“How One Marine Colonel Forced AI Into America’s Wars, And What Project Maven Reveals About India”

 

 Origins of Project Maven

• Initiated in 2017 by Marine Colonel Drew Cukor after witnessing AlphaGo’s victory over a human champion.
• Inspired by the idea that human + machine collaboration outperforms humans alone.
• Aim: integrate AI into US military operations to close the gap between firepower and poor information systems.

The Problem: Data Chaos

• US forces relied on Microsoft Office tools and fragmented defence software, often unreliable in combat.
• Example: GPS reset error in Kandahar (2001) led to friendly fire casualties.
• Cukor concluded: America collected vast data but failed to use it effectively.

Building Maven

• Officially named Algorithmic Warfare Cross-Functional Team.
• Initial funding: $40.8M scavenged from Pentagon reserves.
• Recruitment: Marine reservists for grit, startups for innovation.
• Strategy: horse racing model – multiple startups competing on 90-day cycles.
• Resistance: Air Force, Marine Corps, and Google employees opposed AI weaponisation.

Industry Partnerships

• Startups like Clarifai shifted from wedding photo recognition to drone footage analysis.
• Google withdrew after employee protests; Microsoft, Amazon, and Palantir stepped in.
• Palantir revived through Maven, later becoming a top defence contractor.

 Field to Learn Doctrine

• AI deployed before maturity → improved through real-world use.
• Early failures in Somalia (2017) corrected by retraining models.
• Ukraine war (2022): rapid retraining boosted accuracy, compressing the kill chain from <100 to >1,000 targets/day.
• By integrating large language models, targeting capacity increased fivefold.

Internal Resistance

• Pentagon culture resisted “broken software” deployment.
• Cukor faced investigations and reprimands, ending his career as colonel.
• His wife noted: “Every Marine learns the Corps will never love you back.”

China Factor

• Fear of losing to China drove Maven.
• China adapted US concepts, compelled tech firms to cooperate, and fielded capabilities faster (7 years vs US 16 years).
• As Cukor said: “If you want to see the cutting edge of AI, you go to Shanghai.”

Lessons for India

• Industry gap: India lacks Silicon Valley-scale AI firms; defence sector relies on DRDO, PSUs, and imported kit (“screwdriver-giri”).
• Acquisition gap: Defence Acquisition Procedure (2020) too slow; startups die between prototype and production. iDEX/ADITI funding too small (Rs 450 crore).
• Cultural gap: India’s military is risk-averse, zero-defect, unlike US willingness to field immature AI.
• Data gap: Indian military still digitising legacy paper records; fragmented and siloed.

Conclusion

• Maven shows what preconditions are needed:
• Industry base for AI.
• Procurement flexibility.
• Tolerance for failure.
• Structured, accessible data.
• India is “not yet in the room” – still building foundations before it can ask the deeper question: “Are we the best custodians of this tech?”