Virtual T&E, talent, and data management
Applied Intuition
Accelerating the world’s adoption of safe and intelligent machines
The Nexus Newsletter?
Welcome to the sixth edition of The Nexus Newsletter - a bi-weekly email newsletter covering recent announcements from Applied Intuition and important news bridging national security and autonomy.?
This edition of the newsletter includes the second in a two-part blog series on test and evaluation (T&E) for autonomous military vehicles, recent comments on the importance of attracting and retaining talent, thoughts on enterprise data management for defense, recent news, and a list of upcoming conferences and events we’re attending.?
Read below for updates and commentary from Applied’s government team.?
Virtual testing enables comprehensive T&E for autonomous military vehicles
We published the second in our two-part series covering T&E considerations, approaches, and solutions for autonomous and semi-autonomous military vehicles! Read part 2 for more about:?
Read part 1 to understand how T&E challenges differ across autonomous vehicles built for commercial and defense applications and why relying exclusively on real-world testing results in significant gaps in test coverage.
Talent acquisition and retention are critical to national security innovation
In the conversation around the development of innovative technologies for national security, one key aspect is often overlooked: Talent, not only in the Department of Defense, but in the country as a whole.?
As we mentioned in the fourth edition of this newsletter (July 24th), Applied’s Director of Government Colin Carroll participated in a field hearing hosted by the U.S. Chamber of Commerce AI Commission on Competitiveness, Inclusion, and Innovation. At the hearing, Colin argued that government, academia, and industry must do more to attract and retain the world’s best technologists to solve our most important national security problems.?
“The bottom line is our national security depends on our ability to attract, nurture, and retain technical talent […] If we enact nativist immigration policies or even the status quo that we have now, I really do predict that by the end of the decade the transition of the AI education and research power base to basically China and India [will] be complete. And it’s going to be really, really hard if not impossible to overcome. […] We need immigration reform, ideally bipartisan, that prioritizes incentives for foreign talent to attend school, to conduct research, to work here, and to become really U.S. citizens.”?
Read the six recommendations for how the U.S. can lead on AI from the U.S. Chamber’s AI Commission hearing, or watch the recording of Colin’s comments.
Enterprise data management for defense autonomy programs
Enterprise data management is too often overlooked at the start of defense autonomy programs - it is not the shiny object that program managers are chartered to deliver, so there’s not much incentive to invest. At the same time, programmed dollars go to specific programs; there is often no incentive to share resources for enterprise capabilities that span across similar programs. While often treated as an afterthought, a robust data management program is actually the foundation for any kind of autonomous system development and deployment program. Rather than letting sensor data die on the screen or sit unused on a hard drive, a good data management platform ensures that data is rapidly ingested and made available across the program – for instance, to labeling and machine learning teams as part of the machine learning process, or to test engineers for assessment of system performance. Getting this right early will translate to long-term program success. Getting it wrong could translate to years of wasted efforts. At Applied, we believe that defense autonomy programs should share an enterprise data management platform if at all possible. This reduces specific program costs, enables cross-program sharing of datasets, and is generally a commercial best practice. Tesla doesn’t stovepipe its Model Y data and its Model 3 data on two different platforms, so why should the Department of Defense (DOD)?
Over the last eight months of exploration, we have seen drive log data stored on hard drives and stashed under desks at the DOD, unsecured and inaccessible by the enterprise. We have seen the government hemorrhage its data across corporate networks. These are symptoms of a common problem when it comes to software-defined capabilities, including autonomous systems: Too often, DOD wants to jump directly to building the vehicle, rather than building the infrastructure that will actually enable the capabilities to reach the field at scale.?
Our friends across the pond are clearly learning from our mistakes: The UK Ministry of Defence (MOD) is proactively including data management as part of their Human Machine Teaming (HMT) program. Establishing a data management regime will, as we have seen first hand in the commercial sector in the United States, result in resource and time savings, better performance, more targeted testing, and faster development.?
We have a lot more to say about this topic, so keep an eye on our blog and social media accounts.
News we’re reading?
Autonomous systems are gaining momentum in the national security space. Here are excerpts from a few relevant articles, plus brief commentary from Applied Intuition’s government team:
The United States Army | Testing of the army’s first autonomous vehicle speeds ahead
Imagine the possibilities of a self-driving vehicle on the battlefield.
Engineers at the U.S. Army Combat Capabilities Development Command Aviation & Missile Center, or AvMC, don’t have to imagine it, they are building it. The Autonomous Multi-Domain Launcher takes a High Mobility Artillery Rocket System — or HIMARS — and modifies it with hardware and software to be controlled remotely and driven autonomously. [...]
While it resembles the self-driving cars in the news, the Autonomous Multi-Domain Launcher — or AML — will have capabilities that those vehicles will never possess — but also challenges. Regular autonomous autos operate on established roadways, they aren’t navigating a sometimes treacherous terrain. They also aren’t being fired upon by enemy combatants. Then there is the matter of size. [...]
“We are talking about putting a 36,000-pound vehicle in an area where there will be humans running around,” said Lucas Hunter, AML project manager for AvMC. “Tesla and other companies are working on vehicles that can sense cars in front of them and behind them; they have these nice stripes on the road that tell it, ‘I am getting out of my lane.’ Well, we are driving through open country, we don’t have stripes — we have holes, we have cliffs.”
Our take: History shows that independent development efforts for autonomous vehicles within each of the services will be disjointed and slow. Since the effectiveness of autonomous systems during operations will rely less on the capability of individual platforms, but instead on the ability of human and machine teams from across the services to work together, coordination across military departments and domains is essential. Congress can help the DOD accelerate the maturation of these technologies by investing in a joint, all-domain modeling and simulation capability that reduces the development lifecycle for autonomous systems and enables efficient multi-domain mission-level testing. Virtual simulation will accelerate progress, de-risk adversary intelligence collection of repeated live testing, and save taxpayer dollars.?
Defense One | The Pentagon’s Plan to Buy Weapons Software Faster
As the Pentagon’s business systems and its biggest weapons platforms increasingly rely on more frequent software updates, the Defense Department is working to change its processes so it can buy and deliver software as quickly as it’s needed.?
One of the primary ways it’s doing so is through its software acquisition pathway, which allows contracting officers and program managers to separate out the software components of various programs, from unmanned systems to enterprise software, making the buying process faster and easier.?
The Biden administration’s pick to be the Pentagon’s No. 2 weapons buyer has vowed, if confirmed, to make software modernization a priority, and put a spotlight on the two-year-old pathway. But while the Defense Department is likely to keep pushing it for new and existing programs, it may take some getting used to. [...]
But, as it is now, using the pathway requires some creativity, according to Doug Bush, the Army’s chief weapons buyer.?
Bush told reporters July 27 that the service is using the pathway for newer software programs, while working to shift older ones that use a traditional process to one for continuous development—called DevSecOps—through “schedule adjustments on software programs to move them closer to the way we like to do things,” rather than doing it formally. [...]
One of those programs is the Robotic Combat Vehicle. Bush said the Army “separated out the software element as a separate software acquisition pathway program because we want that control software to be common across many robotic platforms.”?
For the likely-to-be-confirmed Plumb, taking a flexible approach and tailoring the Pentagon’s interoperability and requirements processes for software is “essential” to more fully adopt more modern development models. [...]
Our take: As the Army works to create hybrid contracting approaches for its vehicle programs that utilize both Middle Tier Acquisition and Software Acquisition Pathways, it is understandable that those programs will require creative solutions to deliver faster. As we discussed in the fifth edition of this newsletter (August 10th), the hybrid contracting approach will make hardware-software integration more complex, but still represents an exciting first step. Continuous integration and deployment (CI/CD) is a commercial software best practice that will prove essential to the success of defense autonomy programs during development, testing, and after the vehicle reaches the field.?
Defense News | US Army’s ‘Lethality Task Force’ looks to save lives with AI
As the Pentagon seeks to modernize its weapons program, artificial intelligence and autonomous robots could hold the key to improving the lethality and performance of close combat units.
领英推荐
Front-line infantry service members have long suffered casualties in higher proportion to other positions, with infantry soldiers comprising 90% of U.S. military combat deaths since World War II.
During the Trump administration, Defense Secretary Jim Mattis created the Pentagon’s Close Combat Lethality Task Force to examine capability shortfalls to address what he saw as decades-long gaps in the equipping and training of close-combat units. While leadership of the task force now falls to the Army rather than the secretary of defense, the initiative continues to look at shortfalls across multiple services’ small units.
With AI the name of the game in the Pentagon’s modernization efforts, the four-year old task force is turning to academia and industry leaders to regain a competitive edge in close combat and examine how autonomous technology can be leveraged to address small units’ priorities. [...]
The mission of the task force is twofold: to increase the “lethality” of close combat soldiers while decreasing casualties of U.S. infantry members. [...]
The force, which still includes members coming from multiple services, is now looking toward AI and robots as the next step in filling infantry squads’ capability gaps. At the end of July, the force hosted the first meeting of the Artificial Intelligence for Small Unit Maneuver working group at Fort Benning.
Building on the task force’s mission of improving lethality for small units, the group is particularly interested in providing autonomous and AI technology to close combat troops in the Army, Marines and Special Forces. [...]
By Aug. 31, defense leaders from the services will provide the task force with a list of capability gaps for the Army, Marines and Special Operations. From there, he said the task force will turn to industry leaders and academics to brainstorm ideas about what AI technologies could assist in meeting those needs.
“If a technology shows great promise, and it’s successful with development and the decision is made to make it a program of record, it would go through...an accelerated version of acquisition,” Agee said.?
Our take: At Nexus 22 earlier this year, experts agreed that one of the key lessons to be learned from the conflict in Ukraine is that manned units should not make first contact with an adversary in any future conflict. Autonomous systems fill in capability gaps by handling dangerous tasks in risky environments, providing additional intelligence to make infantry more effective, reducing the cognitive load on warfighters, and other tasks that contribute to warfighter survivability and effectiveness. We are interested in what “an accelerated version of acquisition” means in this context - will DOD commit to building the data collection and management infrastructure required for these programs to succeed??
Air Force Magazine | ‘Wildly Successful’ Skyborg Will Become Program of Record but Won’t Stop Developing S&T
Skyborg, the Air Force’s effort to develop an artificial intelligence-enabled system to control unmanned aircraft, is ready to “graduate” and contribute to key programs such as the Next Generation Air Dominance family of systems, a top program executive officer said.
But while the so-called Vanguard effort is seemingly on the verge of becoming a program of record, it won’t stop developing, experimenting, and testing new technologies.
For months, Air Force Secretary Frank Kendall has described his vision of four to five low-cost, uncrewed aircraft flying alongside the crewed fighter that will be the centerpiece of NGAD—or other fighters such as the F-35—to create a formation. Lately, Air Force officials have started referring to those uncrewed aircraft as collaborative combat aircraft, or CCA.
Kendall and other leaders have pushed an ambitious pace to deploy these drones, perhaps by the middle of the 2020s. Such a timeline is a credit to the success of Skyborg, Brig. Gen. Dale R. White, program executive officer for the Air Force Life Cycle Management Center’s fighters and advanced aircraft directorate, told reporters during the Life Cycle Industry Days conference.
“We used Skyborg as the foundation, the stepping-off point, with all of the [science and technology] work that we did, which feeds into it, and we’re continuing to work with industry and with the enterprise, and we’ll continue to refine that approach,” White said, adding that because of Skyborg, the CCA effort is not “starting from zero.”
Since being identified as a priority effort in 2020, Skyborg has already demonstrated capabilities that CCAs will likely need.
In December 2020, a drone included in the Skyborg program flew alongside an Air Force F-22 and Marine Corps F-35, allowing the two crewed platforms to communicate using otherwise incompatible datalinks. The Autonomous Core System, the “brain” of Skyborg, has successfully flown multiple aircraft made by different contractors. Multiple drones operated by the ACS have flown together, demonstrating a drone “teaming” capability. And there have been other flight tests as well, the details of which have not been publicly announced. [...]
But while Skyborg’s technology is maturing and advancing toward operational use, the program’s S&T side won’t slow down, White added.?
“Most people may not know this, but Skyborg is still flying. We just flew again in June,” he said. “They’re still pushing out capability, continuing to push the bounds of what we can do with autonomy.”
Indeed, White indicated that Skyborg has not only proven out technology for autonomous unmanned aircraft, but it has also demonstrated a new way for acquisition and S&T to work together.
“While we may graduate, programmatically speaking, there’s still continued work that needs to be done on the S&T side,” White said. “And what we found is through that process of Skyborg, that relationship between the PEO and [Air Force Research Laboratory] commander [Heather L. Pringle], it has proven to give us so much benefit. I get to steer what I think she should be looking at. And then she goes off and does that. And then that feeds into the decision space of how we build capabilities.”
Ideally, White added, that’s how the Air Force would like for all its programs to work. “In the case of Skyborg, it was just wildly successful in terms of what we got out of it, what we continue to get out of it, and how we use that to present decision space to our leaders on how we set up programs of record,” he said.
Such an approach is especially important for Skyborg given its focus on software, artificial intelligence, and machine learning. Observers and officials have frequently bemoaned the Pentagon’s acquisition process for cutting-edge technology, saying it doesn’t work with the rapidly iterated, constantly improving approach of tech innovators.
By allowing Skyborg’s science and technology progress to continue while integrating it into collaborative combat aircraft and other programs, the Air Force is moving closer to operational AI, said Caitlin Lee, senior fellow for the Mitchell Institute’s Center for Unmanned and Autonomous Systems.
“It is really important to prototype and to do so quickly and to iterate quickly and to do lots of tests and evaluation, and Skyborg is doing that,” Lee said. “What I’m talking about is crossing what they call the Valley of Death into a program of record. And I think the real trick there is that you can’t just stovepipe this AI in a program of record all by itself. It’s got to be constantly integrated and developed within a platform and optimized for that platform.”
Our take: Autonomy programs and software development do not end with the deployment of a new vehicle or capability - it is a continuous process for the life of a program. We’re glad to hear that Skyborg’s S&T process will continue during its integration into CCA and other programs. We’re looking forward to seeing a similar approach applied to other autonomy programs and vehicles in other domains.?
In the four years that he led the Defense Innovation Unit, tasked with bringing the fruits of commercial tech world to the U.S. military, Michael Brown says he added more than 100 new vendors to the Pentagon roster.
Yes, but: Brown tells Axios that the Pentagon needs to make more structural changes in how it buys equipment if it really wants to take advantage of commercial technology. [...]
The big picture: The Defense Department was once at the leading edge of developing key new technologies, including the internet itself, but today most innovation is happening outside its direct sphere of influence. That’s why the Pentagon aims to forge new bridges to the private sector. [...]
Between the lines: Brown says he kept running into the same challenge: the military's process for buying equipment. "We’re still in a large sense buying things the same way we did 50 or 60 years ago," he said.
Our take:?We have had the pleasure of working with a number of DOD programs via DIU over the last eight months, and we are believers in DIU’s ability to positively impact contracting schedules to rapidly deliver commercial technologies to DOD programs that would otherwise be relegated to government-owned software solutions or slower delivery timelines. We are sad to see Mike Brown depart, but excited to see what the future holds for DIU.
Upcoming conferences & events we’re attending
Tracking new events in the autonomy and national security space can be difficult. Here are a few of the upcoming conferences and events that Applied’s government team is attending - find us there!
Thank you for reading The Nexus Newsletter. Stay tuned for more announcements from Applied Intuition, additional information about Nexus 23, and other important news from the nexus of national security and autonomy.