Air Force moves out on Collaborative Combat Aircraft
Plus, register for our upcoming Nexus Talks webinar.
Welcome back to the Nexus Newsletter, Applied Intuition Defense’s biweekly newsletter covering the latest in national security, autonomy, and software-defined warfare.
In today’s edition, we discuss the Air Force’s Collaborative Combat Aircraft program and announce a new Nexus Talks webinar on the future of software-defined warfare. Plus, we’ll be at the SCSP AI Expo next week and share a recap of some recent events we were at.
Anduril, General Atomics win first round of CCA awards
The Air Force recently made its first round of awards for the Collaborative Combat Aircraft (CCA) program, the service’s biggest bet for gaining a competitive advantage in a near-peer competition.
The service announced last week that Anduril and General Atomics Aeronautical Systems will build and test drone prototypes for the CCA effort under the first tranche of awards. This is a resounding endorsement by the Air Force, selecting two contractors — one of which is a nontraditional vendor — who embody manufacturing speed, low cost, and modularity for one of the Air Force’s most strategic programs.
This stands in contrast to traditional highly exquisite and technically complex programs. Recently, before the announcement, Secretary of the Air Force Frank Kendall put an emphasis on the efficiency of design, stating, "We want you [industry] to invest in the technology and think about how you're going to make a very efficiently produced product for us."
Previously, at a media roundtable, Assistant Secretary of the Air Force for Acquisition, Technology, and Logistics Andrew Hunter mentioned that “focus has been speed-to-ramp..So we've focused on things that we believe we understand very well and that industry understands very well how to deliver that capability on the accelerated timeframe that we've established for Increment One.”
While the majority of details surrounding the program are highly classified, Hunter was also quoted by The War Zone at the 2024 AFA Warfare Symposium saying, “The hardest part is the autonomy piece…the more complex the autonomy that you expect the system and platform increment to perform, the harder it is."
With the focus of Increment One being predicated on speed of development, it puts an emphasis on the developmental methodologies of both Anduril and General Atomics. As Hunter was also quoted by Defense Scoop last December, the “continuous competition” across a diverse vendor base is not limited to hardware but is also a core component of the Air Force’s acquisition strategy for software.
Autonomy is hard. The automotive industry has spent the last 10 years experiencing this very pain point — digital transformation and workforce modernization, shifting from hardware-focused manufacturing organizations to highly capable and self-sufficient software hubs.
General CQ Brown’s motto while he was Chief of Staff of the Air Force, “Accelerate, Change, or Lose,” rings evermore true today. Large defense organizations, some of whom are still implementing waterfall software development approaches on current contracts, may struggle to adapt and implement rapid software development methodologies required in order to efficiently develop, iterate, and sustain autonomy technologies.
The mass consolidation of the defense industrial base on one hand limits the total number of available organizations and thus the diversity of technical approaches. On the other hand, this same consolidation culminates with organizations stacked with incredible amounts of available resources to apply to very difficult technical challenges.
This fierce competition is exactly what the defense industrial base needs. Non-traditionals and insurgent companies such as those represented in the a16z American Dynamism portfolio are a great example of how small businesses also have the capacity to meaningfully disrupt the status quo. If there is one thing the West can bet on, it is that the bedrock of American technical supremacy is born and forged by competition.
The work will be difficult, but it is uniquely the American industrial base, new and old, who is pioneering this new era of aerospace with unmanned and autonomous systems.
Register for our upcoming Nexus Talks webinar!
Nexus Talks is back next month with a new webinar featuring Bryan Clark, Senior Fellow and Director of the Hudson Institute’s Center for Defense Concepts and Technology. Bryan will be joined by our Head of Strategy and Communications John Mark Wilson on May 22nd at 10AM to discuss the future of the software-defined force.
This webinar will preview themes and discussions you can expect to hear at Nexus 24, our annual thought symposium with visionaries, leaders, and operators at the forefront of AI and software-defined warfare. We just announced our first set of keynote speakers for our event and we can’t wait to see you all there.
Register for “Beyond the Buzzwords: Accelerating the Software-Defined Force by clicking here or below. And ICYMI: Click here to watch our most recent webinar, “Swarms and speed: What the Pentagon can learn from Ukraine.”
Come see us at the SCSP AI Expo
We’ll be at the Special Competitive Studies Project’s AI Expo for National Competitiveness in Washington, D.C. on May 7th and 8th at booth 256. Come stop by or contact us here to set up a meeting.
Last week, we were at the Association for Uncrewed Vehicle Systems International’s XPONENTIAL conference in San Diego, where we discussed how AI and autonomous technologies need to be prepared for the operating concept, operating environment, and counter-operations. Some key takeaways:
Warfighters need to be part of the product team
The feedback loop on how users engage with the system is critical for development
AI explainability is important, but the real goal should be uncovering failure modes. Continuous simulation, evaluation, and certification is critical to identifying potential failure modes before they erode capabilities and trust
We were also at the National Defense Industrial Association’s Michigan Defense Expo in Detroit, where we demonstrated on the show floor how autonomy stacks and sensor simulation can run in real-time with a driver in the loop and spoke about acquisition strategies for equipping the software-defined force.
News We’re Reading
Breaking Defense | ‘AI-BOM’ bombs: Army backs off, will demand less detailed data from AI vendors
Key quote: “Instead, amid industry outcry, the author of the now-abandoned ‘AI-BOM’ idea said the service will ask for a much less detailed ‘summary card’ meant to feel less intrusive to companies jealously protective of their algorithms and other intellectual property.
‘You can think about more like a baseball card that’s got certain stats about the algorithm,’ principal deputy assistant secretary Young Bang told reporters Monday. ‘It’s not as detailed or necessarily as threatening to industry about IP.’
At the same time, however, the Army remains deeply concerned about potential vulnerabilities hidden in AI algorithms, which are born of complex math and then modify themselves over multiple rounds of artificial evolution until they become, in many cases, incomprehensible to humans…But the Army is keenly aware it can’t take what would, in theory, be the most secure approach and just build all the AI it needs in-house. What it needs, Bang said, is a way to adapt the best AIs from industry, make sure they’re safe, and retrain them as needed for military purposes using sensitive or classified data.”
Our take: Perhaps the supposedly onerous AI BOM is problematic when the Government perceives that it is dependent on IP from AI companies. However, by virtue of the fact that the Government has immense potential leverage to collect, hold, and distribute data of all kinds to whom it wants, traceability concerns should be organically mitigated by creating incentives for openness in exchange for privileged access to data, which would remain compartmentalized on a Government network.
Unfortunately, efforts to enable these incentives have not yet materialized, despite the great potential and need. Moreover, instead of unlocking the full potential of industry to do great things with this data, many Government R&D agencies are competing with industry on model development, while squandering the ability to pool and manage data effectively across the enterprise.
In effect, we have a terribly throttled system whereby the best model developers (industry) have the least access to data and the best data sources (Gov) are trying to create their own models instead of deploying a data enterprise for autonomy and AI.