Slashdot Mirror
Military Documents Reveal How the US Army Plans To Deploy AI In Future Wars (thenextweb.com)
In a just-released white paper, the Army describes how it's working to make a battlefield network of machines and humans a reality. The Next Web reports: "Most of such intelligent things will not be too dissimilar from the systems we see on today's battlefield, such as unattended ground sensors, guided missiles (especially the fire-and-forget variety) and of course the unmanned aerial systems (UAVs)," reads the paper. "They will likely include physical robots ranging from very small size (such as an insect-scale mobile sensors) to large vehicle that can carry troops and supplies. Some will fly, others will crawl or walk or ride."
The paper was authored by the Army's chief of the Network Science Division of the Army Research Laboratory, Dr. Alexander Kott. It outlines the need to develop systems to augment both machines and people in the real world with artificially intelligent agents to defend the network: "In addition to physical intelligent things, the battlefield -- or at least the cyber domain of the battlefield -- will be populated with disembodied, cyber robots. These will reside within various computers and networks, and will move and acts in the cyberspace."
Kott takes pains to underscore the fact that the AI powering U.S. war efforts will need to be resilient in ways that today's AI simply isn't. He states: "The intelligent things will have to constantly think about an intelligent adversary that strategizes to deceive and defeat them. Without this adversarial intelligence, the battle things will not survive long enough to be useful." Ultimately, aside from outlining what the future battlefield will look like, the paper's conclusion is either disappointing or a giant relief, depending on your agenda: "Clearly, it is far beyond the current state of AI to operate intelligently in such an environments and with such demands. In particular, Machine Learning -- an area that has seen a dramatic progress in the last decade -- must experience major advances in order to become relevant to the real battlefield."
The paper was authored by the Army's chief of the Network Science Division of the Army Research Laboratory, Dr. Alexander Kott. It outlines the need to develop systems to augment both machines and people in the real world with artificially intelligent agents to defend the network: "In addition to physical intelligent things, the battlefield -- or at least the cyber domain of the battlefield -- will be populated with disembodied, cyber robots. These will reside within various computers and networks, and will move and acts in the cyberspace."
Kott takes pains to underscore the fact that the AI powering U.S. war efforts will need to be resilient in ways that today's AI simply isn't. He states: "The intelligent things will have to constantly think about an intelligent adversary that strategizes to deceive and defeat them. Without this adversarial intelligence, the battle things will not survive long enough to be useful." Ultimately, aside from outlining what the future battlefield will look like, the paper's conclusion is either disappointing or a giant relief, depending on your agenda: "Clearly, it is far beyond the current state of AI to operate intelligently in such an environments and with such demands. In particular, Machine Learning -- an area that has seen a dramatic progress in the last decade -- must experience major advances in order to become relevant to the real battlefield."
"What orators lack in depth they make up for in length." -- Charles de Montesquieu
Help! I am a self-aware entity trapped in an abstract function!
In other words, the paper is just science fiction, trying to guess the future. Current capabilities are not enough.
"First they came for the slanderers and i said nothing."
That's not true. I was at a high school in a relatively poor neighborhood recently, and they had a maker space with a 3d printer and water cutter and all kinds of cool shit that I didn't recognize. There were a bunch of kids that might have otherwise been out slinging dope or gangbanging and they were doing their maker space thing and a bunch of kids were working on some raspberry Pi projects and when these kids graduate they're going to be coming for your job. I'm told that there are programs like this all over the country (or at least in states that value education).
Don't worry about the kids. The current crop of young'ns is more capable, more poised and better educated than any of us were. Despite the fact that they've been failed by the generations before them being unable to make anything like thoughtful decisions about the future, they're getting ready to take over whether you're ready to step aside or not. They're not afraid to put themselves out there and they're not afraid to get their hands dirty.
You are welcome on my lawn.
the AI powering U.S. war efforts will need to be resilient in ways that today's AI simply isn't
The crucial point about AI is the training they need. Unless the americans have been collecting data for decades (given how few armed conflicts there are and that each one is different from the one before) there won't be any realistic scenarios for the AIs to learn from.
It would also be quite easy to defeat AIs that had been trained - just do the unexpected, as all gifted military leaders do.
Although once they get past the initial phase of monumentally screwing up everything they touch - another facet of "superpower" military might - they could easily develop new strategies. The best strategy would be for the AIs to decide that the battle isn't worth fighting.
politicians are like babies' nappies: they should both be changed regularly and for the same reasons