DARPA cuts academic AI funding

Background

In the 1960s the U.S. defense research agency DARPA — then known as ARPA — provided millions of dollars for artificial intelligence research with few strings attached^[1]. For a young field, that meant laboratories could choose their own directions and work on long-term questions rather than on deliverables.

The Mansfield Amendment changes the rules

In 1969 the Mansfield Amendment required DARPA to fund mission-oriented direct research rather than basic undirected research^[1]. The National Academies Press describes the effect as a swift decline in some of the military's support for basic research, often driving it toward the applied realm^[2]. For AI specifically, Roberts and his successors at ARPA now had to justify AI research programs on the basis of immediate utility to the military mission^[2]. The test applied to a project shifted from whether a direction looked interesting to whether it could serve a military task soon.

Two reports question the payoff

Assessments pointed the same way. Lighthill's report and DARPA's own American Study Group both suggested that most AI research was unlikely to produce anything truly useful in the foreseeable future^[1]. According to the National Academies Press, the Lighthill report and its U.S. equivalent led to a shifting of DARPA funds out of robotics research and toward mission-oriented direct research rather than basic undirected research^[2].

Where the money went

The result was not that DARPA stopped supporting AI, but that it supported different things: money was redirected at specific projects with identifiable goals such as autonomous tanks and battle management systems^[1]. By 1974 funding for AI projects was hard to find^[1]. Wikipedia lists the 1973–74 DARPA cutbacks to academic AI research in general among the episodes leading into the first AI winter^[1] — the first of the periods in which funding and interest in AI fell sharply at the same time.

How much actually shrank

One qualification matters. The funding cuts only affected the major laboratories and the largest research projects, which by 1975 were a very tiny fraction of people working in the field^[1]. What was cut, in other words, was the most visible and most grant-dependent part of the field, rather than the field as a whole.

What came after

The pattern of task-based funding continued. The National Academies Press notes that total federal funding for AI research, excluding the Strategic Computing Program, tripled from 57 million dollars to 159 million dollars between 1984 and 1988^[2]. The Strategic Computing Program itself was organized around tasks: DARPA established three specific applications as R&D objectives — a pilot's associate for the Air Force, an autonomous land vehicle for the Army and an aircraft battle management system for the Navy^[2]. Money had returned, but it still came attached to identifiable applications.