In this Rise of AI Virtual Chat from August 2020, Fabian Westerheide interviews Dr. Ulrike Franke, policy fellow at the European Council on Foreign Relations, about AI and the future of warfare. Franke argues AI will not end war, drawing a parallel to the earlier hope that nuclear weapons would make conflict too terrible to happen, and insists new military technology has historically added capability rather than removing human cost. She explains that military AI spans a full range from simple automation to fully autonomous 'terminator' weapons, citing Google's Project Maven surveillance work for the Pentagon as a real but non-lethal example. She sets out five reasons militaries want AI-enabled systems: speed, stealth, cost and efficiency, fewer personnel needed, and entirely new capabilities such as drone swarming. The conversation covers slow military procurement, the shift of cutting-edge research from state-funded to civilian labs, the U.S. Marines dropping Chinese-made DJI drones over data-security concerns, debates on keeping humans in the kill chain, the unlikelihood of a global autonomous-weapons ban, and Europe's under-studied but real military AI activity and defense posture.
Key takeaways
- Franke argues AI will not end war, comparing the claim to the earlier, unfulfilled hope that nuclear weapons would.
- She lists five reasons militaries want AI: speed, stealth, cost and efficiency, fewer personnel, and new capabilities like swarming.
- The U.S. Marines and other forces dropped Chinese-made DJI drones over concerns that flight and video data could reach China.
- Research on chess shows human-AI 'manned-unmanned teams' currently outperform AI acting alone, an argument for keeping humans in the loop.
- Franke expects Geneva talks on banning autonomous weapons to clarify definitions but not produce an effective ban.
Detailed notes
Will AI end war?
Asked directly whether AI will end wars, Franke answers no, comparing the idea to the earlier hope that nuclear weapons would make war too terrible to occur. She says new military technologies historically add on top of existing capabilities rather than replacing the human cost of war, and rejects the idea of machines fighting it out on humanity's behalf, insisting there will still be, as she puts it, a young soldier dying somewhere in the mud.
What autonomous weapons look like today
Franke describes military AI as a spectrum from simple automation, such as stabilizing a drone in wind, to fully autonomous systems that can engage targets on their own. She points to Google's Project Maven, in which Google worked with the Pentagon to use AI to scan footage from U.S. drones for improvised explosive devices and other targets, as a real but non-lethal, surveillance-focused example (intelligence, surveillance and reconnaissance, or ISR). Westerheide adds examples from the French national AI defense strategy, including communication, collaborative combat swarms, mine detection, land robotics and connected-soldier exoskeleton suits, and raises an unconfirmed report that ISIS hijacked drone GPS links; Franke responds that non-state actors are increasingly capable of this, citing Hezbollah hacking into Israeli drone video feeds.
Five reasons militaries want AI
Franke lists five reasons militaries worldwide are interested in AI-enabled weapon systems: speed, since computers can analyze far more data far faster than humans; stealth, since removing a drone's uplink and downlink to a human operator, which can be detected, hacked or jammed, makes autonomous drones harder to trace; cost and efficiency, though she calls this debatable since some AI systems cost more, not less; the need for fewer personnel, noting armed drone operations can currently require over 100 people and that Boston Dynamics-style robots can carry significant physical loads for soldiers; and, most importantly to her, entirely new military capabilities such as swarming, connecting many airborne or underwater drones to overwhelm air defenses or create what she calls flying minefields.
Procurement, civilian tech, and the DJI drone problem
Responding to an audience question on slow military procurement, Franke says the problem is universal, not just German, and traces a historic shift: since the 1950s, cutting-edge technology such as GPS, the internet and Siri came out of state-funded, often military-linked research (she cites DARPA), but AI development now originates mainly in civilian Silicon Valley companies, forcing militaries to buy from civilian developers on civilian timelines. She warns that Silicon Valley's 'we just build tech, not for the military' stance is inadequate once a system can clearly be repurposed by state or non-state actors. As an example of the risk of buying off-the-shelf civilian tech, she cites the German, Dutch and American militaries adopting Chinese-made DJI drones for their strong price-to-performance ratio, before the U.S. Marines and others stopped using them over concerns that flight and video data could be sent to China.
Keeping humans in the kill chain
On where AI should stop in the 'kill chain' of data collection, analysis, targeting and engagement, Franke argues there are good reasons not to let AI make the final decision to kill a human being, framing this as a civilizational red line even though full autonomy would be faster. She points to research on 'manned-unmanned teaming', drawing an analogy to chess, where Garry Kasparov lost to IBM's Deep Blue in 1997 but human-computer teams have since outperformed AI alone, as evidence that keeping a human in or on the loop is not just an ethical preference but currently produces better outcomes than AI acting alone.
Bans, Europe's defense posture, and the fog of war
Asked about a global ban on lethal autonomous weapons, Franke says talks at the U.N.'s Convention on Certain Conventional Weapons in Geneva are valuable for clarifying definitions but unlikely to produce an effective ban, since autonomous weapons, unlike landmines, are not a distinct enough category to prohibit cleanly, and she is genuinely worried about arms-race dynamics if one state builds such weapons and others feel compelled to follow. On surveillance, she explains how making the oceans 'transparent' through sensors and AI forecasting could locate nuclear-armed submarines and undermine second-strike deterrence, an unintended consequence of more surveillance rather than an unambiguous safety gain. She argues research has focused too narrowly on the U.S. and China while genuine European programs, such as the Franco-German-Spanish Future Combat Air System and the Neuron and Taranis combat-drone projects, are understudied. She dismisses a 'European army' as unrealistic short of a United States of Europe, but assesses that NATO and EU mutual-defense commitments currently make Europe reasonably defensible, while cautioning against complacency and calling on entrepreneurs and investors to consider European defense-technology startups.
Quotes
As much as I like this scenario, although it's a bit tricky, I absolutely do not believe in it.
Ulrike Franke
In the end you always have the kind of 18-year-old soldier dying somewhere in the mud.
Ulrike Franke
I don't think we're gonna have anything resembling a European army until there are a United States of Europe.
Ulrike Franke
Questions answered in this talk
Will artificial intelligence end war?
According to Dr. Ulrike Franke at this Rise of AI Virtual Chat in 2020, no: she compares this hope to the earlier belief that nuclear weapons would make war too terrible to happen, and argues new military technology historically adds capability without removing the human cost of war.
Why did militaries stop using DJI drones?
According to Ulrike Franke at this Rise of AI Virtual Chat, several militaries, including the German and Dutch navies and the U.S. Marines, bought Chinese-made DJI drones for their price and performance, but the U.S. Marines and others later stopped using them over concerns that flight and video data could be transmitted to China.
What are the main reasons militaries want to use AI?
According to Ulrike Franke at this Rise of AI Virtual Chat, the five main reasons are speed, stealth, cost and efficiency, the need for fewer personnel, and access to entirely new military capabilities such as coordinated drone swarming.