Learning Systems, Autonomy and Human-Machine Teaming

By Cheryl Pellerin
DoD News, Defense Media Activity

Aboard a military aircraft, coming home from the Nov. 20 Reagan National Defense Forum, Deputy Defense Secretary Bob Work discussed the department’s future.

The Pentagon may always plan for war but its real goal is deterrence — to keep potential adversaries from even considering initiating a war with the United States, Work says.

Deputy Secretary of Defense Bob Work is briefed on the function and capability of the National Ignition Facility as he tours the Lawrence Livermore National Laboratory in Livermore, Calif., during a visit Aug. 5, 2015. Photo by Master Sgt. Adrian Cadiz

Deputy Secretary of Defense Bob Work is briefed on the function and capability of the National Ignition Facility as he tours the Lawrence Livermore National Laboratory in Livermore, Calif., during a visit Aug. 5, 2015. Photo by Master Sgt. Adrian Cadiz

Offset Strategies

To do this, DoD uses what it calls offset strategies –- new ways to get around an adversary’s strength.

The first U.S. offset was in the 1950s when the nation built up its nuclear deterrent to make up for the Soviet Union’s superiority in conventional weapons. Work calls this a brute-force offset, one you can only use when you have the advantage.

In the 1970s, the nation’s second offset was to develop revolutionary systems like extended-range precision-guided munitions, stealth aircraft and new intelligence, surveillance and reconnaissance platforms, then -– and this was the real innovation — combine them with new operational concepts for the offset. The United States used this system of systems for victory in Desert Storm.

A third offset strategy is needed now, defense officials say, because adversaries are building up their military capabilities using all kinds of commercial technologies that are available to anyone.

So as part of the offset the department is investing in five technology building blocks that also, like in the 1970s, must be combined to work as one highly capable, advanced system to stay ahead of a growing range of adversaries.

These, Work says, are learning systems, human-machine collaboration and combat teaming, and network-enabled and cyber-hardened autonomous weapons.

1. Learning Systems

Work describes learning systems as those that can be taught to represent the world for a specific domain or problem, and he says they’re especially good with big data analytics to find patterns that people can’t find in enormous amounts of data.

Machine learning – the ability of computers to understand data, manage results and infer insights from uncertain information – is the force behind recent revolutions in computing. E-mail spam filters, smartphone personal assistants and self-driving vehicles are all based on research advances in machine learning. But even as the demand for these capabilities accelerates, every new application requires a Herculean effort. This artist’s concept of a new machine-learning paradigm represents DARPA’s Probabilistic Programming for Advancing Machine Learning program, which aims to address such challenges. Using probabilistic programming languages, PPAML seeks to greatly increase the number of people who can build machine learning applications and make machine-learning experts radically more effective. DARPA photo

Machine learning – the ability of computers to understand data, manage results and infer insights from uncertain information – is the force behind recent revolutions in computing. E-mail spam filters, smartphone personal assistants and self-driving vehicles are all based on research advances in machine learning. But even as the demand for these capabilities accelerates, every new application requires a Herculean effort. This artist’s concept of a new machine-learning paradigm represents DARPA’s Probabilistic Programming for Advancing Machine Learning program, which aims to address such challenges. Using probabilistic programming languages, PPAML seeks to greatly increase the number of people who can build machine learning applications and make machine-learning experts radically more effective. DARPA photo

“But most important,” he says, “we’ll focus on those machines in areas where human reaction time is just too slow. So, cyber defense, electronic warfare, large raid attacks. Things like that where the learning machine will actually make decisions faster and better.”

Scientists and engineers are doing a lot of research in this area including those at MIT Lincoln Laboratories, a federally funded research-and-development center that applies advanced technology to national security problems.

After a recent visit there, Work said researchers can show learning machines hundreds and thousands of different pictures gathered from e-mails or Internet sites like Instagram and Facebook, and the machines can quickly search the photos for specific patterns.

“Right now most of the money for machine learning is in basic scientific research,” Work says, “but the place we’re starting to see it primarily in use is in the intelligence community,” which has made a big investment in the technology.

2. Human-machine Collaboration

Human-machine collaboration is focused on the person and helping the person make better decisions, Work says.

The F-35 is the best example, the machine itself being designed to correlate massive amounts of data and display it in an immediately understandable way to the pilot, who can then make better decisions.

“Over and over again,” Work adds, “pilots are saying this is the game changer, this is totally unlike any airplane we’ve ever flown before.”

A Navy example is Naval Integrated Fire Control Counter-Air, in which, according to USNI News, every air, surface or subsurface unit in a carrier strike group is networked through a series of existing and planned datalinks, giving the commander a clear picture of the battlespace.

Desert Storm KUWAIT---The Third U.S. Army threatened to completely close off northern escape routes, forcing the Iraqi surrender. 40th Public Affairs Detachment photo

Desert Storm KUWAIT—The Third U.S. Army threatened to completely close off northern escape routes, forcing the Iraqi surrender. 40th Public Affairs Detachment photo

The Aegis combat system and submarine combat systems also are good examples, Work says.

On Aegis, he adds, “every sensor is crunching into the Aegis fire-control system, which makes sense of the data, prioritizes the threats coming in at many times the speed of sound — much beyond human reaction time — and presents data to the ship’s technical officer [who] can elect to override the system and take out the threats he thinks are more important. Or if it gets really bad he’ll turn it to automatic and the machine services the target as fast as it can.”

3. Human-machine Combat Teaming

An example of human-machine combat-teaming is the in-development operation of the land-based manned P-8 Poseidon with an MQ-4C Triton drone — manned and unmanned systems that operate together, Work says.

An Army example teams the manned Apache AH-64E helicopter with MQ-1 Gray Eagle and two versions of Shadow drones.

“What we want to do on human-machine combat teaming is to take it to the next level, to look at things like swarming tactics. Can an F-35 go into battle with four unmanned wingmen?” Work says.

And, he added, “there’s no reason why the first entity going through a door in a breach shouldn’t be an unmanned system, sending sensor signals back out to the people right behind it.”

These are the kinds of things the department will explore and invest in, Work says, adding, “We will help the services that are exploring this [and] we will encourage it by giving them advantages inside the budget.

4. Assisted Human Operations

New machines and devices can help people in lots of different ways, Work says. “Wearable electronics, [devices] that power your batteries as you walk, exoskeletons, all sorts of different things that the man and machine can operate together.”

Tomahawk202

Tomahawk cruise missiles are designed to fly at extremely low altitudes at high subsonic speeds, and are piloted over an evasive route by several mission-tailored guidance systems. The first operational use was in Operation Desert Storm-1991, with immense success. The missile has since been used in several other conflicts. In 1995 the governments of the United States and United Kingdom signed a Foreign Military Sales Agreement for the acquisition of 65 missiles, marking the first sale of Tomahawk to a foreign country. Naval Air Systems Command photo

5. Network-enabled, Cyber-hardened Autonomous Weapons

Work says a lot of experimentation is underway with network-enabled weapons, and that everything must be cyber-hardened against adversaries.

The tactical Tomahawk Block IX, whose targets can be updated in flight, is a good example, he says, noting, “It’s totally network-enabled, talking back and forth between the weapons.”

The department is also looking at something the Russians already have done, which is exploring using weapons in a salvo, “like seven missiles screaming in,” he says.

One weapon goes high and directs the other six to targets based on electronic emissions that it senses, Work adds.

Putting it All Together

In the Third Offset Strategy, Work said, “we’ll see a lot of unmanned systems, a lot of science and technology on learning systems, a lot more demonstrations and exercises involving network-enabled weapons. Over time we’ll go to exercises where all five of the pieces are operating together. That’s when we’ll know whether this offset really works.”

Kicking Ass

In the latest offset, Work explained at the Reagan Forum, young officers who have grown up in an iCombat world will have ideas that senior officers won’t be able to emulate. And senior officers who know combat at the campaign level will be able to make the leap to operational concepts and organizational constructs, he said.

“But if we can tap into the captains and majors and lieutenants who have grown up in this world, and we can manage that creativity together,” Work said, “we will kick ass.”

(Follow Cheryl Pellerin on Twitter: @PellerinDoDNews)

Related Articles:

Work: Human-Machine Teaming Represents Defense Technology Future

http://www.defense.gov/News-Article-View/Article/628154/work-human-machine-teaming-represents-defense-technology-future