The Pentagon's Strategic Computing program, widely seen as a response to Japan's highly touted Fifth Generation computer project, is rapidly falling behind schedule because of logistical problems and internal squabbling, according to several Defense Department sources.
The program, which was launched last fall, is supposed to create research breakthroughs in high-speed computers and artificial-intelligence software that would provide a new generation of computing power for the defense and the high-technology communities, according to the Defense Advanced Research Projects Agency. DARPA, which has funded most of the Pentagon's research in computer science, is administering the program.
The program has been described as "essential" for maintaining American leadership in computer technology. Japan's Fifth Generation program, sponsored by its Ministry of International Trade and Industry, is that country's attempt to surpass the United States.
DARPA scientists assert that the new technologies would lead to computers that could literally "think" better, "see" better and calculate faster than any previous generation of computers. One of the key benefits of the program is that the technology's usefulness could be demonstrated quickly and then applied to a broad spectrum of military uses, from "smart" bombs to command and control systems, DARPA has said.
Although the Strategic Computing program is in no danger of collapse, both DARPA insiders and outside computer experts are unhappy with its progress. They contend that it will be unable to meet its goals and that personality clashes within DARPA have sent mixed signals to the research community.
The program already has undergone at least one major reorganization designed in part to minimize the personnel disputes. In addition, four of DARPA's top managers of the Strategic Computing effort said in an interview that they are unlikely to be with the program for more than another year.
According to Pentagon sources, the major conflict has been between Robert Kahn, the director of DARPA's Information Processing Techniques Office, which traditionally has funded computer science research, and Lynn Conway, who was hired last year from Xerox Corp. for a top-level management position with the Strategic Computing project.
As a result of the friction, much of the program was removed from IPTO and placed in the DARPA director's office, and a new manager, A. Charles Buffalano, was brought in to mediate.
Buffalano, Kahn and Conway, as well as a top DARPA scientist, Craig Fields, all say they do not expect to be working on the project this time next year.
The projects themselves also are facing problems. The program plan called for five advanced high-technology projects to be demonstrated within three years. These included an "autonomous vehicle" that could "see" where it was going, a computerized "pilot's assistant" that a fighter pilot could converse with in spoken English to assist in missions, and three battle-management programs intended to use artificial intelligence techniques to aid battle commanders in making decisions.
Only two of those demonstration projects -- the robotic vehicle, which DARPA is very pleased with, and one of the battle management programs -- are on schedule. The other three are "duds," one Pentagon source said. Some of them may be discontinued.
Defense sources also say that few contracts have been awarded under the Strategic Computing program and that most of those that have been signed do not represent investments in new technological initiatives.
According to DARPA, the program has awarded close to $23 million to nine contractors. However, nearly $12 million of that is budgeted for the startup of a special gallium arsenide computer-chip pilot production line, which will be switched next year to the Strategic Defense Initiative -- or "Star Wars" -- section of the Defense Department budget.
Other contracts were for work in progress and simply were shifted to the Strategic Computing program from other DARPA offices. These include contracts with Stanford University for high-speed peripherals; with the California Institute of Technology for computer devices; with Columbia University for new computer machine designs; with Bolt, Beranek and Newman Inc. for multiprocessor machines; and with the Massachusetts Institute of Technology's Lincoln Labs for improved silicon-chip-technology processes.
The new contracts include a University of Maryland project to devise vision systems for the autonomous robots and computer language machines from Texas Instruments Inc. Funds for those contracts were released last month.
Virtually all of the Strategic Computing money disbursed has gone to fund sole-source, noncompetitive contracts. Much of the vital Strategic Computing programs are yet to be awarded and are being competitively bid, including programs in computer design, natural language communication, machine vision, and "expert systems" that allow computers to recreate the decision rules that experts use when making decisions.
DARPA officials say that most of those contracts will be announced before the end of the fiscal year, Sept. 30. Several defense sources say that the delays in completing contract details could hold up some projects by more than a year.
"We paid the price for competitive bidding," said Buffalano, deputy director for research at DARPA. "Getting the bids out has taken longer than expected."
Buffalano said that competitive bidding is one way of expanding the amount of industry participation in the Strategic Computing initiative. He maintains that "building an industrial infrastructure for strategic computing" is a critical part of the defense effort.
Buffalano contends that the program is running on schedule and that the problems facing it are those of "coping with risk" rather than personality questions or the ultimate direction of the effort.
However, internal DARPA documents from May indicate that agency officials have serious questions about how to measure progress in the program, that the perception of the program by the academic research community is negative, and that it is doubtful that disparate technologies can be integrated for demonstrable programs in the alloted time.
DARPA sources said that, prior to the start of the Strategic Computing project, leading researchers at MIT, Stanford and Carnegie-Mellon University were consulted to determine whether producing the demonstration projects in the alloted time was feasible. DARPA was assured its schedules were realistic. Agency sources now concede that is not the case.
The Strategic Computing program isn't the only major computer research effort that is having problems, however. Japan's Fifth Generation effort also is experiencing delays and difficulties in developing advanced software. Europe's Esprit computing program is mired in bureaucratic delays and a lack of cooperation between government and private industry.