Return to main story


It's tedious work, but the software giant dearly wants PCs to gab

Visit Microsoft Corp., and you might think the days of mice and keyboards are numbered. Soon after the new century, Microsoft aims to ring in an era of hands-free computing. PCs, it figures, will be able to converse with users in their native tongue.

Chairman William H. Gates III leaves no doubt about the importance of speech-enabled PCs. ''Computers should see and talk, they should be much easier to use, and they should execute much more powerful searches,'' he says. There's a good reason: Speech could open the floodgates, allowing Microsoft to sell software to millions who find today's machines too complex.

A big chunk of the $2.6 billion Microsoft will sink into research and development this year is earmarked for speech. The software maker has a $45 million stake in Lernout & Hauspie, Europe's leading speech-recognition company. And Gates chose Britain's Cambridge University as the site for his first overseas research center, in part because of its expertise in speech technology. Back at Microsoft's Redmond (Wash.) labs, teaching linguistic skills to PCs is the No.1 priority. More than a third of the 240 scientists at Microsoft Research are focused on tackling issues from simple speech recognition to sophisticated natural-language processing.

Yet, for all that, Microsoft's commercial forays have been modest. There's the talking Barney doll that came out last summer and Auto PC, a family of programs unveiled last fall that give drivers limited voice control over phones and other car electronics. Though cool, these products fall short of state of the art. Auto PC, for example, must be trained to recognize each driver's voice. And it will only recognize words separated by pauses, while products from IBM and others let users talk in unbroken streams. Says Xuedong ''X.D.'' Huang, Microsoft's speech research manager: ''We'll be able to do much more in two years.''

BRONZE MEDAL. Nathan P. Myhrvold, Microsoft's chief technology officer, argues that speech technology, across the board, isn't ready for prime time. The products now on the market, he says, are just not good enough yet. ''If the contest is to ship a marginal product, then I'll take the bronze,'' says Myhrvold.

Instead, Microsoft is aiming at the next generation of speech products. Outside observers familiar with its efforts say it could be worth the wait. ''Their systems are poised to make some big advances,'' says Victor Zue, associate director of the Laboratory for Computer Science at Massachusetts Institute of Technology. And users can count on finding entertainment along with the new technology. Microsoft's animated menagerie includes a wizard character for handling questions and Genie, an onscreen helper with the ability to synthesize speech, which lets it do such things as read items from a Windows menu.

Ultimately, Microsoft hopes to infuse all its products with basic speech capabilities. ''Someday when it's good enough, we want to put it into Windows,'' Myhrvold says. Microsoft declines to forecast when Windows will understand speech. Indeed, Gates fears that the Justice Dept. could interfere with those plans. ''Their approach would block us from ever putting speech recognition into the operating system,'' says Gates. For now, Microsoft is concentrating on making sure third-party speech programs work with Windows.

So when can you throw out your keyboard? Even Myhrvold concedes that jabbering to computers may never be acceptable in crowded offices. But some situations cry out for it. Just watch kids playing video games. ''They shout at it. There's this one-way dialogue going on,'' says Myhrvold. ''The computer needs to understand and talk back.'' We might even end up learning something.

By Otis Port in Redmond, Wash.

Return to main story


Updated Feb. 12, 1998 by bwwebmaster
Copyright 1998, Bloomberg L.P.
Terms of Use