Rep. Seth Moulton believes that killer robots could be here… today. And, not that much further ahead in the future, they could replace actual soldiers on the battlefield — which is why the Massachusetts Democrat is picking a fight with the Pentagon about it. In his eyes, the Defense Department hasn’t “done much of anything” to address concerns about the role artificial intelligence could play in the military down the line, he told Digital Future Daily in an interview following his op-ed in the Boston Globe on Wednesday warning of the imminent danger the tech poses to humanity. He likened the threat to nuclear weapons – and warned the United States has been ceding ground to foreign rivals. From his vantage point on the House Armed Services Committee – and as a top Democratic voice on military matters – the Pentagon military needs to be doing more, now. It’s worth noting that the DoD recently updated its autonomous weapons directive to follow the department's AI Ethical Principles policy, and military officials have repeatedly said there will always be a “human in the loop” when it comes to autonomous weapons killing people. But Moulton thinks that’s not enough. For starters, the Defense Department needs to lead by example and set ethical rules to prevent an automated or unintended atrocity on future battlefields, and capitalize on the military’s huge data reserves to advance artificial intelligence, he said. We spoke with the lawmaker about his fears and hopes for the future, and how he thinks Congress should respond to ensure that we’re all still living and breathing when killer robots become reality. Our interview has been edited for clarity and brevity. You’ve said the Pentagon isn't paying enough attention to AI's potential threats in military situations. What has DoD done poorly, and what can they improve upon? I don't think they've done much of anything. I talked to two service chiefs in the past week who both said the Pentagon is way behind on this. They've got no direction whatsoever, no guidance… The Pentagon is a $760 billion enterprise, and right now, we're spending half as much as China is on AI as a percentage of our defense plan. You're the first lawmaker I’ve heard make the direct comparison to AI being a threat comparable to nuclear weapons. What made you draw that conclusion? Why isn't everyone freaking out about this? Actually, I think there are a lot of people freaking out about it. What sets us apart from the nuclear age is that as soon as we developed nuclear weapons, there was a massive effort to curtail their use. It was led primarily by many of the scientists who developed this technology in the first place and recognized how dangerous it was to humanity, and it resulted in an international effort to limit nuclear arms and limit their proliferation. I just haven't seen anything comparable to that with AI. This is much more dangerous. China is investing tremendous resources in AI. Putin has come out and said that whoever wins the AI race will control the world. All of our serious adversaries are in a real competitive race with us on AI, and so we're losing the leverage to help set these international standards. How plausible is a “Geneva Convention on AI,” like you called for in the op-ed? I think it's plausible because I think everybody fundamentally recognizes the risk. It's easy to be a cynic here and say, well, we're at war with Russia in a proxy war and tensions are rising with China, but let's not forget that we had a lot of nuclear arms agreements during the Cold War. The Geneva Conventions were negotiated with a lot of tensions in the world. I think that this is hard, but it's absolutely worth trying. You also said that we're not far away from seeing killer robots in use. How far away are we? I think that we could produce them today, if we really wanted to. The question is just what a killer robot looks like, right? If you're just talking about autonomous drones, we basically already have that technology and we're developing very quickly into swarm and even more sophisticated forms. If you're talking about killer robots replacing an infantry man, that's probably still five or 10 years off, but it's not decades away. Data quality is a big part of moving on AI quickly and safely. What are your thoughts on the military's data classification system? The United States collects more military sensor data than any other country in the world. And we delete more data than any other country in the world, because we just don't have the place to store it. So, part of this is quite simple. We just need to build the storage capacity for all the sensor data we're collecting on a regular basis. Thank God, we did have the radar data to help us identify those balloons that had flown over from China. We looked back at Big Data to find that. A lot of that data is being deleted every single day. Is there anything in particular you're looking out for regarding AI in the National Defense Authorization Act? Well, there's competing philosophies about how to deal with this. One of the conversations I'll have with [Rep. Rob Wittman, R-Va., vice chairman of the House Armed Services Committee] is that I like the idea of jumpstarting thinking on AI, but don't want to set up an office that immediately exerts so much control over the services. In a public hearing, I encouraged the chiefs to go fast on AI and figure this out. The Pentagon secretary is not going to lead on this. So you should do it yourself. Much like the Marine Corps independently developed a strategy for the Pacific and now everybody else is trying to catch up.
| 
