This past week, Navy leaders called for sailors, civilians, and researchers to commit themselves to emphasizing and adopting robotics and artificial intelligence (AI) to solve warfighting challenges. In a memo to service chiefs, Secretary of the Navy Ray Mabus called for the DON to consider “how to adapt recent private sector advances in fields such as machine learning, natural language processing, ontological engineering, and automated planning for naval applications.”
Why do commercially developed AI and robotics offer such promise to the sea service? Are these advances decades away? And how can sailors in the fleet help drive the change Secretary Mabus is calling for? Let’s examine these questions further.
The Virtuous Technological Cycle: Faster and Cheaper Computing
Berkeley Robot for the Elimination of Tedious Tasks – Source Link)
Pop culture is familiar with the concept of Moore’s Law of Integrated Circuits. Simply put, this maxim states that computing power has tended to double every 18 months for the last several decades. This leads to steady advances processing power and resulting technical advances.
But Moore’s Law is not the end of the story. As speed and computing power have increased, the cost of these capabilities has decreased rapidly. Consider the cost required to execute a gigaflop, a standard measure of computing power. In 1984, it cost $42,780,000 in hardware to complete this task. By the year 2000, this figure had dropped to $1,300. Today, it costs less than eight cents in hardware to complete this task.
These factors create a virtuous cycle. More advances in power lead to more applications where a technology might be adapted. More applications lead to more demand, which in turn lead to larger numbers of chips being manufactured. More investments in manufacturing lead to more investment in research and therefore quicker development. The cycle feeds on itself.
As computing power becomes faster and cheaper, it allows scientists to harness machines to complete new and more challenging tasks. Artificial intelligence programs can sift through massive repositories of data to learn patterns they can then recognize. Software can be programmed to observe situations and “learn,” just as a human does from experience.
Consider the Berkeley Robot for the Elimination of Tedious Tasks, or BRETT, under development at UC Berkley. BRETT is programmed to utilize “deep learning” techniques to observe a problem, orient itself, and solve the issue. While it takes several hours to solve a simple task, with increases in computing power, its speed will grow. Just as a child’s simple brain grows into an elegant masterpiece, so too will such machine learning technology develop rapidly as computing power continues to race forward.
Adopting Rapid Technological Solutions: How to Outfit a Truck
In an article in Proceedings in 2012, CNO Jonathan Greenert wrote about budgetary and acquisitions challenges. Due to lengthy development of new platforms, Adm. Greenert suggested that rather than buying “luxury cars” with numerous built in features, the Navy ought to buy “trucks” that can carry modular payloads. Such open architecture systems can easily and rapidly adopt new sensors, weapons, and technology at relatively low cost.
This flexibility combined with rapidly advancing computing technologies makes the near future very bright. While DoD has been and remains at the forefront of research and development, there are many commercial entities building robots and AI products that have dual military uses. Tools like autonomous robots, facial recognition databases, and speech recognition and translation software have all been developed in the civilian sector and offer great promise in military applications. The speed of commercial innovation is regulated by market forces and Moore’s Law. The speed of our acquisitions system is regulated by a bloated process developed by legislators and implemented by managers with a vested interest in its perpetuation. Which system do you think is faster?
By adopting commercial technology in open architecture systems, the pace of adopting new capabilities can accelerate. Enhancements to ensure information assurance and security will be required. Acquisitions processes will have to be respected as well. But this will minimize costs as well as cut down on the multi-year interval between requirements for a weapons system being frozen, and initial operating capability milestones. Open architecture systems in the aviation, submarine, and surface forces that will enable these capabilities to quickly “plug and play,” with upgrades coming in months rather than years. This will bring new capabilities to match the pace of technological advances as closely as possible.
Imagining the (Not so Distant) Future
How realistic, though, is the introduction of machine learning and advanced artificial intelligence into military service? Certainly, the Navy has adopted systems like the X-47 Unmanned Combat Air System. But are these other technologies more pipedream than reality? Let’s conduct a thought experiment.
While writing, I imagined flying a mission in the near future in my most recent fleet aircraft, the P-8 maritime patrol aircraft. Such a jet would have an AI system that could analyze the ocean environment, predict the actions of a threat submarine, and recommend to its operators where to search. Acoustic operators using SSQ-125 multistatic sensors would be assisted by an AI system that used machine learning techniques to analyze reflections from underwater targets and provide its judgment whether the return was a submarine or a shipwreck. The aircraft would be equipped with an autonomous communications intelligence (COMINT) recording and translating system. This system would automatically record, translate, and transcribe chatter it received.
Sound like science fiction? If it does, the reader may be surprised to know that all these technologies either already exist in various forms, or are very close to reaching fruition. For over a decade, the MH-60R helicopter has boasted an advanced decision aid called the Acoustic Mission Planner (AMP).[7] By analyzing the ocean, AMP can provide a crew with recommendations on where to employ sensors and search. Updated in real time, its algorithm provides a changing search plan as the hunt unfolds. Similar tools for fixed wing aircraft are being developed.
To detect quiet diesel submarines, the navy has turned to high-powered active sonars. These systems, in theory, are subject to high false alarm rates, and require operators to decipher the returns. The Naval Research Laboratory is developing machine learning software that observes how humans classify returns, and then mimics that behavior. Such “human mimetic” behavior can augment the performance of a less-experienced human operator or speed up classification by a seasoned aviator.
While automatic translation seems to be the realm of Star Trek, such technologies are becoming increasingly common, to the point where they are freely available through services such as Google Translate. Earlier this year, DARPA announced that speech identification and translation software could be available to intelligence analysts and combat troops as early as 2017. Such automated tech could remove the need to carry a linguist onboard, while providing the P-8 a new intelligence gathering capability with no additional manning.
Challenging the Warfighter
Adopting robots and AI systems will not just require warfighters and support personnel to consider how new technology can be employed. It will also require that we consider our relationship with these tools. Far from fearing this technology as a threat to us, or our eventual replacement, we should acknowledge that our role will shift and embrace that reality.
While machines increasingly take on monotonous or computationally intense tasks, we will take on the role of supervisor and analyst. For example, airline pilots frequently discuss their role as one of a “systems manager,” allowing the autopilot to conduct much of the physical task of flying while they observe system performance and make decisions regarding malfunctions, weather, and optimizing their route.
Joining the Conversation
New technologies and warfighting challenges will require solutions from all corners of the fleet. The Navy’s Office of Strategy and Innovation has recently launched a crowd-sourced website known as the Innovation Hatch. In the next month, leaders are challenging sailors fleet-wide to offer their ideas and thoughts on how advances in AI can solve problems they see every day on the deckplates.
The Naval Warfare Development Center has also recently launched a crowd-sourced website known as Navy Brightwork to harvest ideas from the fleet. Brightwork is more focused on warfighting applications and as such has both NIPRnet and SIPRnet portals.
It’s an exciting time both in the Navy as well as society at large as we watch technology grow and change around us. Tools that were rare just years ago are ubiquitous and cheap today. As advances in computing race forward, let us hope that sailors adopt the technology around us to seize the intellectual high ground and win the conflicts of tomorrow.
