Roanoke Times
                 Copyright (c) 1995, Landmark Communications, Inc.

DATE: WEDNESDAY, October 13, 1993                   TAG: 9403180007
SECTION: EDITORIAL                    PAGE: A-6   EDITION: METRO 
SOURCE: 
DATELINE:                                 LENGTH: Medium

SHHH

The Christian Science Monitor reports that Microsoft has upgraded its Windows Sound System - the feature that, in its previous lives, has allowed you to record important messages like "Hi, Joe!" for co-workers, who then are greeted by scratchy noises and a muffled "Wawoe" when they fire up their program.

Now, you can tell the computer to "go to sleep" and it will ignore your voice commands, then to "wake up" to return to voice control. Of course, once you wake the thing up, you still have to know what you want the computer to do, and how it is to do it. That's a bug they haven't worked out yet.

The thing can talk back, too. One advantage cited is that voice computer communication can carry with it the expressive nuances missing from written words. A co-worker, who presumably would be out of earshot otherwise, will know how enthusiastic you are about the job she is doing when she hears her computer say, "GREAT job!" rather than reading a flat "great job."

Of course, using capital letters and an exclamation point does help in drab, old-fashioned writing, but ... well, you'll just have to take our word for it when we tell you it sounds just so much more enthusiastic when we say it out loud. Oh, happy day.

Upon reflection, though, this talking feature promises to be a mixed blessing at best. The program's parents speak proudly of how it can read data back to you so you can check figures against your original numbers. But how long will it be, you have to wonder, before it's giving you a lot of lip, like an obnoxious, know-it-all human, pointing out gleefully that you've made a mistake?

We know this kind of thing takes 12 or 13 years to develop in people, but how long is that in computer years? Is there a 7-to-1 ratio, as with "dog years"? In that case, we can expect to be arguing with our computers in less than two years.

Which is certainly cause for alarm. We suggest that before these advances go any further, programmers teach computers to listen in a more sophisticated way. Computers should, for example, be able to follow more than one simple instruction at a time. Instead of closing their ears immediately when told to "go to sleep," they should wait for possible further instructions. Like maybe even, on occasion: "Go to sleep. And don't wake up."