WEBVTT - What Could Possibly Go Wrong? 0:00:04.600 --> 0:00:07.040 Sleep Workers is a production of I Heart Radio and 0:00:07.200 --> 0:00:16.560 Unusual Productions. I was on a patrol up in the 0:00:16.600 --> 0:00:19.960 mountains of Afghanistan, and I was doing a reconnaissance patrol 0:00:20.079 --> 0:00:23.760 with a very small ranger team, and I saw someone 0:00:23.800 --> 0:00:29.120 approaching us, young man, maybe early twenties. That's Polar, a 0:00:29.160 --> 0:00:32.640 former Army ranger, an elite special operative in the U. S. Military. 0:00:33.360 --> 0:00:35.519 There were sort of a couple of possibilities that came 0:00:35.560 --> 0:00:38.640 into my mind. One was that he was a goat hurter, 0:00:38.920 --> 0:00:41.440 just out, you know, taking his goats out to eat grass. 0:00:41.960 --> 0:00:45.160 Another was that he might be like a woodcutter that's 0:00:45.159 --> 0:00:48.240 protecting his property. And another one might be that he 0:00:48.640 --> 0:00:51.080 was someone who was had seen us. There wasn't a 0:00:51.080 --> 0:00:53.360 lot of vegetation in the area, so we were pretty exposed, 0:00:53.720 --> 0:00:56.240 and he was coming to kill us, And those all 0:00:56.280 --> 0:01:01.160 seemed like very real possibilities. PULL knew of similar situations 0:01:01.200 --> 0:01:05.160 where enemy fighters pretended to be civilians, concealing their weapons 0:01:05.240 --> 0:01:08.920 until the last minute ambush. I fairly id degree of 0:01:08.959 --> 0:01:11.040 confidence that, you know, if we got into a gunfight, 0:01:11.160 --> 0:01:14.120 we could outmatch him or three of us, but if 0:01:14.120 --> 0:01:16.280 you surprised us, he might have usially killed one of 0:01:16.319 --> 0:01:18.440 us in the process first, so I wanted to maneuver 0:01:18.480 --> 0:01:20.760 to a position where I could see him, and I 0:01:20.760 --> 0:01:22.200 found him. He was sitting on the edge of a 0:01:22.240 --> 0:01:25.760 cliff and looking out over this by valley, and he 0:01:25.800 --> 0:01:28.080 had his back to me, and I settled into a 0:01:28.080 --> 0:01:30.840 position where I could watch him very closely through my 0:01:30.880 --> 0:01:34.120 sniper rifle. It was close enough that the wind carried 0:01:34.120 --> 0:01:37.000 his voice towards me and I could hear him talking. 0:01:38.200 --> 0:01:40.440 That alarmed me because I couldn't see who was talking to. 0:01:41.240 --> 0:01:45.240 Paul didn't speak the local Afghan languages Dory or Pashtow, 0:01:45.720 --> 0:01:49.040 so he couldn't understand what the man was saying. I 0:01:49.080 --> 0:01:50.880 thought he might have had a radio and he could 0:01:50.920 --> 0:01:54.560 be reporting back to maybe a group of fighters a nearby. 0:01:54.960 --> 0:01:58.000 I was wearing that possibility versus maybe he's just you know, 0:01:58.080 --> 0:01:59.960 talking to himself or talking to his goats or something 0:02:00.000 --> 0:02:04.480 in And then eventually I heard him start singing, and 0:02:04.520 --> 0:02:06.880 it struck me that if he was singing, you know, 0:02:06.920 --> 0:02:09.440 singing out over the radio to someone in our position, 0:02:09.480 --> 0:02:12.040 it seemed like a very strange thing for a fighter 0:02:12.080 --> 0:02:14.320 to be doing. And the most like the explanation was 0:02:14.360 --> 0:02:17.239 he was just an innocent go to hurd Um had 0:02:17.320 --> 0:02:20.040 no idea that we were there, and he was singing 0:02:20.080 --> 0:02:24.639 to himself just to pass the time, and so Paul 0:02:24.800 --> 0:02:29.079 lowered his sniper rifle and he left, no provocation, no 0:02:29.160 --> 0:02:34.080 harm done to either side. But that incident really stuck 0:02:34.120 --> 0:02:35.720 with me because there was a period of time where 0:02:35.760 --> 0:02:38.720 I didn't know whether he was, you know, an innocent 0:02:38.760 --> 0:02:40.639 person or a fighter who might have been trying to 0:02:40.680 --> 0:02:43.200 kill us Um, And the actions I would take were 0:02:43.320 --> 0:02:46.400 very different in those two worlds. And now I look 0:02:46.440 --> 0:02:49.160 back when I think about autonomous weapons and I think 0:02:49.200 --> 0:02:51.920 what would a machine do? How could a machine understand 0:02:51.960 --> 0:02:55.600 the context that I did, and sort of realize it 0:02:55.639 --> 0:02:57.840 doesn't make sense that a person might be singing in 0:02:57.840 --> 0:03:00.640 a tactical way that's some strange is probably just an 0:03:00.800 --> 0:03:04.360 in person be able to do that. It's a profound question. 0:03:04.440 --> 0:03:07.880 Paul poses what happens when something like a Predator drone 0:03:07.919 --> 0:03:11.000 has as much autonomy as a self driving car and 0:03:11.160 --> 0:03:15.080 can an AI system ever understand context, which can sometimes 0:03:15.120 --> 0:03:19.360 mean the difference between life and death. In this episode 0:03:19.520 --> 0:03:22.440 will examine AI on the battlefield and the future of 0:03:22.480 --> 0:03:40.279 technology driven warfare. I'm as Velosen and this is Sleepwalkers. 0:03:40.280 --> 0:03:42.720 So care last episode we talked about the future of 0:03:42.760 --> 0:03:45.360 work and we focus on one big question, which was 0:03:45.720 --> 0:03:49.200 what can AI not do and what Paul is talking about? 0:03:49.360 --> 0:03:53.320 Identifying whether someone's a shepherd or an insurgent. Identifying targets 0:03:53.320 --> 0:03:56.320 on the battlefield seems to be one of those things, sure, 0:03:56.400 --> 0:03:58.080 But on the other hand, it's important for us to 0:03:58.080 --> 0:04:00.320 ask what is AI good at. You know, good at 0:04:00.320 --> 0:04:03.320 making predictions based on data about what might happen next, 0:04:03.560 --> 0:04:07.280 It's good at seeing patterns. So it makes me think, 0:04:07.320 --> 0:04:10.640 you know that with enough training data about battlefield interactions, 0:04:10.680 --> 0:04:14.600 it could get just as good or better than humans 0:04:14.600 --> 0:04:16.600 at this task. Yeah, and of course we don't have 0:04:16.680 --> 0:04:19.840 the counter effectual to pull story. Sadly, there's probably many 0:04:19.880 --> 0:04:22.359 stories where the army ranger doesn't wait to hear the 0:04:22.360 --> 0:04:25.040 shepherd seeing before deciding what to do and just pulls 0:04:25.040 --> 0:04:28.360 the trigger. And equally, there's probably many sad cases where 0:04:28.440 --> 0:04:30.160 it does turn out to be an insurgent and an 0:04:30.160 --> 0:04:33.000 ambush happens to the soldiers. So just from that story, 0:04:33.040 --> 0:04:35.599 it's not necessarily clear to me that humans are always 0:04:35.680 --> 0:04:38.400 or will always be better than algorithms that identifying targets. 0:04:39.000 --> 0:04:41.159 You know, I do think it raises an important question. 0:04:41.760 --> 0:04:45.240 We are increasingly comfortable outsourcing many parts of our lives 0:04:45.279 --> 0:04:47.159 to technology. You know, who are we going to fall 0:04:47.160 --> 0:04:48.839 in love with? How are we going to get to 0:04:48.880 --> 0:04:51.839 our cousin's house? You know, that's decision making. Though, in 0:04:51.880 --> 0:04:54.680 the world of war, we can also build tools to 0:04:54.720 --> 0:04:57.240 do things that people don't want to do. A lot 0:04:57.240 --> 0:05:00.760 of people have heard about Boston Dynamics. They have this 0:05:01.240 --> 0:05:07.640 extremely terrifying video of this ford legged robodog aptly named Spot, 0:05:07.839 --> 0:05:11.040 you know, that can run across bumpy ground. It can 0:05:11.080 --> 0:05:13.600 go into places that aren't safe for human beings, you know, 0:05:13.680 --> 0:05:16.919 And they've just announced this model with a claw on 0:05:16.960 --> 0:05:19.080 the back of it that can open doors. There are 0:05:19.080 --> 0:05:22.719 so many potential uses for this, from assisting military rates 0:05:23.240 --> 0:05:26.919 to sending Spot into buildings that are unsafe for humans. Yeah, like, 0:05:27.320 --> 0:05:30.159 a whole pack of Spots can actually pull a truck. 0:05:30.200 --> 0:05:32.440 There's a video of it online. Look it up. It's frightening. 0:05:32.680 --> 0:05:36.239 Teamwork is the dream work exactly? But no, like, what's 0:05:36.240 --> 0:05:39.120 going to happen when that comes to the world of 0:05:39.160 --> 0:05:42.600 war and we have packs of robodogs? Well, who knows? 0:05:42.680 --> 0:05:44.880 Right the future is unclear, and that's really the question 0:05:44.920 --> 0:05:49.159 of this episode. How will AI and robotics change the battlefield? 0:05:49.560 --> 0:05:51.479 And how good of an insurance policy is it to 0:05:51.520 --> 0:05:53.960 keep a human in the loop, to have somebody a 0:05:54.040 --> 0:05:57.640 person controlling the pack of spots. We have from Paul 0:05:57.720 --> 0:06:00.000 char at the beginning. He's now the director of Technology, 0:06:00.040 --> 0:06:03.360 g and National Security at a bipartisan think tank called 0:06:03.400 --> 0:06:06.679 the Center for a New American Security. He's recently written 0:06:06.680 --> 0:06:09.359 a book called Army of None, Autonomous Weapons and the 0:06:09.400 --> 0:06:14.440 Future of War. We're moving to a world where machines 0:06:14.480 --> 0:06:16.880 may be making some of the most important decisions on 0:06:16.920 --> 0:06:20.680 the battlefield about who lives and dies. And as Paul's 0:06:20.720 --> 0:06:24.640 experienced in Afghanistan taught him, even elite military training isn't 0:06:24.640 --> 0:06:27.200 always enough to tell who is and isn't likely to 0:06:27.200 --> 0:06:30.640 be a threat. So understandably there's a great worry in 0:06:30.760 --> 0:06:34.680 handing over target identification to a computer, especially as the 0:06:34.680 --> 0:06:37.920 stakes are life and death. And as Paul tells us, 0:06:38.080 --> 0:06:41.800 autonomous weapons are already being produced and sold around the world. 0:06:42.120 --> 0:06:46.120 The best example today of fully autonomous weapon is a 0:06:46.240 --> 0:06:49.360 drone built by Israel called the Harpy Drown and he's 0:06:49.400 --> 0:06:53.320 been sold to a number of countries Turkey, India, China, 0:06:53.480 --> 0:06:57.360 South Korea. It's an anti radar webbon that is lying 0:06:57.360 --> 0:07:00.520 a switch pattern in the sky looking for enemy radars, 0:07:01.040 --> 0:07:03.560 and when it finds one, it can then attack the 0:07:03.640 --> 0:07:07.080 radar within any further human permission. And that processes the 0:07:07.120 --> 0:07:10.880 line to an autonomous weapon that's able to go find 0:07:10.920 --> 0:07:13.640 targets and then attack them all by itself. These are 0:07:13.720 --> 0:07:17.320 not the autonomous killer robots of science fiction, setting their 0:07:17.360 --> 0:07:20.120 own goals and killing it will. We still send them 0:07:20.120 --> 0:07:22.520 into battle and tell them what to look for, but 0:07:22.600 --> 0:07:25.080 we no longer control exactly what they do when they 0:07:25.120 --> 0:07:28.280 get there. So by analogy, you might think about a 0:07:28.280 --> 0:07:31.800 self driving car. Um, there are really degrees of how 0:07:31.920 --> 0:07:34.240 much a car could be self driving. You could have 0:07:34.360 --> 0:07:38.400 some like the Tesla autopilot today, where there's a steering 0:07:38.400 --> 0:07:41.280 wheel and the human could intervene and grab control of 0:07:41.280 --> 0:07:44.320 the vehicle. UM, you might have some like Google self 0:07:44.400 --> 0:07:47.040 driving car that is no steering wheel and a person 0:07:47.200 --> 0:07:50.960 just merely a passenger. But even in a totally self 0:07:51.040 --> 0:07:54.400 driving car, the human is still choosing the destination. You're 0:07:54.400 --> 0:07:56.160 not getting in your car just in car, you know, 0:07:56.200 --> 0:07:58.280 take me where you want to go. So it's some 0:07:58.360 --> 0:08:01.120 level humans are always involved, and it's going to be 0:08:01.120 --> 0:08:03.960 the case in warfare. Um. The question is, you know, 0:08:04.000 --> 0:08:06.880 when do we cross the threshold where the humans have 0:08:06.920 --> 0:08:10.960 transfer control of some important and meaningful decisions two machines, 0:08:11.400 --> 0:08:14.760 and then what are the legal or ethical complications of that. 0:08:15.560 --> 0:08:18.720 It is a hugely important question. In the aftermath of 0:08:18.720 --> 0:08:21.560 the Second World War, a hundred and ninety six countries 0:08:21.600 --> 0:08:25.280 signed up to the Geneva Convention setting standards for behavior 0:08:25.360 --> 0:08:28.640 in battle, But how do we enforce those standards if 0:08:28.680 --> 0:08:32.760 the combatants are machines not people. Beyond having a human 0:08:32.760 --> 0:08:35.160 in the loop, one important piece of the puzzle is 0:08:35.160 --> 0:08:38.560 a healthy testing and review process, a clear understanding of 0:08:38.559 --> 0:08:41.160 how a weapon works and the decisions it will make. 0:08:41.840 --> 0:08:45.840 But according to Richard Danzig, a former secretary of the Navy, 0:08:45.880 --> 0:08:48.840 that's easier said than done. One of the things that 0:08:48.920 --> 0:08:54.280 concerns me is that the technologies are frequently highly classified. 0:08:54.800 --> 0:08:59.040 So we're for self driving cars, we typically require that 0:08:59.160 --> 0:09:03.240 there be millions of miles driven, and we insist that 0:09:03.400 --> 0:09:08.720 external regulators review them for safety. In the military context, 0:09:08.880 --> 0:09:12.560 we don't have millions of miles of experience before combat 0:09:13.320 --> 0:09:17.440 and we don't typically have any kind of third party 0:09:17.520 --> 0:09:20.680 review that says, wait a minute, here, the risks associated 0:09:20.720 --> 0:09:24.199 with this system spinning out of control. We ought to 0:09:24.240 --> 0:09:27.880 be using teams to say, hey, what could go wrong 0:09:27.920 --> 0:09:31.400 if an adversary wants to attack these and subvert them, 0:09:31.640 --> 0:09:34.960 or when they interact with other systems. This kind of 0:09:34.960 --> 0:09:38.360 war gaming is valuable, but the best laid plans and 0:09:38.440 --> 0:09:42.280 standards can crumble in the face of existential threats, real 0:09:42.440 --> 0:09:45.920 or perceived. You only need to think about Hiroshima and 0:09:46.080 --> 0:09:50.480 Nagasaki to understand how quickly restraint can give way to 0:09:50.600 --> 0:09:55.240 the desire for victory his pool Again, as countries feel 0:09:55.240 --> 0:09:57.880 that their national survival might be at stake, they're gonna 0:09:57.880 --> 0:10:00.520 be willing to take more risk put up more experimental weapons. 0:10:00.720 --> 0:10:02.920 The use of poison gas in World War One is 0:10:02.960 --> 0:10:05.600 I think a terrible example of this happening and practice. 0:10:05.679 --> 0:10:07.960 Germany was in a panic to find some kind of 0:10:08.000 --> 0:10:12.040 wonder weapon that might break that stalemate. So a desire 0:10:12.120 --> 0:10:16.080 to get an upper hand will clearly Historically we've seen 0:10:16.200 --> 0:10:19.719 lead militarios to take risks and deploy more experimental technology. 0:10:21.480 --> 0:10:23.439 This is one of the scariest things about war to 0:10:23.559 --> 0:10:27.599 me Kara, particularly war involving new technology, the potential for 0:10:27.679 --> 0:10:32.679 misunderstanding and unintended consequences. Paul mentions the almost accidental use 0:10:32.720 --> 0:10:35.520 of poison gas in World War One, and then there 0:10:35.559 --> 0:10:38.200 was a Cuban missile crisis where we almost stumbled into 0:10:38.200 --> 0:10:40.880 a nuclear war. And all this potential for misunderstanding is 0:10:40.920 --> 0:10:44.000 compounded exponentially in the world of AI because it's not 0:10:44.080 --> 0:10:46.120 just humans who are trying to read each other and 0:10:46.160 --> 0:10:49.600 make decisions. It's algorithms sort of loose in the wild, 0:10:49.960 --> 0:10:53.200 interacting with one another. The Guardian actually had a great 0:10:53.200 --> 0:10:57.000 story about this called Frank and Algorithms, The Deadly Consequences 0:10:57.000 --> 0:10:59.880 of Unpredictable Code, and the article makes the point that 0:11:00.080 --> 0:11:03.560 the stock market flash crash of was actually caused by 0:11:03.600 --> 0:11:07.120 algorithms interacting with one another. You know, it's hard not 0:11:07.200 --> 0:11:09.600 to think that this could happen in the wild, so 0:11:09.679 --> 0:11:11.960 to speak. It's a scary thought, and it's made even 0:11:12.040 --> 0:11:15.600 scarier because there's just so much potential all around for misunderstanding. 0:11:15.840 --> 0:11:19.280 And that's something Richard Danzig is seriously concerned about. When 0:11:19.400 --> 0:11:22.840 we talked about sending a ship on a mission, policy 0:11:22.920 --> 0:11:27.080 makers by and large understand what that means and how 0:11:27.160 --> 0:11:31.280 others will perceive it if the ship comes into their waters. 0:11:31.760 --> 0:11:36.520 Whereas when we start talking about artificial intelligence, policy makers, 0:11:36.559 --> 0:11:38.800 if there are five people in a room, may readily 0:11:38.920 --> 0:11:42.840 envisioned five different things. We need the people making decisions 0:11:42.880 --> 0:11:46.320 to understand both the situations they're dealing with and also 0:11:46.400 --> 0:11:48.920 how the tools they're using actually work. And that's not 0:11:49.000 --> 0:11:51.840 easy when it comes to AI. Part of the issue 0:11:51.960 --> 0:11:55.120 is the so called black box problem. Currently, we understand 0:11:55.160 --> 0:11:58.320 the principles of how a neural network uses probabilities to 0:11:58.360 --> 0:12:01.640 reach a conclusion, but we can't interrogate the millions of 0:12:01.679 --> 0:12:04.720 micro decisions it makes along the way. This is a 0:12:04.840 --> 0:12:09.000 huge barrier to understanding weapons systems powered by AI. I 0:12:09.040 --> 0:12:11.880 wanted to know how the research is developing new military 0:12:11.880 --> 0:12:15.280 technology think about the black box problem. So I spoke 0:12:15.320 --> 0:12:18.560 with Artie Prabaka, the former head of DARPA, the Defense 0:12:18.600 --> 0:12:22.959 Advanced Research Projects Agency, and Arthie shared a story about 0:12:22.960 --> 0:12:26.880 how the black box problem plays out. Many years ago, now, 0:12:26.920 --> 0:12:29.199 there was a wonderful paper from Stanford that showed a 0:12:29.280 --> 0:12:32.400 machine system that could label images. This was a girl 0:12:32.440 --> 0:12:35.440 blowing out the candles on her birthday cake, or a 0:12:35.480 --> 0:12:39.520 construction worker doing something so fairly complex analysis of what 0:12:39.600 --> 0:12:42.000 was going on in this picture, and it would get 0:12:42.200 --> 0:12:43.599 one right, and it would get ten right, and we 0:12:43.720 --> 0:12:45.720 get a hunder right. And then there was a picture 0:12:46.240 --> 0:12:49.200 that every human being would say, that's a baby holding 0:12:49.200 --> 0:12:52.560 an electric truth brush, but the machine said it's a 0:12:52.600 --> 0:12:55.880 small boy holding a baseball bat. And you know, you 0:12:55.920 --> 0:12:58.240 just look at it and you think, what what what? 0:12:58.080 --> 0:13:00.319 What were what were you thinking? And this I think 0:13:00.320 --> 0:13:03.520 it's a great illustration of the black box nature of 0:13:03.559 --> 0:13:07.040 these learning systems because they've been trained on all this 0:13:07.360 --> 0:13:11.280 volume of data, but when you look inside to say, well, 0:13:11.320 --> 0:13:13.480 what went wrong there, you know, you just see a 0:13:13.480 --> 0:13:15.800 bunch of nodes with weights from being trained, and so 0:13:15.960 --> 0:13:20.959 they're really opaque. Offi's example is kind of cute, but 0:13:21.120 --> 0:13:23.960 think about it for a moment. The difference between a 0:13:23.960 --> 0:13:27.560 baby holding an electric toothbrush and a small boy holding 0:13:27.559 --> 0:13:31.840 a baseball bat could also be in Afghanistan, the difference 0:13:31.880 --> 0:13:35.400 between Paul Shepherd and militant, in other words, the difference 0:13:35.400 --> 0:13:38.359 between life and death. Yeah, and it's a bit daunting 0:13:39.160 --> 0:13:42.080 that we are becoming more reliant on something that we 0:13:42.160 --> 0:13:45.959 continue not to understand fully, don't you think absolutely and 0:13:46.120 --> 0:13:47.920 Henry kissing you actually wrote a piece on this for 0:13:47.960 --> 0:13:52.200 the Atlantic called how the Enlightenment Ends Big Stuff for Sure, 0:13:52.200 --> 0:13:56.040 Hetty and what's kissing You? He argued that because we're 0:13:56.120 --> 0:13:59.080 unable to interrogate the output of algorithms, as we rely 0:13:59.120 --> 0:14:01.720 on them more immorti classify the world around us, we 0:14:01.760 --> 0:14:04.400 may actually start to lose the ability to reason for ourselves. 0:14:04.640 --> 0:14:07.680 It's not inevitable though, that AI will always be opaque. 0:14:08.160 --> 0:14:11.439 The EU are actually working on this policy that decisions 0:14:11.440 --> 0:14:14.800 made by AI need to be explainable to people they affect. 0:14:15.320 --> 0:14:17.880 That may be a policy that's easier said than done. 0:14:18.360 --> 0:14:21.720 Although the so called next wave of AI is all 0:14:21.760 --> 0:14:24.880 about explainable AI, and it's actually a major initiative right 0:14:24.880 --> 0:14:29.080 now at DAPPA. Here's off the again. Explainable AI has 0:14:29.120 --> 0:14:32.960 been part of starting an entire new field of inquiry 0:14:33.080 --> 0:14:37.360 and artificial intelligence to couple that kind of statistical power 0:14:37.920 --> 0:14:42.040 that machine learning systems have with systems that explain how 0:14:42.080 --> 0:14:44.480 they got the results that they got. In order for 0:14:44.600 --> 0:14:47.840 us human users to be able to know when to 0:14:47.920 --> 0:14:51.680 trust those machine learning systems and when not to trust them. 0:14:51.760 --> 0:14:55.440 What authe is describing would be a huge breakthrough in AI. 0:14:55.680 --> 0:14:58.960 Understanding how neural networks make their decisions would allow us 0:14:59.000 --> 0:15:02.480 to honess the power the technology much more safely, and 0:15:02.560 --> 0:15:06.000 not just on the battlefield. When we come back, we 0:15:06.000 --> 0:15:07.840 look at how much of the technology we take for 0:15:07.880 --> 0:15:11.520 granted in our everyday lives actually originated in the military. 0:15:17.800 --> 0:15:21.360 So DARPA, the defense agency with an annual budget of 0:15:21.440 --> 0:15:25.840 three point five billion dollars, its motto is to cast 0:15:25.880 --> 0:15:29.720 a javelin into the infinite spaces of the future. What 0:15:29.840 --> 0:15:32.000 you may not know is how much of the technology 0:15:32.040 --> 0:15:35.400 we all use every day came right out of Darper. 0:15:36.560 --> 0:15:39.040 I think about this every time I use my smartphone 0:15:39.080 --> 0:15:43.880 because that's a beautiful, seamless integration of a whole host 0:15:43.960 --> 0:15:48.440 of technologies that were sparked many many years ago by Darper. 0:15:48.560 --> 0:15:51.200 So the chip in your cell phone that talks to 0:15:52.000 --> 0:15:56.520 the cell tower is based on materials and electronics technology 0:15:56.560 --> 0:16:00.240 that was developed for communications, and radar systems that up 0:16:00.240 --> 0:16:03.600 that knows when you've rotated your phone is MEM's technology 0:16:03.640 --> 0:16:07.880 that had huge early support from DARPA. But also Series 0:16:07.960 --> 0:16:12.040 or other intelligent agents are based on the artificial intelligence 0:16:12.080 --> 0:16:15.160 research that was done. But what fueled this wave of 0:16:15.240 --> 0:16:22.240 incredible innovation at DAPPA. Well, actually, war DARPA is a 0:16:22.360 --> 0:16:27.480 very American concept. In nineteen fifty seven, the Soviet Union 0:16:27.640 --> 0:16:30.760 put the first artificial satellite on orbit. That was Sputnik. 0:16:31.200 --> 0:16:33.360 There was a lot of excitement because human beings had 0:16:33.360 --> 0:16:36.400 never done that before, but of course also quite a 0:16:36.480 --> 0:16:38.480 shock for the United States at the height of the 0:16:38.520 --> 0:16:42.120 CULD WARP. Sputnik was a reminder that in addition to 0:16:42.200 --> 0:16:45.040 working on the problems that you knew you had, you 0:16:45.240 --> 0:16:47.840 also needed to have people who came to work every 0:16:47.920 --> 0:16:51.760 day to think about those kinds of technological surprises. And 0:16:51.840 --> 0:16:56.040 so DARPA was started as a reaction to that technological 0:16:56.040 --> 0:16:59.960 surprise of Sputnik. It's mission for sixty years has been 0:17:00.120 --> 0:17:05.080 to create those kinds of technological surprises, and its history 0:17:05.400 --> 0:17:09.080 is one in which it's accomplished that mission. DARPA is 0:17:09.640 --> 0:17:13.119 known as the place that made the early stages of 0:17:13.320 --> 0:17:17.879 each revolution and artificial intelligence, and of course must memorably 0:17:17.960 --> 0:17:21.280 for starting the ARPA net and writing the protocols that 0:17:21.359 --> 0:17:24.880 became the Internet that we have today. Think about that 0:17:25.359 --> 0:17:28.560 the technology that forms the architecture of our daily lives. 0:17:28.560 --> 0:17:32.040 In the twenty one century, the Internet was created by 0:17:32.040 --> 0:17:35.760 a defense agency whose mission was to outthink the Soviet Union, 0:17:36.600 --> 0:17:39.000 and in some sense all of the technologies we've looked 0:17:39.040 --> 0:17:42.600 at so far in Sleepwalkers are the outgrowth of Dapper's work. 0:17:43.520 --> 0:17:46.760 What really enabled this was DARPA's decision to allow their 0:17:46.760 --> 0:17:50.520 technology into the US private sector and to let entrepreneurs 0:17:50.560 --> 0:17:55.679 build on top of it. Absolutely as vital was that 0:17:56.040 --> 0:17:59.480 were the companies and the entrepreneurs and the investors who 0:17:59.520 --> 0:18:03.440 saw that you could turn those those raw research results 0:18:03.560 --> 0:18:07.640 into this seamless, beautiful product that we've now we all 0:18:07.680 --> 0:18:10.959 live with all the time. Characters. Amazing to think just 0:18:11.040 --> 0:18:13.600 how much a Silicon Valley really stands on the shoulders 0:18:13.600 --> 0:18:16.399 of DARPA. But even outside of what we think of 0:18:16.480 --> 0:18:19.240 as the tech world, there's plenty of examples of military 0:18:19.320 --> 0:18:22.879 technology living with us. For example, microwaves, which were an 0:18:22.920 --> 0:18:26.800 accidental byproduct of radar technology, and then as good old Rumba, 0:18:26.880 --> 0:18:30.280 which was originally developed a mind sweeping technology and is 0:18:30.280 --> 0:18:34.760 still a mind sweeping technology in my house. Um. Yeah, 0:18:34.840 --> 0:18:37.520 you know. There's this concept of dual use technology, which 0:18:37.560 --> 0:18:40.639 we've talked about a few times. It's the idea that 0:18:40.720 --> 0:18:44.120 technology developed for the military can have civilian applications and 0:18:44.200 --> 0:18:46.840 vice versa. We talked about this with blink identity and 0:18:46.880 --> 0:18:51.359 facial recognition. Do you remember last year the project may 0:18:51.359 --> 0:18:54.080 even walk out? Yeah, Google, right, So that was a 0:18:54.080 --> 0:18:57.359 project for the Pentagon. Over three thousand employees protested that 0:18:57.400 --> 0:19:00.720 they didn't want to develop that technology. Google pulled the 0:19:00.760 --> 0:19:03.720 plug on the project. The problem is though that you know, 0:19:03.800 --> 0:19:06.880 you can say you're developing you know, AI and target 0:19:06.880 --> 0:19:09.920 recognition for the military. All you can say you're developing 0:19:10.000 --> 0:19:13.199 AI to recognize what's happening in images. But it's the 0:19:13.200 --> 0:19:16.080 same technology and once it gets into the wild, anyone 0:19:16.119 --> 0:19:18.440 who wants can use it. And that's actually something that 0:19:18.520 --> 0:19:22.720 Arthur you spoke about about innovation traveling from DARPA to 0:19:22.760 --> 0:19:25.520 the private sector. Yeah, and now DARPA actually has this 0:19:25.640 --> 0:19:29.679 younger sibling called the Defense Innovation Unit d i u X, 0:19:30.160 --> 0:19:33.280 whose job is actually to invest and incubate technologies from 0:19:33.280 --> 0:19:36.600 the private sector that could be helpful for defense. So 0:19:36.760 --> 0:19:39.720 this is basically the bridge from Silicon Valley to d C, 0:19:40.320 --> 0:19:43.960 which is taking things from the consumer space and applying 0:19:44.040 --> 0:19:46.919 them for military use. And one of these technologies is 0:19:46.920 --> 0:19:50.800 called Halo. They're basically headphones that electrify your brain in 0:19:50.880 --> 0:19:55.240 order to stimulate growth. That's right. So I went to 0:19:55.280 --> 0:19:58.600 Connecticut to test them out with former Navy Seal John Wilson, 0:19:59.760 --> 0:20:02.680 OH a seal for twelve years. I've served in Iraq 0:20:03.080 --> 0:20:08.240 multiple times, Afghanistan, I went to Mogadishue, and then South America. 0:20:08.480 --> 0:20:10.920 As you can imagine, drug warfare is still a really 0:20:10.960 --> 0:20:13.520 big issue, so we have military units down there to 0:20:13.880 --> 0:20:16.600 combat the cartels. We were gone three hundred days out 0:20:16.600 --> 0:20:19.080 of the year training and then we were deployed from 0:20:19.080 --> 0:20:23.080 once on end. So that's what we lived, breathed day 0:20:23.080 --> 0:20:24.800 in and day out. We weren't going home at night. 0:20:25.440 --> 0:20:28.000 These days, John is back with his family and he's 0:20:28.080 --> 0:20:31.480 especially interested in how new technology can help seals past 0:20:31.520 --> 0:20:35.520 and present. One such technology is the Halo headset, which 0:20:35.600 --> 0:20:38.520 d I u X invested in. The headset uses an 0:20:38.520 --> 0:20:42.720 electric current to prime the brain for so called neuroplasticity, 0:20:42.760 --> 0:20:45.960 in other words, the ability to learn and learn quickly. 0:20:46.400 --> 0:20:49.160 For us, I've recognized and when we do a pistol draw, 0:20:49.359 --> 0:20:52.600 that movement is a repetitive movement that we've done thousands 0:20:52.640 --> 0:20:55.280 and thousands and thousands of times over and over again. 0:20:55.520 --> 0:20:58.040 And what this does is it primes the brain to 0:20:58.200 --> 0:21:01.440 learn those repetitive movements faster? And have you genuinely knows 0:21:01.520 --> 0:21:04.680 the difference. When I first came across this, I've got 0:21:04.680 --> 0:21:06.480 a bunch of seals together and went out to the range. 0:21:06.520 --> 0:21:09.080 We neuro primed and we started shooting. So we got 0:21:09.119 --> 0:21:11.639 the baseline. Did this for a month, looked at our 0:21:11.680 --> 0:21:15.000 scores and our scores were light years better. And light 0:21:15.080 --> 0:21:17.840 years I mean milliseconds, right, But milliseconds on the battlefield 0:21:17.840 --> 0:21:20.720 equates the life or death. I may be as far 0:21:20.760 --> 0:21:23.679 away from being a military person as anyone could be, 0:21:24.200 --> 0:21:26.240 but John has kindly agreed to lead me through a 0:21:26.359 --> 0:21:29.200 Navy seal workout. So what we have here it looks 0:21:29.200 --> 0:21:33.160 like a beats headset right with some strange noduals. Yeah, 0:21:33.560 --> 0:21:35.760 so what we have on the top has mentioned were 0:21:35.800 --> 0:21:38.679 these little nods. What those do? Or those are going 0:21:38.760 --> 0:21:42.640 to send a current into the cortex or frontal cortex 0:21:42.680 --> 0:21:46.000 into your brain? Is that safe? That's a great question. 0:21:47.000 --> 0:21:51.000 I'm gonna bypass that question. It's safe. Turns out it 0:21:51.080 --> 0:21:53.680 is safe. It's been tested by Dapper and others. So 0:21:54.080 --> 0:21:56.240 I had to put it on and John agreed to 0:21:56.240 --> 0:21:59.320 help me use the headset to neuro prime before putting 0:21:59.320 --> 0:22:01.879 me through my pace is all right, let's let's crank 0:22:01.880 --> 0:22:04.160 it up so you can. So it's probably a good 0:22:04.400 --> 0:22:05.800 I don't know if you can feel a difference there, 0:22:05.800 --> 0:22:07.879 but we've got twenty minutes of neuro priming. It feels 0:22:07.880 --> 0:22:10.439 like something pinching in my head a bit. Yeah, it's not. 0:22:10.600 --> 0:22:13.960 It's not a comfortable feeling, right, yeah, and you imagine, no, 0:22:14.080 --> 0:22:17.159 you never do. But it's worth it, right, What is 0:22:17.160 --> 0:22:20.720 that uncomfortable sensation? What's happening that? That's electricity? Yea, so 0:22:20.800 --> 0:22:23.320 it's it's going into your brain right now, and it's 0:22:23.320 --> 0:22:26.680 getting your brain into a state of neuroplasicity. Hyper Learning, 0:22:26.800 --> 0:22:28.879 essentially is what that state allows you to do, and 0:22:28.880 --> 0:22:32.119 it just allows you to learn faster and learn more information. 0:22:32.800 --> 0:22:35.199 After a warm up, which was in fact a lot 0:22:35.280 --> 0:22:39.719 more intense than my normal workout, two three, come on up, 0:22:39.880 --> 0:22:44.760 drive up and hold right still the warm up, Yes, 0:22:44.760 --> 0:22:46.760