That maturation, however, might be further away than many think. DoCoMo's trial 3G handsets, for instance, operated at 200 Kbps, overheated badly and were recalled. These bugs are being addressed in current models. Even more disappointing, 3G will probably mean speeds in the 256 kbps range -- nowhere near high enough for any kind of decent video experience. As a result, 3G expectations have cooled. The service will dribble into large metropolitan areas over the next five years but won't be available elsewhere until after 2010.
Now that the early fizz has dissipated, vendors are left with the bittersweet taste of stale soda. Less than 10% of AT&T Wireless and Sprint PCS 2G customers have signed up for the wireless Web, and for good reason. Web access via a handset has so far been a largely disappointing experience.
Despite the gloom, 3G continues to have its proponents. In all probability, the telcos will lick their wounds and step back into the 3G fray over the next year or two. After all, wireless subscription figures continue to climb worldwide and even the wireless web is gaining ground. Regardless of big losses to date, they can't ignore the potential.
"Those who treat the arrival of wireless access to the Internet as just another terminal device and protocol to worry about will find themselves as far behind as those that thought the Internet was not relevant to their industry or community," said Simon Hayward, research director at Gartner Group in Stamford, Conn.
Speech Recognition Rhetoric
Just a couple of years ago, speech recognition (SR) was the darling of the investment community. Market leaders such as Dragon and Lernout and Hauspie (L&H) got a little carried away with their enthusiasm. Accuracy rates of 98% and the capability of recognizing more than 100,000 words were going to make the keyboard and even handwriting obsolete within a matter of months. So what happened?
At the height of the Internet bubble it all started to unravel. L&H (which earlier had gobbled up Dragon and other competitors) went bust after year 2000 losses of $1.5 billion. Its fall from grace followed revelations of secret connections to German intelligence, overstated revenues to the tune of $300 million, and imprisonment of top management for insider trading and stock market manipulation.
L&H assets were snapped up ScanSoft, now targeting its SR efforts at call centers. IBM, too, continues to develop SR, as do companies such as SpeechWorks and Nuance. Together, these few still grapple with the real reason for SR's collapse -- failure to deliver the goods.
Many users are disappointed with reported accuracy rates. Despite the hype, the bottom line is that accuracy is often found trailing around 80% to 85%. Dictation programs garble too many words, struggle with punctuation and seem reluctant to understand high voices, people with colds and various accents. Clearly, an evolving technology had been misrepresented as the finished article. It might yet take many years before the technology is even close to the promises made at the end of the millennium.
Check it out for yourself at 1-877-VIAVOICE. I tried to find phone numbers of IBM personnel in its directory assistance. Instead of Silver Spring it heard Santa Fe Springs. Instead of Robinson it gave me Gunther. I did get it to understand Washington as the city, as long as I didn't say D.C. afterwards. Maybe my Scottish accent was too much for it, but based on that experience, I can't see that demo convincing anyone of the value of the technology.
"Two years ago, they were telling us that speech recognition would be ready in two years, but it's nowhere near it yet," said Neil Strother, senior analyst at Cahners-in-Stat, based in Scottsdale, Ariz.
The point about speech-recognition is not that the technology lacked merit. Rather, it needed far more time to mature than the market allowed. Truth be told, it is widely used in niche markets such as directory assistance, call centers and public libraries around the country. It isn't that impressive, but it continues to be used and people appear to be willing to put up with it in small doses.
Despite its shortcomings, though, SR's future may not be bleak. IDC predicts the market will surge to $3.5 billion by 2005, though the analyst firm continues to express reservations.
"Limitations do exist, however, since implementation is highly customized and still relatively high cost," said Steve McClure of IDC.
So what are the great lessons to take away from all these sordid tales of excess and misrepresentation? All have in common a massive investment in hype about technologies that were not quite ready for the big stage. Speech recognition and 3G are not yet technically there, yet were sold for a while as the finished article -- a big mistake that may add many more years to the eventual rollouts. Broadband was in better shape, but suffered from a different delivery shortfall: vendors failed to take into account what it would take to traverse the last 50 feet to the customer.
In reality, that small leap is every bit as large a hurdle as crisscrossing the globe with fiber optic cable. Telcos were more than up to the challenge of laying out fiber optic superhighways, yet have never really grasped the magnitude of tracking individual lines to homes and small offices.
The telcos didn't hire and train enough people for the task at hand. Until they do, broadband will continue to rise in adoption at a far slower rate than it really deserves. Yet the entire field stands ready for someone with their act together to deliver good service fast to everyone who wants it. When someone finally figures out how to do that, they stand poised to eat up much of the broadband terrain. And due to financial conditions, they may even consume the assets of their rivals for pennies on the dollar.