Sifting Software Goes Commercial

FOR MORE THAN 15 YEARS, snoops in the U.S. intelligence community have been teaching "neural networks" to comb through vast streams of electronic traffic. The software excels at spotting tidbits that can tip the balance in political negotiations and military confrontations.

Now, these intelligence-gathering capabilities are heading into the business world. David C. Hoppmann formed Intell.X earlier this year when he realized the technology would be better at filtering business information than intelligent programs known as software "agents." Agents can mitigate information overload by learning to take on routine chores, such as quashing unwanted E-mail, says chief technologist Jeffrey P. Massa, formerly of the National Security Council. But they're based on rules, and can't learn much that's really new. Neural nets start with a clean slate. They can "learn what information fits together, and why, without needing human intervention," says Massa.

Initially, the Intell.X software will help DataTimes Corp., Intell.X's parent, sift through data streaming daily through its DataTimes database and select only the most pertinent information for specific customers. This fall, Intell.X will unveil another program, called Summarizer, that will digest E-mail and create a synopsis to help managers decide which messages to read in full.

Before it's here, it's on the Bloomberg Terminal.