Monday, February 28, 2005

I take back anything I ever said disparaging or negative about ice cream. Ben & Jerry's is a triumph of food science, unequalled in all the world of mass production foodstuffs.

An interesting aspect of debugging is that when performing triage, you often must leave errors on multiple levels of code organization, requiring you to keep in mind the way layers above and below 'will work' or at least ought to, in order to code effectively. But one must take care not to let this idealization get in your way, when you need to decide what to do next. You can make a great deal of 'progress' figuring out how things ought to work, and not actually get any closer to your goal. You still need to remember and change things.

Programming is very close to the world of ideas, and more than any other kind of engineering, it's easier to blend the two. But if it's not written and compliling, it doesn't count, no matter how deep your understanding.

As you may know, or not. We're building a demo now here at AdaptiveAI, as our news updated indicate. Progress is good, and very interesting. Of course, when it comes down to it, we're really talking about two demos, one to convince people we're doing an interesting and opportunistic thing, and the other, the second, more unstructured demo, to convince them that we can do it.

Unfortunately, nobody is going to pay for low-level intelligence and perception. So it's all about the path. What is the path from here to there? On the mountaintop of intelligence is what everybody is interested in, but that doesn't mean that's where the deep magic is. What we have is very interesting on it's own, to me, and no doubt to many others. But the implications is really what makes it shine. I'm finding that those implications need to be explained in more practiced language and examples than I have yet.

AI is a serious problem, it needs people attacking it with serious projects using real tools. I don't think it's going to be completed by small teams in the near term. As time goes on, of course, it will get easier. Maybe in a few years we'll have the theory to push more and more of the complexity to tools and crunch, but for now it's going to take people and time. That's why we need to expand our project, I think. Other a2i2 members may have different thoughts, of course. Today Josh and I were talking about the hypothetical case of a more public, more mainstream a2i2. What would it be like to publish more regularly, to have an active website, maybe a discussion list? Partially public demos and explanations. It might be interesting. But I don't really think it would get us there any faster. At least not at the moment. That's not to say I wouldn't enjoy managing something of that nature. For now, it's about finding people and time. And being very very clever and careful, and being first.

No comments: