Military Funding

As Tyler Folsom points out in his essay, "The Search for an ŒElectronic Brain': An Introduction to Neural Networks," the Defense Advanced Research Projects Agency proposed in 1988 an eight- year, $390 million neural network research program. DARPA identified seven defense applications that it believed could use neural networks, all of which were complex sonar, radar, or infrared sensor systems. The director of DARPA's Tactical Technology Office, Janet Lupo, announced proudly, "The future of machine intelligence is not AI" (Folsom 301), and considering how new connectionism was at the time (the Bible of neural network research, Rumelhart and McClelland's Parallel Distributed Processing, had come out in 1986), the claim that AI had been upset by a new paradigm was not too ludicrous a claim. DARPA claimed that it hoped in five years to have used neural networks to create a system with the intelligence of a bee: "Bees are pretty smart compared to smart weapons," said deputy director Dr. Craig Fields. "Bees can evade. Bees can choose routes and choose targets" (NY Times, Aug. 18, 1998, quoted in Folsom 301).

The question of whether military spending on AI research is complicated by the number of previous ill-considered military expenditures on computer science research: the Strategic Computing Program, for instance, spent its money looking into specific military applications of AI without first performing the necessary general research, and the Strategic Defensive Initiative turned out to be a huge pit to thrown money into, an impossible project considering the form of the system and the performance of computers at the time. On the other hand, DARPA is immediately responsible for the birth of the Internet, which began as a network to connect educational institutions with defense grant money and used a massively interconnected packet-switching scheme for communication in order to maintain the network's integrity in case it was bombed.

This question must be addressed by each AI researcher individually; although defense money is highly available for research money, the Defense Department seems to historically have overoptimistic or misguided assumptions about the possibilities and applications for computer science research. Research money for AI projects can be difficult to come by, and the choice must be left up to the researcher as to whether taking money from a given source for a given project is an ethical decision.

back