Remember HAL in Stanley Kubrick's "2001: A Space Odyssey," the computer that got a little uppity and locked its master out of the space capsule? I’m happy to report that HAL has straightened himself out. He's alive and well and living in San Diego, where he's working as a neural network by day and an intelligent agent by night.
    Thanks to HNC, San Diego’s largest software company and a pioneer in artificial intelligence, smart software now knows its place. One place is at the Charles Schwab & Co. call center in Omaha, Neb., where a latter-day HAL pores over the daily deluge of e-mail messages coming into the huge online financial services company and "auto-routes" them to various departments for final handling. At Sure Trade, a hyper-discount brokerage, HAL saves even more on staff overhead by "auto-responding" to queries with canned responses.
    Through a unique combination of neural networks and real-time content analysis called "content mining," HNC subsidiary Aptex is helping a growing array of companies automatically "read," categorize and respond to e-mail queries coming into customer support help desks. "And since the vast majority of messages fall into one of a handful of recurrent types of questions, such as 'I’ve forgotten my account password and need a reminder,' we can easily send one of a group of preformed messages back to them," notes William Caid, Aptex chief technology officer. The system, called SelectResponse, not only greatly reduces the head count of technical support staff, but its categorization method also is more consistent than the vagaries of human decision making.
    But there's more, much more, that this software can do. While the algorithms used in Aptex's neural network can’t actually think, they can certainly learn. Show its SelectCast software a data bank of 1 million buyers' profiles (age, sex, income, etc.) along with their recent purchases, and it will tell you what the one-millionth-and-one buyer is likely to acquire. Highly likely, in fact.
    So now you have a tool that can profile a surfer's buying preferences in real time merely by watching his or her behavior on a Website and comparing it to previous buying patterns. "The metaphor we use is color," notes Caid. "Every behavior is assigned a number or vector that we then translate into a color, like at a paint store. As our system watches a surfer click through the site, we can add more and more data richness to the color, fine-tuning it to uniquely identify the visitor. When a surfer initially comes onto a computer Website, we know only that he's interested in hardware and he's just basic blue. But when he clicks through to view portable disk drives, then examines the cost of bundling a high-end workstation with 3D-rendering software, he's now a lighter shade of cobalt blue."
    And therein lies the rub, for "a cobalt blue type is highly predisposed to purchase top-of-the-line zip drives," beams Caid. "At which point the site is triggered to flash a special offer on high-end Iomega Zip drives and presto, the customer is sold." (Looks like HAL may soon be wearing Armani suits and driving a Porsche.)

    Intelligent Agents Prowl
    Meanwhile, at Centrax Corp., a different kind of HAL works the night shift as an intelligent agent, where it roams large corporate networks to sniff out aberrant behavior by "insiders" that may signal corporate monkey business.
    What kind of monkey business? The kind that took place during the spring of 1993, when a General Motors exec named Jose Ignacio Lopez began exhibiting some rather compulsive-looking behavior: He began downloading literally tens of thousands of sensitive documents. It wasn’t obsession, it was larceny — on a grand scale. When Lopez took the huge cache of trade secrets to his new employer, Volkswagen, both criminal and civil actions ensued that made worldwide headlines for years. And although Volkswagen ultimately agreed to pay GM $100 million and purchase $1 billion worth of parts from GM, industry observers say the ultimate damage was probably much worse — perhaps incalculable.
    How could such a thing have happened? Didn’t GM have all the latest information risk management techniques — secure files, passwords, firewalls, encrypted messaging, electronic shredders?
    "They had all the passive barriers in place," notes Chris Byrnes, CEO at Centrax, "but these are largely designed to keep outsiders out." The real danger lies within, with those most trusted employees that already are in the castle, so to speak. Furthermore, with most passive barriers, "anybody can eventually get in if given enough time — like rats gnawing through warehouse boxes during the night." So, much as surveillance cameras constantly scan a warehouse for suspicious activity, corporations need intrusion-detection software that will constantly scan their networks, noting unusual or extreme behavior.
    Centrax's software is constantly on the prowl, scanning mainframes and desktops alike for subtle changes in user behavior that can often signal that, well, somebody's up to no good. "If an 8-to-5er begins staying late and copying files, or requesting passwords they've never cared about before, our system will detect these anomalies and trigger alarms."
    Often, though, the situation is quite a bit more complicated than that. "One company’s larceny is another company’s routine," notes Byrnes. "While copying voluminous files is quite normal for a biotech that must send reports to the FDA, file copying in the archly competitive chip business is a definite no-no. So we’ve developed a range of 'audit' policies covering the financial, banking, insurance, legal, manufacturing and other industries." These knowledge-based monitoring systems can be installed onto a workstation at the touch of a button, and distributed to every network on the computer. In seconds, a comprehensive line of proactive defenses has been applied.
    One can’t help thinking if Kubrick had only known that hard-working, God-fearing little HALs would be such productive members of society three years before his mythical futuristic tale, he might have rewritten the script and had little Hals checking up on big HAL before he went bad!

Gregory McQuerter, CEO of the McQuerter Group, has been providing marketing support to San Diego’s top high-techs for more than a decade.