Saturday, April 18, 2009

Ay

Once again, I'm on an AI kick. I tell myself that this is a good thing; it is a way to learn more about a subject that's interested me for years, off and on, and, who knows, perhaps I'll find a way to apply some AI concept to the real world, and make some money doing it.

(Pause for muted laughter)

Okay, so I know that is pure-D rationalization, but I still do it anyway, since the idea of reading about a technical subject with no possibility of getting any money out of it at all somehow seems silly and wasteful and wrong to me. I can't really explain why, it just does. Even now, when I don't have a job to go to, to take time from to read this stuff, doing it seems like something that I ought not admit to.

Technical stuff that I don't understand has always had a hold on me. From the time I first read about Claude Shannon of Bell Laboratories, doing his seminal work in information theory (which I read about when I was about twelve years old; I didn't really understand it then, and I barely understand it now), I've been fascinated by technical concepts that I didn't understand. Once I get to the point where I barely, generally, somewhat, to a certain extent, understood them, my enthusiasm for the subject dropped, and I forgot about it for a while.So I know little bits and pieces about a fair number from things, from why the I Beam looks that way to what mitral value prolapse means, and all of which, as someone once said, stuff that I can't make money from. Others can, and do; I don't. But it still interests me.

At the moment, I'm trying to come with this: how can you identify a situation which could benefit from the application of artificial (or as I'm trying to think of it, simulated) intelligence? For example, this afternoon, I was sitting in the car, waiting for my wife to come out of the store, and I was watching a guy mowing his lawn. I've heard of lawn-based Roomba like products that work based on you setting markers in the ground -- I'm guessing radio-based, but I don't know -- that tell the robot mower how far to go. Well, I mused, what if you could teach the mower how far to go, and then let it figure out what the best way to approach the task would be? How would you handle the mower that thought it was supposed to go mow the neighbors lawn, too...or the street? Could you do it without a GPS system, let alone, external markers? What would it have to 'know' that a human mower would know -- and how could you get it to 'know' that stuff by itself?

So thats the kind of thing I'm trying to think about, every so often. And I'm trying to note when I come across an article that talks about the application of AI concepts, like an old article in Wired that I just found the other day which describes the efforts of a fellow named Gerhard Widmer to understand what makes music work. And a different one that laid out some rough delineations of types of AI applications. That kind of thing. It's fun. Odd, okay, pointless, of no value but to my curiosity -- but still: fun.

No comments: