At our Town Council meeting tonight, a member of the police force we share with our neighbors just south of us appeared to give us an overview of a week-long course he’d recently completed in composite ID sketches. There are very few sketch artists in the US, despite what we see on TV, which means that most times sketches aren’t made until several weeks have gone by and the witness’s memory is no longer fresh.
When his presentation ended, I asked him if computer aided art was ever used (I thought it would speed up the process.) His answer surprised me. Computer aided art isn’t used, because the police have determined the results are too perfect to be recognized as human.
Apparently, the human mind needs to be able to fill in the spaces between the pixels, so to speak. But it’s more than that. No face is truly symmetrical–and the witness’s mind registers the asymmetry and registers it in a way the computer program can’t do.
I see AI as still being a long way off. What about you?
I’m going to disagree with you if your only basis is that computers draw too well. There’s a difference between Artificial Intelligence and Artificial Human Equivalent intelligence, which, by definition, is not perfect intelligence anyway.
Great, Pav–can you explain the differences. It may help me to understand why science seems to be looking for ways to create Artificial Intelligence, which is what we hear about, while we’re sort of in the dark about AHEI. AI frightens people on many different levels, not the least of which is the Paradise Lost aspect of ii.
What’s the aim–the goal–of AHEI? If fully developed, will it have use?
Well, AHEI is a term that I just made up that seemed suitable for the purpose. Secondly, AHEI, to the extent that I meant it, would simply mean intelligence that is subject to human error. The goal of AI, of course, is to eliminate the human error, so, they’re not striving to make a human equivalent.
Anyway, that’s what I’m getting at. Humans look at a description and they see imperfection where the AI looks at the description and defaults to producing perfection, or specifically, to whatever description it is told to produce. I don’t think that makes the AI in any way inferior to a human, it just means that, because the AI’s intelligence (with respect to what it knows and knows how to do) is perfect, it can only render a, “Perfect,” result. In other words, it can’t be used for situations in which a perfect result is not desired, which are few and far between.