How to Build a Robot Army
Making DoD's Replicator Initiative succeed where others have failed
The aftermath of the Global War on Terror, the ongoing War in Ukraine, the lingering threat of a PLA invasion of Taiwan, and money; these are the ghosts of war past, present, future, and forever. The GWOT haunts us in our training, our equipment, our mindset, and for many, in memories of the fallen. In Europe, the bloody, low and high tech battles of attrition in Ukraine dominate headlines and the desks of analysts everywhere. Globally, we’re in a cold war with the PRC, and in the Pacific specifically, both sides are scrambling to prepare for war over Taiwan (among other flashpoints). At home, there is no longer a singular party of national security. There is a consensus to “do something” RE: the PRC, and not much else. Congress won’t appropriately fund the budget the DoD needs, every department in Washington seems to have its own uncoordinated strategy for dealing with the PRC specifically, and a not insignificant portion of the population thinks Russia is one of the good guys. Planners are staring at Ukraine with one eye and Taiwan the other, while their hands are tied with the only dollar bills left in Washington for new defense spending. The DoD budget is massive, much of it spent on systems we already have, systems that are aging, exquisite, and armed with half-empty magazines. Faced with these challenges, what’s the Pentagon to do? Recently, we got an answer: we’re gonna build a robot army.
Now, it would be extremely hypocritical of me to be nothing but skeptical of the DoD’s Replicator program. One of the primary themes of my Prometheus Award-nominated novel, EX SUPRA, is that it would be incredibly harmful to national security to neglect the development of weaponized artificial intelligence and warbots. Man-machine teaming is the next evolution in warfare, and it is objectively a Good Thing for the Pentagon to pursue and accelerate the development and deployment of machines on the battlefield. It would also be hypocritical of me to reject DoD’s increased collaboration with startups and Silicon Valley, something I also advocate for in the book. However, I do have my questions and concerns about how this program came about and will move forward, and my general skepticism of the promises of both primes and startups in this space. So today, rather than just blow smoke or screech about Offset 3.5, I want to deep dive into what a successful Replicator program requires, should look like, and things to be wary of as the program progresses. In other words, I want to tell you how to build a robot army.
What’s a Warbot?
There’s a bunch of terms, varying by military branch, academia, pop culture, etc. for machines on the battlefield. Personally, I use Weaponized Artificial Intelligence (WAI) to apply to all smart military systems from logistics networks to drones, and I use “warbots” as the catch-all for armed and unarmed machines physically on the battlefield. These can range from TF 59’s sail drones to unmanned ground vehicles (UGVs) to the AMASS drone swarm project. In the case of the Replicator Initiative, the DoD is talking about warbots operating in all physical domains of warfare (sea, air, space, land). The general concept is to accelerate the development of these machines over the next two years and begin deployment within two years. The focus of this program is to, as opposed to most DoD programs, create unmanned systems that are low-cost, autonomous, fast to produce, and attritable. Given what we’ve seen in Ukraine and what we can expect in a war with the PRC in terms of ammunition expenditure and casualties, focusing on attritable systems that don’t break the bank is a very smart idea. The challenge is nailing down how we are defining terms like low-cost, attritable, and autonomous, and how we are going to turn them into reality. Let’s talk about a few of the challenges in even defining our robot army.
Image Source: DVIDS. GHWB Operates the MQ-25 Aircraft.
Culture Eats Strategy
In most places in the DoD, $10,000 is cheap for a machine, but if most attritable systems today resemble “suicide drones” should they be classified as a system or as a munition? $10,000 is a lot of money on a commander’s hand receipt, and "Chase the Raven” culture remains alive and well down on the line. What I'm referring to here is the GWOT-born culture within the services to either not use or become incredibly passive about the employment of cheaper systems like the Raven UAS because of the cost if they are damaged or lost, where many commanders decide the cost of losing the system is greater than the ROI for successful employment on the battlefield. I’ve heard rumor there are efforts to change this, but culture can take a while to transform, the Army still struggles to get commanders to follow explicit SECARMY directives even when it doesn’t cost them tens of thousands of dollars.
That’s risk-aversion at the micro level; but the macro level investing in these systems will undoubtedly require DoD to incur a loss on a lot of potential warbots. Many will fail in development, and so long as they are fairly vetted, that shouldn’t be a problem, and yet many will undoubtedly make it a problem. Moreover, I can say from personal experience that there will be vendors that seek to exploit the noise and hide their junkware through savvy marketing pitches. A few of these will undoubtedly get through, it happens, where my concern lies is how they process is being reformed, and how we are ensuring that the origin of this initiative isn’t rooted in those kinds of vendors. Trial and error, and most importantly failure, are things we have to get back to accepting especially when we’re talking about attritable systems. We talk about all of the amazing things we built in World War II, but for every system that succeeded, tens if not hundreds of prototypes failed, and of those that made it to the field, everything from binoculars to tanks still failed more than they were supposed to and in unanticipated fashion. That is simply the nature of attrition warfare on the industrial scale. Being able to identify issues, adapt, throw out failures, and rebound without years of institutional sputtering is a cultural problem for the DoD. The warbots might be able to take on the PLA, but the DoD’s cultural will be their first and greatest challenge. I won’t even touch the cultural challenges of robot-replacement paranoia.
It’s not impossible to change this culture problem. The Marine Corps, under General Berger, successfully transformed its culture in order to implement Force Design 2030. Under CMC Berger, the Marine Corps led a top-down and bottom-up transformation in how the USMC thinks, trains for, and fights wars. It was the culmination of years of work before him, and not without opposition from ill-informed has-beens on the outside. The Marines are continuing that transformation and reworking their designs as they experiment. This needs to be the model that all of DoD adopts if it wants to successfully develop and integrate its robot army.
Your Tax Dollar at WERX
Of course, what’s a program without money? Congressional funding for…anything these days remains up in the air. Nothing should be assumed to be guaranteed, but I’d say a funding for a program like this is fairly safe, for now. The problem here is more how DoD may allocate its money. DoD contracts are notoriously gargantuan, messy projects, even with AFWERX and similar initiatives that seek to streamline the process for smaller vendors. I’m a huge advocate of greater Silicon Valley and DoD cooperation and beyond helping startups and “new” defense companies through the process, DoD and Silicon Valley need to get better at educating each other on what their processes look like. Both sides view the other with a sense of mystical unknown, and that leads to break downs in communication that are then exploited in marketing pitches and lobbying to Congress. Throw money at the new companies, let things break and find new successes, but the success can only be as honest as the money thrown at it. I’ve seen too many pitches that abuse the ignorance surrounding the acquisitions process, not to mention concepts like AI, quantum, and robotics. DoD has done a lot of work in the last decade with DIU and other programs improve its outreach to new companies, but it’s still filled with people that couldn’t be further from Silicon Valley and vice versa. It’s still going to take a while before the right people are talking to the right people to really jumpstart the development process.
One of the ways to get around the old DOD pipeline while still ensuring project integrity is to transition away from programs of record and replace them (or at least some of them for smaller dollar, software intensive accounts) with capabilities of record. Disclaimer: this is not my original idea. In the most basic sense, rather than locking in money for a specific company for a contract, we are locking in money for the capability offered, and every so often, companies can rebid/lost the contract if they’re not up to snuff without having to cancel an entire program. Think of it like changing phone plans but retaining your phone number across carriers. This would undoubtedly ruffle feathers on the Hill (I think it already has) but it has to be done. This is the real way to preserve competition and reduce barrier to entry in the thinning military industrial sector without playing new favorites.
Build the Bot Around the Gun
What’s a warbot without a weapon? We undoubtedly have the most experience mounting things that go boom to robots. The Ukrainians are a close second these days. From hunter-killer HARPY’s and artillery spotting microdrones to quadcopters with unguided rockets and underwater vehicles that can launch mines and torpedoes without a human in the loop, the future of warbots means building the machine around the proverbial gun. At its core, that’s what the Pentagon wants and needs. Sensors are only as effective as the effects they help generate. In our case, even the unarmed sensor bot may need to be attritable while still contributing to putting warheads on foreheads. Migrating from focusing on the development of warbot systems themselves to ones that are explicitly designed to be on the pointy end of the killchain (there’s that ugly word again) is going to be Replicator’s proverbial valley of death. Putting the war in warbot is the difference between a bunch of vaporware and actually pushing the US military to the next technical level. That might sound simple, but that space between machine and killing machine is also the no man’s land between DoD and Silicon Valley. We can do it, but it’s going to take time to do it right and at scale. And in order to do that, we have to ensure we build the warbot around the gun.
“These things take violence and timing”
Timing is everything. Get to an objective too early and you might give the enemy advance warning, miss a time hack and you might link up with the wrong end of an artillery barrage, spend two decades on a bunch of useless exquisite junk systems and you might find yourself on the wrong end of a new arms race. If there’s one thing I’ve emphasized in this article, its the timeline for development, deployment, and integration of warbots into the Joint Force from Replicator. DoD knows that breaking through the valley of death is its most serious acquisitions challenge, Silicon Valley knows it too. You can’t just force things through, if you want less of a squeeze on the pipeline, we have to play by “wormhole rules”: it takes less energy to pass smaller objects through a tear in spacetime than it does to tear a larger hole in the fabric of reality. For DoD, this means investing more in getting a longer list of new, low-cost systems out to the Force, and then deciding which ones to invest heavy in. The Army did this quite well with the latest batch of night vision devices and a few other systems. Obviously, this methodology doesn’t work for necessary exquisite systems like fighter jets and ships, but if your goal is attritable warbots, then this is your methodology. I think DoD knows this, and it means communicating to Silicon Valley that they can’t expect huge infrastructure and supply chain investment on the front end. Hopefully this has already happened.
So where does timing fit in here? Well, it means accepting that fielding a system doesn’t mean its war-ready. In other words, the 18-24 month timeline that DoD has planned for fielding apparently thousands of warbots is not representative of how the force will operate, nor how many of those thousands are actually war ready. My concern here is that DoD’s messaging has oversold the process, and when that timeline hits, a lot of folks are gonna be looking around asking questions that DoD can’t answer. I also, somewhat cynically, suspect DoD is sitting on a pile of prototype systems that vaguely resemble their end goal but don’t really meet the definition of an attritable warbot ready for combat. But who knows, I could be wrong! Again, I want this program to succeed, my worry is not DoD as much as it is DoD playing with fire. The fire in question is the perception of Offset 3.5, LCS 2.0, FCS 2.0, etc., where we get a bunch of buzzwords and we’re no closer to actually fielding our robot army. The more DoD can communicate on this program, the better it will be for everyone involved. Now is not the time to play communications games.
Finally, my last question on timing is the ability to scale production with a secure supply chain. It wasn’t long ago that no one could make a drone in the US without at least a Chinese screw or two, 3D printing at scale is still about a decade away, and I haven’t seen a bunch of new factories pop up in Silicon Valley or Northern Virginia. Where’s the supply chain? Where are the production lines that can do more than prototype or pop out second-rate products with a +/- 10% margin of error on design. Attritable shouldn’t mean poor quality. We can’t trade one defense supply chain crisis for another, and that is my most significant concern here: can we actually ensure we have a wartime production chain for the future of the force? I honestly don’t know. There’s a few companies out there doing work, but I’m not quite confident we’re there yet. And this supply chain issue runs into the wormhole problem. If you don’t have an existing supply chain that can withstand some shock, that shock is going to find your sooner or later and hopefully it’s not in the valley of death.
Deterrence, Competition, and the Bleeding Edge
When it comes to defense tech, never forget where the blood in the bleeding edge comes from. If we are going to bet big on attritable warbots, remember what we’re betting. I say this because I don’t like that we’re talking about these systems in terms of deterrence. Individual systems, no matter how evolutionary or revolutionary, are not automatically deterrents. Successfully integrating warbots into new force structures and doctrines, paired with good training and exquisite systems, as with FD2030, is what makes a conventional deterrent credible. Moreover, that credibility comes from the opposition seeing what these systems can do to their systems. Once again, we’re back to communication being key here. In our arms race with the PLA, the development of our robot army gives us an edge, but only if our bots work and make sense for how we want to fight and for how the enemy may force us to fight. A robot army would undoubtedly have to be the product of a successful overhaul of the arsenal of democracy. It would certainly be a demonstration of our ability to withstand an attritional war like what we’re witnessing in Ukraine, but that requires a lot of back end work before we get there. It requires funding the munitions, maintenance, and personnel for those exquisite systems. Warbots are a worthy compliment to the existing joint force, they are not, however, a substitute for a failure to properly run and fund American national security.
If you enjoyed this article, check out my novel, EX SUPRA. It’s the story of the war after next. Nominated for a Prometheus Award for best science fiction novel, it’s the story about the war after the next war. From the first combat jump on Mars to the climate change-ravaged jungles of Southeast Asia, EX SUPRA blends the bleeding edge of technology and the bloody reality of combat. In EX SUPRA, the super soldiers are only as strong as their own wills, reality is malleable, and hope only arrives with hellfire. Follow John Petrov, a refugee turned CIA paramilitary officer, Captain Jennifer Shaw, a Green Beret consumed by bloodlust, and many more, as they face off against Chinese warbots, Russian assassins, and their own demons in the war for the future of humanity. And don’t forget to subscribe to Breaking Beijing! It’s free!