Voice Recognition

By MICHAEL KRANTZ

Talking to a computer is always easier than typing, which is why engineers have been working with such fervor on voice-recognition technology. They've made great progress. The accuracy rates of dictation programs from industry leaders--Lernout & Hauspie, Dragon Systems, Microsoft and IBM--have been improving around 10% a year for a decade now. During the next year, voice software geared to their specialized vocabularies is expected to gain a foothold in the medical and legal professions. Popular search sites like AskJeeves and TellMe Network hope to become "voice portals," letting users seek information by speaking simple phrases into the phone.

That's just the start. The next few years will begin the era of "pervasive computing," when the focus of digital software will migrate from desktop PCs linked to the Internet by phone wire to a plethora of newfangled, Web-ready products ranging from TVs and cell phones to dashboards and Palm Pilots. "You'll have the Internet in your pocket, anytime, anywhere," says Kai-Fu Lee, Microsoft's v.p. of user-interface platforms, of tomorrow's wireless and handheld devices. Most of them will be too small to have a keyboard. "The only way you're ever going to get lots of data into small devices," says Dragon Systems founder Janet Baker, "is by talking to them."

Eventually they'll even understand you. The 10-year horizon promises the birth of both natural-language software that "understands" many complex sentences, and broadband data speeds that make online video ubiquitous. The average software product in 2010 could well have a face, a voice, ears and something resembling a brain, which suggests that our next great challenge will be figuring out what we really want to say.

--By Michael Krantz