every ounce of my being since I first read that line. Machines that learn MUST be watched carefully.

Learning doesn’t mean storing information. Learning isn’t being able to answer the question “what is 2+2?” Learning, by my definition, is the ability to make intuitive leaps in order to fabricate a personal moral code. I don’t see that as a human trait only either.

For example, my cats have learned that after I eat, they can usually expect food. Since I am the dominant animal in my house (and I work VERY hard to maintain that position, no joke) they have come to see me as the provider and the nurturer. So when I have a snack, and they see/smell/hear me eating, they want something as well, even if it was only minutes after they have finished a meal. They have also learned that after I’m at the stove for a while, I will be eating. That intuituve leap is what I consider learning.

I just read an article posted on Slashdot that details some research work being performed in Italy. In the project, they are adding intelligence to an AIBO dog to allow it to share the things it has been learning with another AIBO.

Now, I’m trying not to imply a dooms-day scenario here, but it’s bad enough that we’re teaching machines how to learn. I find that concept really scary because I share Frank Herbert’s fear of machinery. If you make a machine that learns, maybe one day it will learn more than you. Maybe it will learn how fallible and vulnerable people are.

And it may learn that it can do better things without us.

This entry was posted on Friday, June 23rd, 2006 and is filed under blogpost. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.