One of the great presentations I saw at SXSW was the one by Doug Lenat on Artificial Intelligence. Lenat has a company called Cycorp where he works on an A.I program called Cyc. Of course, a lot of the promise of A.I. is overhyped and there is a lot of work to do before computers can truly replicate human thought in any significant way. Still, the presentation by Lenat was very interesting.
He began with a review of some A.I. milestones, such as efforts to pass the Turing Test, which is basically an imitation test where the computer has to fool a person into thinking that the computer is actually a human too. Joe Weizenbaum's 1965 psychiatrist program, which mimicked an empathetic shrink by asking simple questions was an early effort in this regard. Lenat pointed out the program produced some funny interchanges like this:
Patient: "My dog's mother died recently"
Computer: "Tell me about your mother."
The computer obviously missed that the focal point of the sentence was the dog's mother, but Lenat pointed out that a lot of real psychiatrists might assume the patient really wanted to talk about their mother too (it's not a bug, it's a feature).
The trick with A.I. is to get the computer to understand the context of language, something which people are good at, but which is really hard for a machine loaded with software. Another example:
Human: "I'm going to ask you some simple questions."
Computer: "Do you think your plan will succeed?"
And what happens when the computer doesn't know the answer, or even understand the question? It has to improvise in order to appear helpful.
Human: "What does the letter 'M' look like upside down?"
Computer: "Try searching the Web."
You gotta love it; the computer sounds like the fifth grade student whose dog ate his homework.
Lenat admitted that his discipline sometimes seems overwhelmed by a monumental task: teaching machines things that are common sense to most people. But, he says whenever he starts getting depressed he reads product warning labels (e.g. the chainsaw warning that cautions against putting your hand on the blade while it is spinning). So maybe the key to A.I. isn't for computers to become smarter, but for people to become stupider.
No, actually, the key to the project is to load the computer up with lots of basic rules. Lenat thinks that he is close to reaching a threshold point, after which the computer will actually be able to bootstrap itself into a new level of "understanding" where really useful things will start to happen (e.g., the computer can become an expert witness for plaintiffs in products liability trials).
He still needs help with the 'teaching of basic information' and, guess what? You can help. If you are interested click here to see how you can tutor a computer from home.
I met Mike at SXSW and he graciously introduced me to various wonderful people. I found out that his law blog (one of the earliest in blogland) has an RSS feed. Check it out.
6:26:26 PM
Paula Brantner of the Workplace Fairness Organization emailed me to say that they have now added an XML Feed to their Blog, which is entitled Today's Workplace Blog.
5:57:17 PM
If I put information up on my website and anyone can access it, is it wrong for someone to report that information elsewhere? Apparently, it is in Texas. Sheesh, the next thing you know someone will be claiming I can't link to their site without permission. Oh, wait that'salreadyhappenedtoo.
4:40:14 PM
Funny thing is, people don't like being "Googled." In fact, I heard several comments about this at SXSW from people who either (1) didn't like that others could know about them, or (2) were afraid that they couldn't really figure out, from their own search results, if the information that they found about someone else was true. Obviously, as we move toward a world of pervasive access to world-wide information we are going to have less privacy. At least privacy of a certain kind.
I, personally, don't care if people Google me. I know that what they learn will be limited in scope, no matter how powerful Google is (I am not the embodiment of the results of a Google search). I also don't mind that people will delude themselves into thinking that they know me from a Google search: people have had misguided notions about things like this for millenia. Finally, I'm not even worried that people will learn the inner private me from reading this weblog.
What do you guys think you know about me, besides trivial, factoid stuff? The real me is not in Google's search pages, nor in the pages of this weblog. People who obssess about what people might learn from them from Google are the modern day equivalent of the noble indians of a hundred years ago who believed that having their picture taken would somehow rob their soul.
Kim Plonsky of KiMformation is gathering resources for online court records. See this post below:
Thanks for the mention . . .. . . . to Naked Ownership for this post, in response to my recent post about the status of online access to Louisiana state court records. Thanks, also, to Tim Hadley, of Math Class for Poets: Law and Life, for the info on online access to court records in Colorado and Wisconsin, which gave me the idea that I should start compiling links here to online state court resources throughout the country -- hoping, of course, that some of you blawgers across the land will help me along by punting me some more links from other states to get started (hint, hint).
Won't you help? It's the gift that keeps on giving. Thanks.
