Thank you very much, Craig [Martell], and good afternoon everyone.
You know, between executive orders and strategy rollouts and summits and hackathons and symposia like this one, we have all been doing a lot of AI-ing lately.
So I want to be clear: whenever I speak at a podium, there will always be a human-in-the-loop. (Laughter.) I know.
I know this community has been on quite the journey over the last few years.
In some ways, it's a journey that began before many of us were born — even before Moore's Law was etched into tablets of silicon handed down from the mountaintop.
You're building on a legacy that goes back to the notes and algorithms of Ada Lovelace.
To the code-breaking of Alan Turing, and the code-making of Admiral Grace Hopper.
And to the 1960s vision and decisions of J.C.R. Licklider, who directed DARPA's funding of the first AI academic research hubs at Stanford, MIT and Carnegie Mellon.
It's a legacy that continued through the post-war, post-Cold War, and post-9/11 eras — over 60 years of DoD investing in AI and fielding data and AI-enabled systems — from air defense networks to logistics planning tools.
And it's a legacy of sparking breakthroughs that tech companies advanced and innovated on, transforming them into products that millions of Americans use every day.
Of course, increasingly over the last dozen years or so, advances in machine learning have brought new generations of AI innovation. Much of it, happening outside of government.
So our task in the Defense Department today is to adopt those innovations wherever they can add the most military value. And that's been a priority for Secretary Austin and for me since day one.
We knew we had to build rapidly and responsibly — iterating and investing to deliver a more modernized data-driven and AI-empowered military now.
There was no debating the "why" — it's because these technologies give us an even better decision advantage than we already have today. And that's imperative given the pacing challenge we face from the People's Republic of China.
In deterring and defending against aggression, AI-enabled systems can greatly improve the speed, quality, and accuracy of commanders' decisions — which can be decisive in deterring a fight, and winning a fight.
But also in managing the world's largest enterprise. We've got nearly 3 million people on the payroll; a health care system serving over 9 million troops, retirees, and family members; and assets spread out worldwide over 25 million acres, roughly the size of Kentucky. The value of DoD's assets worldwide is larger than Amazon, Apple, Microsoft, and Walmart combined — by a long shot. At this scale, we must leverage data and AI to be smarter, faster, and better stewards of taxpayer dollars.
So in 2022, informed by commercial best practice, we integrated four disparate digital, AI, and analytics teams under an empowered Chief Digital and Artificial Intelligence Officer (CDAO) who reports directly to the Secretary of Defense.
At its core, CDAO's mission is to accelerate DoD's adoption of data, analytics, and AI from the boardroom to the battlespace. And in a short time, we have seen remarkable progress.
Thanks to this community, we have data decrees that all DoD data must comply with. Because data interoperability, access, and trustworthiness are critical for AI, for doing command and control across all domains, and for being a modern Defense Department. No excuses.
Beyond that, our ADA initiative deployed data scientists to every Combatant Command. And we've brought that model into the Pentagon. Every component of the Office of the Secretary of Defense is receiving analytic product teams to accelerate their own data modernization.
Meanwhile, Task Force Lima is digging into large language models and generative AI tools.
Through teamwork across DoD, we've worked tirelessly to continue being a global leader in the fast, responsible development and deployment of AI technologies in the military sphere: creating policies appropriate for their specific uses and leading in — on — international cooperation.
And we're making the necessary investments to enable and scale data and AI across DoD — with cloud computing contracts, network infrastructure, data integration, faster ways to buy commercial software and keep it up-to-date, and more.
All of this is helping to realize Combined Joint All-Domain Command and Control. Of course, CJADC2 isn't a platform, or a single system. It's a fusion of concepts, technologies, policies, tools, and talent that's advancing how we command and control forces with key allies and partners.
Over the last year, CDAO and its partners have used a series of global information dominance experiments to set a blisteringly-fast pace for this work: every 90 days, we're iterating on capability development and delivery. And we'll be keeping it up in 2024.
Now, last August, I said that CJADC2 is not some futuristic dream, and these investments are rapidly yielding returns to the warfighter — making us even better than we already are at joint operations and combat integration. Today, I want to share a bit more.
You see, that summer, I had challenged CDAO and its mission partners to deliver a minimum viable capability for CJADC2 by the end of 2023. For security reasons, I can't say where or what that capability is for, but I can tell you it was no easy task — especially in six months.
But with a lot of hard work across many teams — pairing operators across multiple commands with engineers from DoD and industry — they delivered, on time and on target — combining software applications, live data integration, real-world networks, new cross-domain operational concepts, and warfighters around the world, to provide even better decision advantage for DoD and our military commanders.
The minimum viable capability for CJADC2 is real and ready now. It's low latency. And extremely reliable.
That's the beauty of what software can do for hard power. Delivery doesn't take years or a decade. Our investments in data, AI, and compute are empowering warfighters today.
So, the software rebels have won. The message they carried is clear. But for the rebels to govern, to achieve the change we need, we must move from cementing our foundations to building and deploying an enterprise data and communications architecture.
That means acquiring where we can, developing and fielding new solutions where we need to, and ensuring we have an integrated, interoperable framework that delivers the capabilities our warfighters need.
I've said it before: we should be unafraid to shift approaches as we evolve — our goal is to be ahead of the curve, not chasing the curve. CDAO will play a critical role in this next phase of our development, but so too will our operators and research, development, and acquisition professionals throughout the Defense Department.
These efforts are advancing U.S. capability in material ways, but they are not signs of an "AI arms race." That's not how we see it. Nor do we seek such an arms race.
After all, data and AI are not weapons. They're fundamental, general purpose technologies — like electricity. And history is instructive: with electricity, the greatest military utility was not in electrocuting enemy troops. (Laughter.)
Instead, electricity's value came in capabilities that, looking back, seem almost innocuous: Radios. Radar. Semiconductors. Satellites. Night-vision goggles. Data links. Battle networks. Precision-guided munitions. Avionics that enabled dagger-shaped stealth bombers to fly. WiFi.
All things that were so fundamental and transformational to the military, at first they seemed like magic. Today, the technologies continue to evolve and improve. Yet they've become so ubiquitous, many people now take them for granted. And they're almost boring.
The same is true for data and AI. The most transformational use cases may in fact seem boring. But they are essential because of what they enable for the warfighter and all who support them.
And that means the advantage in data and AI goes to the country that uses them better, faster, smarter, and safer.
Today, that is the United States of America. We have better chips, better talent, better tech, and better values that guide how we use data and AI. And we intend to stay in the lead.
Our advantage comes from who we have building them, who we have using them, and how we do so. And in these arenas, America will always be unbeatable.
And just as with other foundational technologies, we've got to keep improving, to make AI better, more secure, faster, and smarter — and safe, too. It's a DevSecOps mindset.
CDAO is a vital engine as we continue accelerating change with urgency, from the E-Ring of the Pentagon to the tactical edge, to deliver at speed and scale.
Of course, we can't do it without resources. One of our combatant commanders recently reached out to me noting that the advances they are relying on from CDAO are dead in the water without our fiscal year 2024 appropriations. So, we need Congress to come together and pass appropriations for 2024, ASAP. It is long overdue, and the delay is devastating.
We're also hamstrung by our internal culture. You see, our processes date back to the 1960s, when defense dollars were the main driver of American innovation, and the private sector less so.
The script flipped long ago — with commercial tech companies now doing the lion's share of innovation, especially with data, digital analytics, and AI. Yet Pentagon practices didn't keep up, even though it's been decades since DoD was alone at the tip of the innovation spear.
So for three years, we've been taking a comprehensive, iterative, warfighter-centric approach to defense innovation writ large — recognizing we face an accumulation of challenges and barriers, and there is no silver bullet that will lower them all. And that, too, requires a DevSecOps mindset.
Because of you, and many others, we are succeeding in overcoming our historical baggage. It's a constant effort, but every bit is necessary to drag DoD into the modern era.
Never get complacent, though. It's a protracted campaign. Technology will keep changing. Today's good habits and best practices will one day be outdated and entrenched. And our global competitors will also continue to advance their capabilities. So we must also be constantly learning, growing, evolving, and pushing ourselves.
These competitors should know: we're in this for the long haul. After all, the PRC pacing challenge is a generational challenge. It will change over time, but it is not going anywhere. But that's okay. Neither is America.
As a nation and a military, we are deeply committed to staying ahead of our competitors — today, tomorrow, and for the foreseeable future. No matter how long it takes.
Within the Defense Department, we know more work lies ahead.
We have to get better at data sharing and accessibility, and keep pushing to make all our data clean, labeled, trustworthy and secure. That is foundational to the AI hierarchy of needs.
As we take in new data, we have to continually update our models and algorithms, to account for changes in the world, in our competitors, and in the operational environment.
And not just with technology; we must also continuously improve how we recruit and retain top talent that's in high demand by the private sector.
We're fortunate to grow our own talent across the joint force, and to have so many brilliant data scientists and software engineers and analysts on our team — many in this room — who are world-class technical minds, and choose to serve despite many other career choices out there.
And just as captains of commercial industry left the C-Suite during World War II to help us out-build the Axis powers, today we are also fortunate that top tech talent from the private sector has increasingly chosen to come work for and with us — leaving lucrative jobs at places like Lyft, Apple, Google AI, and others.
Every one of them sees and knows the stakes of this strategic competition that we're in with the PRC. And it's not just patriotism that motivates them. It's a genuine desire to serve. They feel a responsibility to give back to the country that gave them so many opportunities.
Because they know that so much of America's innovation success was made possible by America's democratic institutions, and government investments, and good schools, and equal rights, and the blanket of liberty that safeguards us all.
Patriotism and the call to public service are powerful things. But they're not all-powerful. At least, not for everyone.
Pay matters. Workforce culture matters. Team-building matters.
How we treat people matters. Training, professional development, time with family — it matters.
Having tools and IT you can use to do your work without tearing your hair out really matters. (Laughter.)
Getting people in the door. Reducing time-to-hire by using all the flexible authorities Congress gave us. Jobs that offer people meaning and purpose. Making sure everyone can take leave and recharge every once in a while before they get back in the game. All of that matters.
And just as we have a DevSecOps mindset for the capabilities and products we deliver, we're applying that to our organizations, too. To continually learn, improve, and iterate.
It's work that never ends; and again, it never should. Getting it right takes active management, and leadership at all levels.
And I want to be clear that we care about it at the highest level. So I am grateful for all the work that's currently underway — even as I know we have a lot more to do as a department.
Let me close with this: Governing at the enterprise level is wicked hard. DoD is the most complex organization in the world. And powered by clean data and talented people, CDAO is paving the way for how we're bringing to bear digital analytics and AI at speed and scale.
Yet even as we're embedding these technologies in many aspects of our mission — from battlespace awareness, cyber and reconnaissance, to logistics, force support and other back-office functions — we do so mindful of the potential dangers. And we're determined to avoid them.
We don't use data and AI to censor, repress or disempower people — just as we don't seek to control innovation or make it toe the party line.
Indeed, our country's vibrant innovation ecosystem is second-to-none precisely because it's powered by a free and open society committed to responsible-use values and ideals.
By putting those values first and playing to our strengths — our people foremost among them — we've taken a responsible approach to data and AI that will ensure America continues to lead.
Now, at a time when people make snap-judgments based on something like a "p(doom)" stat, I know such nuance can easily get lost.
But here's the thing: at the Pentagon, we don't fit neatly into square boxes or round holes. And we reject the false dichotomy that AI is either all good, or all bad.
Data and AI are necessary to empower our warfighters. We have to capitalize on that. We have to move swiftly, since both technology and our competitors are also moving swiftly. And it must be safe and secure to be effective — because if it's not, then it's not useful to the warfighter.
So we need speed and safety. We have to be responsible and rapid.
We don't have the luxury of choosing one or the other. It has to be both.
For the good of the nation. For the good of our mission. And for the good of the world.
Thank you very much.