Law and regulation · October 28, 1983
Strategic Computing: DARPA's plan
On 28 October 1983 the US Defense Advanced Research Projects Agency (DARPA) set out its plan for a Strategic Computing programme: to create 'machine intelligence' within a decade and demonstrate it in three military applications — an autonomous vehicle, a pilot's associate and a battle management system for a carrier battle group. The first five years were estimated at approximately 600 million dollars.
Why it matters
Artificial intelligence became a line of the defence budget with its own schedule and demonstrations: vision, speech, language understanding and expert systems were to work together in a machine rather than in separate papers. The plan counted how much computation was missing and tied intelligence to parallel architectures and microelectronics.
The plan's cost table: 50 million dollars in fiscal 1984, 95 in 1985, 150 in 1986, with 1987-1988 to be set by progress. At the peak, about 150 million a year; about ten technology communities and five to ten application communities are needed, each of over a hundred people, for eight to ten years. For the autonomous vehicle's vision the plan counts 10-100 billion instructions a second against 30-40 million in the most powerful machines of the day; for battle management, continuous speech recognition with vocabularies of one to ten thousand words. The vehicle is a land vehicle only as an example: the plan allows an underwater one instead.
What the record does not claim. The plan does not name Japan: the words Japan, Fifth and ICOT do not occur once in the text, and foreign programmes are one paragraph of an appendix whose list is open only to readers with clearance. So the record does not say the programme was an answer to the Fifth Generation. How much DARPA finally spent is beyond the plan, which counts only the first five years.
Event record
- Event date
- October 28, 1983
- Timeline date
- Event date
- Verification
- Sources gathered automatically · September 25, 2026
- Lines
- ID
- evt-0831
The date on the title page of DARPA's plan.
Related events
- Related Japan's Fifth Generation project
The second state programme for machine intelligence in the same decade; the plan itself does not name the Japanese one.
- Related PROSPECTOR finds molybdenum
Among expert systems already at work the plan names the geological evaluation of sites, which is what PROSPECTOR did; the system itself is not named in the plan.
Records that link to this one
- Related Britain's Alvey Programme
The American plan of the same year, with money from a different place: a defence agency rather than industry and universities together.
- Builds on Resource Management: a shared speech recognition test
The database, in its authors' words, was designed and recorded within DARPA's Strategic Computing speech recognition programme; its naval domain matches the application the plan set for the Navy.
The DARPA 1000-word resource management database for continuous speech recognition (P. Price, W. M. Fisher, J. Bernstein, D. S. Pallett), ICASSP-88, New York, 11-14 April 1988, vol. 1, pp. 651-654 - Builds on ALV: the Strategic Computing land vehicle drives itself
Leighty's report names the land vehicle as the first of three military applications of DARPA's Strategic Computing programme, begun in October 1983, each with yearly demonstrations.
DARPA ALV Summary (Robert D. Leighty), U.S. Army Engineer Topographic Laboratories, ETL-R-085 - Builds on FRESH: an expert system at the Pacific Fleet headquarters
DARPA's budget of February 1987 lists FRESH among the military applications of the Strategic Computing programme, whose 1983 plan promised a battle management system for the fleet.
Justification of Estimates for Fiscal Years 1988 and 1989, Submitted to Congress, February 1987 (Defense Advanced Research Projects Agency), RDT&E descriptive summary, project ST-10 Strategic Computing - Builds on Pilot's Associate: a minute of flight, two hours of computing
According to the Wright Laboratory paper, the Pilot's Associate programme began in February 1986 as an application demonstration for DARPA's Strategic Computing programme.
Pilot's Associate: Evolution of a Functional Prototype (Carl S. Lizza, Sheila B. Banks, Michael A. Whelan), in AGARD CP-499, Machine Intelligence for Aerospace Electronic Systems