22 November 2004

Computer Generation

Today's news includes three articles that relate to "computer generation." The most obvious is one discussing computer-generated fiction, which really doesn't have anything to say that's "news"—the "advances" described are all two years old or more, and Orwell anticipated them with the novel-writing machines in Pornosec. Come to think of it, that explains a great deal about Barbara Cartland et al. Maybe I should have someone at Romantic Times check the next conference it sponsors for cyborgs… especially those still running BSD.

More to the point, perhaps one should see whether the film and TV industries have gotten a jump on print publishing. Based on the typical TV script of late, Hollywood has long been in "just add character names and corporate sponsorships and stir" mode. How else to explain the continued prevalence of really, really awful movies that cost a lot to make and nobody goes to see (or at least will admit to going to see)? Frankly, it also explains exactly why two studios are so much less cooperative than the others in WGA arbitrations: They don't want HAL forced to testify, possibly out of fear that their "scriptwriter" will start singing "Daisy" in public#8230;

On a more serious note, computer are all about the repetitive processing of information. Their results are always subject to human interpretation, and frequently to human overruling. This, as much as anything else, explains why the three big debt-rating agencies were so inexcusably slow in downgrading debt from WorldCom, Adelphia, and Enron. The motivation for doing so will probably remain a mystery in those three instances—anybody who believes that most of the relevant documents weren't shredded hasn't ever seen just how efficiently a shredder operates on the low volume of information created in merely financial contexts—but, at least in the eyes of the Washington Post, the general practice could stand some greater scrutiny. And that's before we get into the racism invisibly (and probably unconsciously) built into the model used to create consumer "credit scores."