Military-Industrial Origins of Computing

In the AI Story

ENIAC, the first electronic general-purpose computer, was built at the University of Pennsylvania under U.S. Army contract to calculate ballistic firing tables. The institutional context was military, the funding governmental, the problem being solved the projection of lethal force with greater accuracy. No commercial market for electronic computation existed—it was created by military need and sustained by military funding through the long gestation fundamental breakthroughs require. The pattern repeated: ARPANET (Defense Department communications survivability), time-sharing (DARPA-funded interactive computing), GPS (DoD precision targeting), microprocessor acceleration (military demand for compact electronics), and the neural network research keeping the AI foundation alive through multiple commercial winters.

The values military institutions embed in technologies they develop are specific: efficiency (systems must perform under resource constraints), optimization (incremental improvement is continuous requirement), control (outcomes must be predictable and directable), reliability (failure costs lives), and scalability (systems must handle operational demands orders of magnitude beyond laboratory conditions). These values are rational—even imperative—within military contexts. A communications network that fails under stress or a computation system producing unreliable results represents operational catastrophe. But when technologies embodying these values migrate into civilian applications, the values carry consequences their originating context did not anticipate.

The migration is structural, not conspiratorial. Engineers trained in institutional environments rewarding efficiency and optimization carry those priorities into commercial products, treating them as technical requirements rather than institutional choices. The benchmarks evaluating AI systems descend directly from military requirements for measurable, demonstrable capability—the ballistic tables ENIAC computed, the signal detection tasks early pattern recognition systems performed, the intelligence analysis tasks funding natural language processing research. Systems optimized for benchmark performance develop specific capabilities (breadth, fluency, speed, confidence) reflecting benchmark values rather than the values intellectual inquiry or democratic deliberation would prioritize.

Civilian deployment of military-origin technologies requires deliberate construction of new institutional arrangements embedding civilian values. The internet's migration from ARPANET to civilian infrastructure required institutional innovation—new governance structures, new protocols, new conceptions of what the network was for beyond military communications. AI's migration from research laboratories to knowledge-work ubiquity requires analogous innovation: evaluation criteria prioritizing intellectual development over operational performance, design principles preserving cognitive autonomy over engagement maximization, regulatory frameworks protecting democratic deliberation over efficiency optimization. These civilian values do not emerge from military-industrial genealogy—they must be deliberately constructed through institutional effort.

Origin

Smith's Military Enterprise and Technological Change (1985) documented the pattern across multiple domains, establishing that military institutions function as technological catalysts not through direct production but through sustained funding of research that commercial markets abandon during long development cycles. The finding challenged both the garage-inventor mythology and the market-driven innovation narrative, revealing that the state—specifically, the military state—was the primary engine of American technological development in domains from precision manufacturing through computing.

The implications crystallized in Smith's later work: technologies developed in military contexts for military purposes carry military values into civilian applications, often invisibly. Understanding these genealogies is essential for evaluating whether civilian deployment serves civilian needs—a question the myth of the neutral tool prevents from being asked.

Key Ideas

Military funding sustained computing through commercial indifference. ENIAC, ARPANET, time-sharing, GPS, neural networks—all developed in military-funded research when no commercial market justified investment, establishing institutional origins that shaped civilian applications.

Technologies carry institutional values across contexts. Efficiency, optimization, control, reliability, scalability—values rational for military systems—persist in civilian AI, shaping characteristics users experience without recognizing their origins.

Benchmarks descend from military evaluation criteria. The standardized tests measuring AI performance inherit the military's demand for measurable, demonstrable capability, rewarding operational values over intellectual or democratic ones.

Engineers trained in military contexts carry priorities forward. The migration is structural: engineers learning their craft in efficiency-rewarding environments treat efficiency as technical requirement rather than institutional choice when building commercial products.

Civilian deployment requires deliberate value-embedding. Military-origin technologies do not automatically serve civilian needs; institutional innovation must construct evaluation criteria, design principles, and governance frameworks embedding civilian values.