Will AI give the Army a secure ‘Snapchat of information’?

By: Kelsey D. Atherton  

Source Link

The Army wants to explore the ways in which robots, artificial intelligence and autonomous systems could supplement, aid and transform ground power. Among the service’s more ambitious ideas: Is it possible for AI to enable a platoon to take on a force 10 times its size?

The hypothesis put forth by Ted Maciuba, deputy director of robotics requirements at the U.S. Army Maneuver Center of Excellence, is that AI and robots will offer such an outsized advantage to infantry that a given force could face a similarly equipped rival that is an order of magnitude larger (minus the robots and AI) and come out on top.

To test these hypotheses, Maciuba plans to host a September 2020 demonstration at Fort Benning, Georgia, with industry supplying machines, software and other tools to test.

C4ISRNET’s Kelsey D. Atherton spoke with Maciuba recently about the planned exercise and the tricky nature of releasing classified information to infantry without clearances.

C4ISRNET: What’s one thing industry should keep in mind when working on AI?

TED MACIUBA: The military folks know something that the commercial folks may not know, which is that most of our soldiers don’t have security clearances. We’re talking about getting an AI system out there that is going through and getting all this classified mission command, classified intelligence. And the product from that is going to a 19-year-old soldier who basically has had a background check to make sure that he hasn’t been too much of a problem for the law, has gone through the training necessary to be an infantry man or an armor crewman or a scout. And they are going to be getting the products of this the same way they get the products coming out of our radios.

They are listening to what’s going on. And they are a consumer of what is essentially classified information. But the view is that it’s ephemeral classified information and it is something that impacts them now and the fact that it’s coming to them will not change anything because they act on it and then they move onto the next mission. So, whatever we come up with out of AI has got to be transitory, ephemeral: It is presented and then you move on.

C4ISRNET: Let’s talk about what this means, because we’re talking about the robotically equipped platoon and the AI-enabled platoon.

MACIUBA: These are the two different Joint Capability Technology Demonstrations we’ve proposed. I’m telling industry that we’re getting ready to put out the request for white papers through some kind of other transactional agreement process. Part of that is determining what is the scale of the resources that the Army must provide to be able to do both of those things. That AI-enabled infantry platoon, kind of by definition, has information that’s coming to it that’s going to be classified even though the people who are going to be consuming it do not have a security clearance.

C4ISRNET: Would the AI hold the clearance?

MACIUBA: I would say that the AI is going to be that fusion, an element that takes all that information and then provides a product and that product will go through filters and there will be top secret products. There will be secret products, products that are coming out of it that — because they are an ephemeral, maybe secret product as they go in — are going to be declassified by the time they get to the user, though that is not the right term. We will just accept the fact that the user right now and for the foreseeable future will be able to use that initially classified information to do their mission.

C4ISRNET: The Snapchat of information?

MACIUBA: Absolutely.

C4ISRNET: What kind of tasks do you envision? Is this just to see how much they can do with AI?

MACIUBA: We know what soldiers do right now. At this point we don’t have the granularity yet to be able to understand how AI will help them make better decisions faster, but some of that is going to be gathering that data and then putting AI tools, AI products that are classified but ephemeral information. Maybe I’m providing that to a soldier as a Heads Up Display that will provide the interface, populating that synthetic picture so you will be able to see, for instance, that at 2,000 meters on this azimuth there is some threat. I think [we will have] material recognition, potentially facial recognition, maybe cyber recognition to say, “There is a threat that the soldier will need to be able to address.”

In the future, AI will be able to show you what you can’t see but is within your area of operations. Again, you can think of a platoon being kind of a hemisphere that probably goes out about 500 meters. We want to be able to see subterranean; we want to be able to see the multidomain; we want to be able to see what’s going on in the water; we want to see what’s going on in the air.

Now we want to see all those domains and we want to have those soldiers work within that synthetic real-world environment so that they can be more effective than the adversaries they go up against.

C4ISRNET: You wanted a platoon that can take on a battalion.

MACIUBA: One of the metrics is 10 times better. Ten times better means that if you can see something at 500 meters, now we want to go out to 5,000. When it all comes down to it, we now have a platoon that can do the mission of 10 platoons.

Our platoon goes up against a comparably equipped foe without all this robotic stuff, now we can defeat an enemy battalion. All bets are off if that enemy battalion has a comparable constellation of robots and has a comparable AI system, a comparable mission command system.

What we’re looking at is how we are positioned right now versus a threat. If we had this constellation of robotic systems, we would expect to defeat a force that’s 10 times larger than the force.

C4ISRNET: Would the AI be looking at awareness or offering ways to outmaneuver enemy positions?

MACIUBA: We’re looking at the full range of lethality protection, mobility, situational awareness. We’re talking to industry, saying, “This is our vision, what does that mean for what you can provide, especially the technology you could deliver to Fort Benning in September 2020 for senior leaders to see what is in the realm of the possible?” These are going to be either prototypes or systems that we put together in a different way that make us more effective.

C4ISRNET: Might someone submit a unit management AI that can collect and interpret intelligent products and then spit out orders? An AI that says, “You could be over here instead.”

MACIUBA: A platoon is the largest organization whose sole mission is to close with and destroy the enemy. Once you get to company and battalion and brigade combat teams, you now start to have that more formalized orders process.

That is probably more the mission command part of AI, whereas we’re focused on bringing AI down for the actual operation. Now the platoon has been given a mission. That mission allows it to leverage all that it has — that constellation of ground, air, potentially water, potentially cyber, potentially some space capabilities — so that we can do multidomain operation at the platoon level.

C4ISRNET: What kind of machines do you see? Are you looking at ground robots, air robots, sea robots?

MACIUBA: In a perfect world, you’d have a robot that would fly in. Because flying takes so much energy to beat the air into submission, it would be able to drive around and if there was some reason that it needed to see what was going on underwater, it would go underwater. I think that’s a future version. There will be some that are air, some that are ground. There are some that are water, underwater. The constellation necessary to be able to get to that 10 times increase in capabilities, and we don’t know what that is. We want industry to come back and say, “Hey, you know, have you thought about this?” And probably we have not. We are not the experts on this.

What we want to do is integrate all those components to get to that 10 times capability. We want those robots to be our picket line. Back in Civil War days, you put out the pickets and those were soldiers that were out there giving you that early warning. They were providing you a certain level of protection, of lethality. They would come up with the mobility solutions. That’s what we want robots to do.

C4ISRNET: The robot takes point. The robot makes first contact.

MACIUBA: We want our soldiers to do their mission enabled by robots. We need to have a soldier-machine interface. There needs to be enough AI built in, not only to that system, but also to the squad, to the platoon, to the company, to the battalion, to the brigade and the echelons above brigade. As you lose the interface with the AI cloud, you will have the graceful degradation of being able to still be effective but not as effective. And then when you get down to probably that platoon level — that platoon then still needs to be able to interact with its specific constellation of systems. And then when all that goes out, then you still have a soldier, a trained platoon that can do the things that our platoons can do.