Over There

I occasionally put notes into another blog that I write. The blog is oriented around a project that I'm working on, one which can drive me absolutely batty (no, even more so than normal). I started it because I needed to do something as an outlet, and since writing is about the most creative thing I do, that suggested itself. I can't say that its helps all that much, but its a bit amazing to look at it and think of all the time thats passed on this project, one that I really thought would be, if not done, then substantially done by now. Whereas we're not even started, to speak of.

The project involves installation of a piece of software that's relatively straightforward to install (I can say that with a straight face) and relatively complex to run. It will run without any tuning at all, but if you want it to do useful things for you, you have to tune it -- set values, configure it. Most people don't bother. Its not that they don't want it to do useful things for them; its that they have neither the time nor the expertise to do the configuration or the tuning. The product is not quite shelfware....but in some ways, it is. Its the kind of thing where you come into a shop and say wow, you have that? And they say Um...yeah. Why?

For years, I have been interested in the concept of artificial intelligence. I am embarrassed to say just how long it took me to realize how absolutely down to the gravel actual artificial intelligence was. Its not that I thought that the places which worked on it had secret devices that they'd gotten from crashed alien spaceships -- but I did think that they must have some programming techniques far beyond mortal man. Things that only truly amazingly superior people could do. With perhaps a little bit of alien science thrown in.

Finally, I came across Patrick Winston's book on the LISP programming language. I didn't really understand it, but for the first time I realized that artificial intelligence didn't actually mean 'intelligence' so much as it meant 'producing actions that are similar to those which could be produced by intelligence'. I was awed by this. I began to think that normal people -- maybe even me! -- could do this. And they can -- because, properly done, there is nothing amazing about AI -- its just a computer system producing output, like the payroll system or the traffic light controller. No big deal. All you've got to do is code it. Very, very carefully. With a lot of attention to detail. And multiple iterations, and prototypes, and trials. And a goodly amount of ingenuity.

Its like that great description of writing: nothing to it. Just sit down in front of the blank paper and stare at it until blood comes from your forehead.

Which brings me to the Omegamon product, the tool that I'm installing. Now, I happen to be an Omegamon bigot. I think its a marvelous tool. I think it can do things that no other tool in its field can do. And this is mostly (though not entirely) true. But the only thing that it cannot do is think. It cannot produce the effect of intelligence. You ask it a question, you get an answer. You don't get everything you wanted to know, or even everything it knew about the subject of your question, you get - the answer. And even to get that to any useful degree, you need to tune and configure it. There are about twelve gazillion knobs you can twiddle. Have fun.

I think it should be better than that. I want to make it better than that. I want Omegamon to not NEED the tuning, not NEED the configuration. I want it to observe my environment, make reasonable guesses, and then start running. And every so often, make useful observations to me. Like 'Hey, heres a problem situation building; I saw this last week just before the system cratered, and here it is, happening again.'

Is this possible? Well, no, not in a STAR TREK kind of way, it isn't. But AI is out there, and its making computer systems better. Its not making them HAL-like but its making them more intuitive, more graceful, less UNIVACy. And that's where Omegamon is now. Its a marvelous tool but its not an intuitive tool. I want to move it beyond that UNIVAC-like stage . And it drives -- me-- crazy that I can't do it. Because wanting is not enough. You need more. You need ability, you need a vision, and you need a way to get that vision down here into the real world. I can't do that. I don't know how. I don't know anyone who DOES know how. And though I may not be too old to learn, I'm too old to do anything really sexy with anything that I do learn, along those lines. And other than my wife, I don't know anyone who even wants to talk about it, contemplate it, try to shape the vision. Its just me, muttering in the dark.

Drives me crazy.