No PriorsThe Future of Warfare: How the US Department of War Thinks About AI
Sarah Guo and Emil Michael on pentagon’s AI push: applied deployment, autonomous systems, industrial-base rebuilding strategy.
In this episode of No Priors, featuring Emil Michael and Sarah Guo, The Future of Warfare: How the US Department of War Thinks About AI explores pentagon’s AI push: applied deployment, autonomous systems, industrial-base rebuilding strategy Emil Michael, Under Secretary for Research and Engineering (and leader of the DoW’s chief AI office), describes a restructuring meant to unify technology efforts across DARPA, DIU, and other innovation arms to move faster and reduce duplication.
At a glance
WHAT IT’S REALLY ABOUT
Pentagon’s AI push: applied deployment, autonomous systems, industrial-base rebuilding strategy
- Emil Michael, Under Secretary for Research and Engineering (and leader of the DoW’s chief AI office), describes a restructuring meant to unify technology efforts across DARPA, DIU, and other innovation arms to move faster and reduce duplication.
- He argues the urgency is driven by China’s unprecedented military buildup and by U.S. supply-chain fragility, requiring new investment levels, new entrants, and faster acquisition pathways for startups.
- A centerpiece is GenAI.mil: a secure, DoW-network GenAI platform deployed quickly to the enterprise, reaching ~1M unique users in 30 days, with multiple models and expansion across classification levels planned.
- The conversation emphasizes a shift toward distributed autonomous systems (drones/robots), scalable production (not just prototypes), and financing mechanisms (DIU, manufacturing scale programs, and the Office of Strategic Capital) to “collapse the Valley of Death.”
IDEAS WORTH REMEMBERING
7 ideasThe DoW is trying to act like a focused product organization, not a committee.
Michael says 14 “critical” tech areas created diffusion; he cut them to six and put “sprints” behind them to drive execution, signaling a cultural shift toward prioritization and delivery.
Applied AI is the #1 priority because the private sector is funding the frontier.
Rather than build a DoW foundation model, the plan is to adapt commercial models to defense use cases, focusing on deployment, security constraints, and integration with DoW workflows/data.
GenAI.mil is a template for rapid, secure AI rollout at massive scale.
Because public chatbots are generally prohibited on DoW networks, they designed a different data-flow architecture to prevent sensitive data from leaking into public model training, then shipped to millions quickly with a “tiger team.”
Defense AI deployment is organized into three lanes: enterprise, intelligence, and warfighting.
Michael frames near-term progress as “pace-setting” pathfinder projects that both deliver value (efficiency, intel fusion, planning/wargaming) and prove feasibility to overcome internal resistance.
Autonomy will take an increasing share of budgets because it’s cheaper per firepower and reduces risk to personnel.
He predicts a meaningful budget mix shift toward autonomous systems (air/sea/undersea) and suggests 20–30% of the defense budget could go to these systems within 10 years as AI enables sensing and independent action.
The bottleneck for startups is scaling flawless manufacturing, not prototyping.
Michael warns founders that selling to DoW isn’t like enterprise sales; beyond a demo, they must show a credible plan to produce at scale (e.g., “ten thousand of these”), especially for hardware-software systems.
Capital mechanisms are being used to shrink the ‘Valley of Death’ between pilot and scale.
He cites DIU’s rapid contracting funds, programs aimed at manufacturing scale, and the Office of Strategic Capital’s low-cost loans (Treasuries + ~100 bps) to expand capacity in brittle supply chains (e.g., critical minerals).
WORDS WORTH SAVING
5 quotesFourteen priorities mean no priorities at all.
— Emil Michael
We’re not building a foundation model at the DoW… So how do I adapt or use what’s being developed in the private sector and apply it to the Department of War use cases?
— Emil Michael
It’s crazy to me that SpaceX, and Anduril, and Palantir all had to sue the Department of War for their first contract. So the idea is, you don’t have to sue anymore. Come through the front door.
— Emil Michael
We’ve had over a million people, uniques, use it in the last thirty days.
— Emil Michael
If you come with a tech pitch and a prototype, that’s great. What’s the plan to get me ten thousand of these?
— Emil Michael
QUESTIONS ANSWERED IN THIS EPISODE
5 questionsGenAI.mil architecture: what specific technical controls ensure prompts/data don’t flow back into public model training, and how is auditing handled?
Emil Michael, Under Secretary for Research and Engineering (and leader of the DoW’s chief AI office), describes a restructuring meant to unify technology efforts across DARPA, DIU, and other innovation arms to move faster and reduce duplication.
You mention “one-third of the enterprise on one model.” What criteria decide which model(s) get approved, and how do you evaluate performance vs security tradeoffs?
He argues the urgency is driven by China’s unprecedented military buildup and by U.S. supply-chain fragility, requiring new investment levels, new entrants, and faster acquisition pathways for startups.
For the three AI lanes (enterprise, intelligence, warfighting), what are the named “pace-setting” pathfinder projects and what measurable outcomes are expected in 6–12 months?
A centerpiece is GenAI.mil: a secure, DoW-network GenAI platform deployed quickly to the enterprise, reaching ~1M unique users in 30 days, with multiple models and expansion across classification levels planned.
On autonomy: what are the main blockers today—policy (human-in-the-loop), reliability, comms in contested environments, or procurement—and which are you tackling first?
The conversation emphasizes a shift toward distributed autonomous systems (drones/robots), scalable production (not just prototypes), and financing mechanisms (DIU, manufacturing scale programs, and the Office of Strategic Capital) to “collapse the Valley of Death.”
Scaled hypersonics and directed energy: what does “at scale and reasonable price” concretely mean (target unit costs, production rates, timelines)?
EVERY SPOKEN WORD
Install uListen for AI-powered chat & search across the full episode — Get Full Transcript
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome