Bill Gates Talks Up Next 'Holy Grails' of Computing

by Alex Woodie

Speech recognition software and visual application development techniques will be two of the next holy grails of technology to fall to human ingenuity, Microsoft chairman and chief software architect Bill Gates said this week at the Gartner Symposium 2004 conference in San Diego, California. An important factor driving these developments will be the huge increases in processing power available from computers, which Gates said will be "almost free" by then.

"Many of the holy grails we've worked on in the last 30 years will be solved in the next 10 years," Gates told his interviewer, Gartner chief executive Michael Fleisher, during a packed Mastermind Keynote session. Some of these holy grails are speech recognition, tablet computing that uses handwriting recognition, forecasting applications, and the model structure of application development, he said. Moore's Law--the doubling of the performance of processors every 18 months or so--will continue to be a key driving factor for processor-intensive workloads such as speech recognition, Gates said. "Ten years out, in terms of actual hardware costs, you can almost think of hardware as being free," he said.

When those improvements in speech recognition software happen, it will enable an entirely new way of interacting with computers, Gates predicted. History has shown an exponential rate in errors when computers try to decipher human speech, he said, implying recent advances have brought us rapidly closer to the section of the curve where error rates are sharply reduced. Last week, Microsoft launched Speech Server 2004, a software product for helping companies launch automated call centers. While the technology is not quite there yet, Microsoft knows how to address the problems with speech recognition, and will have done so by the end of the year, Gates said.

Tests show humans still hold a 2-to-1 advantage over computers in their ability to accurately recognize words played from a random word generator, Gates said. This advantage can be attributed to certain factors such as background noise and context. Gates called this the "wreck a nice beach" syndrome. Instead of "recognize speech" computers will translate that phrase as "wreck a nice beach."

But we've all heard this before. Microsoft has said in the past that it intended to solve the speech recognition problem, but didn't, and Gates acknowledged this. "People should be skeptical. People got up on stage" and said they would solve this. But "we're putting our money where our mouth is," he adds. For example, Microsoft software recently outperformed a human dictation typist in China, Gates said. Of course, he admitted, it didn't hurt that the typist was using a Chinese keyboard, a fact which delighted the crowd. (Chinese keyboards have the equivalent of 96 different character keys, much more complex than the 26 character keys used in the West, and it generally takes three keystrokes to type a single simplified Chinese character.)

Besides speech recognition, another holy grail of technology that Microsoft is chasing is creating new applications visually by rearranging and connecting blocks of code that fit together as components. Today, this practice is referred to by many names, including process modeling, which is high on Microsoft's, and Gates', list of priorities.

In the next 10 years, Gates predicted that the amount of code being written will decrease by a factor of five. "They key breakthrough in coding is to write less code," Gates said. Instead of having programmers write code, Gates said, companies will employ business analysts who will use "glue code" to visually stitch together components of business applications. This modeling software and "glue code" (which undoubtedly will be delivered in .NET, although Gates didn't say so) will all be standard and be delivered as a commodity in a $50 operating system (which we all know is Windows, even though Gates didn't say it). "This is what the IT industry owes to its customers," Gates told Fleisher. This is the vision of object-oriented programming (OOP) that really took hold with the Smalltalk language decades ago. This OOP approach has been borrowed by various C compiler creators to create C++, has been borrowed by Sun to make C into Java and J2EE, and has been borrowed by Microsoft to make Visual Basic and C into C# and .NET. While the idea of modular computing is very old, we're still talking about it more than doing it largely because the computer power that OOP needs has not been affordable and the mindshift from monolithic programming to OOP is very tough to make.

This modular vision of application development's future will go nowhere without standards, which is where extensible markup language (XML) and Web services come in. "Web services are super, super important," Gates said. "There's never been a protocol to allow software to connect to software. XML is a starting point. XML is such a phenomenal technology. It's everything you dream about what it would become."

Gates said the partnership that Microsoft and IBM formed in 2001 to lead the development of Web services standards has been critical to making the future of Web services a possible reality. "We both put super good people on it," Gates said. He also acknowledged that the Big Blue partnership is not all rosy. "We compete on having the best tools, the best implementation of Web services."

During his one hour conversation with Fleisher, Gates touched on several other topics, including Microsoft's $6.8 billion research and development budget for 2004, which Gates says is split 25-75 to research and development, respectively. Fleisher asked pointedly whether "Longhorn," the much-anticipated and oft-delayed kicker to Windows 2003 on the server and Windows XP on the desktop, was considered research, and Gates quickly responded that it was, indeed, covered on the development side of things. At $6.8 billion, Microsoft's R&D budget for 2004 would, for the first time, be larger than that of industry juggernaut IBM, which spends around $5 billion a year on myriad projects and products. Microsoft, which is less than half the size of Big Blue, spent $4.3 billion on R&D in its fiscal 2003 year.

Gates also explained Microsoft's position on outsourcing, which is somewhat unusual among large software development companies in the fact that Microsoft is not aggressively pursuing outsourcing. (Of course, most companies don't have $50 billion in the bank.) "We're not interested in doing Longhorn 20 percent cheaper. We're interested in doing it six months earlier," Gates said. "We do have India and China development groups . . . [but] the bulk of Windows development and testing is going to stay in one place."