I can do that, Dave: AI and warfare

Hey, Ryan Reynolds here wishing you a very happy half off holiday because right now Mint Mobile is offering you the gift of 50% off unlimited. To be clear, that's half the price, not half the service. Mint is still premium unlimited wireless for a great price. So that means a half day. Yeah, give it a try@mintmobile.com Switch upfront.

Payment of $45 for three month plan equivalent to $15 per month required new customer offer for first three months only. Speed slow 135 gigabytes of network busy taxes and fees extra. See mintmobile.com.

Hello and welcome to the Intelligence on Economist Radio. I'm your host Jason Palmer. Every weekday we provide a fresh perspective on the events shaping your world. Today. Executives from the Fukushima Daiichi nuclear plant, which melted down following a tsunami in 2011, were found not guilty of negligence. The verdict closes the book on criminal proceedings. But the disaster's aftermath is still playing out. And there's a new one of a kind restaurant in Cuba's capital. You can get authentic Chinese food surrounded by reminders of the close political ties between the two communist states. The fact that it took five years to open is one hint that business relationship a bit trickier. But first, There seems to be no aspect of human endeavor that isn't going to be affected by artificial intelligence. For the most part, AI will be hard at work behind the scenes. From recommending your next binge watch to recommending whether that spot on an MRI needs a closer look. Perhaps inevitably, military types are considering how AI can make recommendations on the battlefield. The concern is that war could be waged with a speed outside human control and a complexity beyond human understanding. An AI arms race is just beginning. But it's probably not going to look like you think it will.

When most people think about artificial intelligence, AI and warfare, their mind goes to kill the robots.

Shashank Joshi is our defense editor.

You know, smiling Terminator figures, wandering the battlefield.

I'll be back.

Actually, that's a way away that's very hard. You do have unmanned vehicles that can wander around a battlefield, fly in the air, but they tend to be quite nascent. They don't work in a very wide range of circumstances. They faint very easily. And so if you look at the robots that are out there today in battle, very few of them are really killer. The ones that are exceptions are loitering drones that can smash into radars hours after they've been released into the wild, and defensive guns that can shoot down incoming projectiles much more quickly than a human could. Defending a Base or a ship. They're not two legged creatures with cannons in their hand. That isn't really what, what AI is about today when you look at military applications.

So what is it that you think of then when thinking about AI and warfare?

I think there are a pair of other applications. One of them is about taking the torrents of information that you have in the modern age and processing it. So for example, in 2011, America's 11,000 or so drones sent back 37 years of footage. Obviously even the Pentagon cannot sit through all of that. But machine learning algorithms are able to sift through it, look for objects, look for suspicious actions. The other one is acting on that information. The point of military decision makers, of colonels and generals with all those medals on their chest, is that they have to make tough choices. Who do we attack? What's our plan? What's our strategy?

So when you say acting on those data AI making strategic decisions, how do you mean?

Well, you for example, have expert system platforms that could tell up Italian commander, what's the safest route I can take to cross that hill? And, and a machine can crunch all sorts of data on the terrain, on where the enemy is in a fraction of the time it would take staff officers to do it with old fashioned maps, charts, tools, and with much more accuracy. So that's at the basic level. But of course there may be bigger questions of military strategy. That is, what's the optimal way for me to repel a Chinese invasion of Taiwan that may look like they're the preserve of humans, but, but in due course, AI algorithms, particularly those that incorporate elements of randomness, elements of more sophisticated artificial intelligence techniques like deep learning, they may be able to offer up humans answers to those sorts of more complex questions.

And so it seems clear then that AI could make for good military strategists. We think of it as an inherently human thing because it brings in so many things. You think the world thinks that AI can do it?

AI is obviously quicker than humans, but I think what's more important is it may also be able to think in, in more creative ways that surprise us. Any of our listeners will be familiar with DeepMind's program AlphaGo, which in 2016 beat one of the world's best players in the ancient strategy game of Go, a game that was thought to be perhaps not tractable to an artificial intelligence. What people may not realize is that it did so with such aplomb that the Chinese People's Liberation army held a conference the next month to say, hey, what lessons do we get from this landmark victory on of an AI for military strategy. I think there's a tangible sense that modern machine learning, particularly deep learning, may be able to come up with moves in strategic ways that are so creative, so powerful, they might really begin to catch up or exceed humans ability to do the same.

But I mean, what was notable about AlphaGo's win was that it made moves that human players simply wouldn't and that confounded the great players. The AI came up with something that humans didn't really understand. Why on the battlefield. That's a tricky point.

It's a legal problem for a start. You know, military commanders have to abide by various principles of the laws of armed conflict. For example, the civilian harm has to be proportional to the military advantage gained. If you don't understand why the artificial intelligence is telling you to attack that building over there, you may not know whether it's complying with those laws. On top of that, even if you could solve the legal problem, you have a cultural problem. If the AI is proposing complex, mind bending moves that it says will yield results in a week's time, what if you don't trust those? I spoke to a Royal Air Force intelligence officer who's been working a lot on artificial intelligence and he suggested to me the example of an AI that proposes funding an opera in Baku in response to a Russian military incursion in Moldova. Now this is a surreal move. Humans may not be able to immediately grasp the reasoning behind this, but it may result from the algorithm grasping a political chain of events that we just don't see. And that's going to tax human judgment in very serious ways.

But if there are internationally agreed boundaries beyond which military commanders will not go, why can't those be coded into the AI?

I think the problem is some of these judgments seem very human for us, right? So judgments about proportionality may depend on how vital a military objective is to fulfilling your political goal. For example, that's going to be very, very hard for an algorithm to judge. Similarly, a critical principle is discrimination. You have to be able to discriminate between competence and non combatants. Now, okay, sometimes they're wearing uniforms and carrying guns and it's easy. But what if they're a local villager in Syria who you suspect is aiding Islamic State by conducting reconnaissance on your forces? Is he a competent? Is he a civilian? Humans judge these things in fuzzy ways that even they can't articulate sometimes. So how are they ever going to tell an algorithm how to make that judgment itself?

I mean, the logical extension of this is an entire nation's military system under the control of AI against another country's AI. Like the same sort of thing that we see in flash crashes on the stock market, where essentially the algorithms are battling it out. How do we mitigate the risks of the war equivalent of a flash crash?

Western governments and militaries will say, we will always have a human on the loop that is supervising things. The problem, of course, is that if an AI is able to act far more quickly than a human and you suspect your adversary is going to turn his or hers on, there is a strong incentive not to go. Second, if they simply push a button and the AI system is able to spot all of your tanks, all of your air bases and destroy them so fast you can't respond, you may be tempted to say, hang on a second, maybe we should turn over control to our artificial intelligence. And of course what that does is it compresses the time available for old fashioned diplomatic escalation.

But this technology is out there and its potential is recognized. That's surely making for an AI in warfare, international race. Is anybody winning it yet?

It's very hard to judge. We have no metric for progress in AI. Lots of this is taking place in secret. But we. What I will say is a lot of Americans are anxious that China's political system gives it an advantage both in terms of access to data from things like facial recognition systems in China. And data is what you train many of these algorithms on, but also in terms of being able to harness their technology companies for military purposes. Whereas in the Western world, lots of the Silicon Valley firms, lots of these tech firms don't want anything to do with the armed forces. One of the world's most advanced AI companies, DeepMind in the UK, owned by Google, has said it will never work on military or surveillance applications. So a lot of people say, well, we may have incredibly sophisticated private sector AI companies, but if the Chinese are able to strong arm their companies into working for the military, they will ultimately get a leap ahead of us. Clearly this worries lots of Americans. At the end of August, General Jack Shanahan, who heads up the Pentagon's Joint Artificial Intelligence center, gave a press conference and he said, what I don't want to see is a future where our potential adversaries have a fully AI enabled force and we do not.

Shashank, thank you very much for your time.

Thanks for having me.

Today in Tokyo, a court ruled in the trial of three executives from Tokyo Electric Power Co. Or TEPCo. It was the only criminal case brought against those involved in the 2011 Fukushima nuclear disaster. In March of that year, a tsunami engulfed Japan's northeast coast, flooding towns and farmland and severely damaging the Fukushima Daiichi nuclear plant. Almost 19,000 people were killed by the tsunami, and a further 160,000 were evacuated after the power plant's core leaked radioactive material into the sea.

This is the main road from Fukushima into Tokyo. It's the evening rush hour and the traffic should be going in that direction. Instead, thousands of people are trying to get out of the town and into the main city.

It was the worst nuclear disaster since Chernobyl in 1986.

It was quite eerie because there was nobody there. Pets had been abandoned, so there were feral dogs and so on.

David McNeil reports for the Economist from Japan.

Lights were off. You got the sense of what something must have looked like or what a place must look like during a war, except that there was no physical damage. That was the weird thing. And, of course, you can't see or smell radiation. And that increases the sort of the fear that it generates, sometimes the irrational fear that it generates.

And when you spoke to the people there, what did they say?

They were bewildered because they were getting conflicting messages. So one of the motifs throughout the disaster was the lack of information or the confusing information that was coming from both tepco, the company which ran this plant, and the government, and the sense that things were worse than the government was saying or both were saying. So you had this sense that something terrible was happening and people were not being told.

And now, eight years on, what happened.

In the courtroom today, three senior executives with Tokyo Electric Power, they were found not guilty of professional negligence resulting in death and injury. That's what they were charged with. Prosecutors said that those three executives were responsible for the deaths of 44 elderly patients at a hospital a few miles from the Daiji Power Plant. They died after a chaotic evacuation.

But what exactly were they accused of?

They were accused of failing to take action to prevent the meltdown. Now, what the critics, and, of course the prosecutors said was there was plenty of historical evidence of seismic activity in that area of northeast Japan, and the company should have prepared. For example, the cooling system was stored in the plant's basement, and TEPCO had ignored several tsunami studies and built the close the plant, I should say, too close to the coast, without seawall defenses.

So what was their defense?

Well, the defence of the executives is the same as the TEPCO defense, which they've said all along, which was, there is no way of knowing that the Tsunami could have triggered the triple meltdown. They have sort of said for years that the quake and the tsunami were once in 1000 year events and so strong as to be outside their expectations or the calculations of engineers. And specifically Katsumata, the chairman at the time, he said that he had relinquished control of the company day to day control of the company when he became chairman in 2008.

So this verdict is kind of the end of at least the legal end of the story.

It seems to be. There's a number of private lawsuits in the pipeline, but this was the only criminal lawsuit which put the men at the top in the dock for what happened.

And the people around Fukushima and in Japan more widely, who do they feel is responsible?

Well, tepco of course, bore the brunt of the public anger to the extent that people who wore TEPCO uniforms had to hide them. I think that anger has died down and now it tends to be a sort of a more amorphous anger and resentment towards the system, if you like. You know, the fact that tepco, which of course is semi nationalized now, is, was somehow in league with the experts and the government over the years and that they had obscured the problems in the industry and had failed to acknowledge the dangers of building all these nuclear power plants in a country that is so vulnerable to seismic shocks.

Nuclear power used to be a big fraction, a growing fraction of the country's electricity. What's happened to the nuclear industry since?

Well, clearly Fukushima has driven a stake through Japan's entire energy policy. Most of the nuclear power plants will probably never restart. There's about nine reactors currently in operation. There were 54 before the disaster. All of these plants, of course, are the subject of legal battles because of safety concerns. And the longer that these legal battles go on, the older the plants become and the more likely they are never to restart. So the reason why this decision today is quite important is because the world's biggest nuclear power plant, which is a seven reactor facility, Kashiwazaki Kariwa over in Niigata, is owned by Tepco. And TEPCO wants to restart several of those reactors. This gives them a boost.

David, thank you very much for joining us.

You're very welcome. Thank you for having me.

On the circular doors of the Beijing restaurant in Havana, you'll find huge Chinese characters that read Shuangxi. It means double happiness, a phrase often found on the cash stuff envelopes given as wedding presents. On the walls you'll find pictures of Cuban and Chinese communist bigwigs past and Present, what you won't find is table number four. It's an unlucky number in China. The newly opened Beijing is the first wholly foreign owned business in Cuba, and its history so far is a window into relations between the two countries.

The Beijing restaurant came about over a conversation between Xi Jinping and Raul Castro while on a flight to the Cuban city of Santiago.

Roseanne Lake writes about Cuba for the Economist.

They were enjoying some Chinese food on board and Xi Jinping said, we will put a good Chinese restaurant in Havana. And that was in 2014. Fast, fast forward five years later. Lots of money, lots of back and forth, and the restaurant is now finally open.

And why did it take so long then?

Well, there are many reasons. This is actually a story that has fascinated me for the past two years. I used to live in China, so anything Chinese in Cuba always gets my attention. And I'd walk past this building many, many times and you could see that it was pretty much fully restored. The first time I came across it, there was a team of architects from Beijing inside the building, which immediately got my attention. And I strolled in and I that they were communicating with Cuban laborers. There was clearly a language barrier. Basically, they communicate in expletives. And you could see that, you know, the construction of the building was pretty much set. Things were painted, things were in place. And every once in a while an inspector would come in and say, well, you know, this fire detector or this carbon monoxide detector, it's in the wrong spot. It would just seem like they were being very fussy about different elements of the building's construction.

But at last it's open. You finally got to go. How's it going in there?

Well, it is very much a Chinese restaurant in Cuba. I mean, the food is delicious. It's certainly very authentic. I sympathize with the staff and with the chefs at the restaurant because lots of basic goods, and this is true for any restaurant in Cuba, are difficult to find. So a dish like shihong shiji dan, scrambled eggs with tomatoes, isn't on the menu because it's very difficult to find the amount of eggs required to produce it and also tomatoes. So the restaurant looks beautiful, and it's certainly wonderful that people can enjoy Chinese food in this Cuban setting. It seems the restaurant has become a hotsp. Lots of Chinese business people in Havana who take advantage of several private rooms that it has to, you know, conduct their business.

Speaking of Chinese businessmen and investors in particular, is this something of a case study? I mean, do you think that the sort of hiccups and the shortages and the things being hung up in ports will, will put off other businesses that might want to set up shop in Cuba.

I do wonder if it could be somewhat of a cautionary tale, especially because this one got what one would think is sort of special treatment. At the time it was announced, the deal was the example of 100% foreign entity owned foreign investment in Cuba.

So with that in mind, do you think China is going to continue to invest in Cuba?

You know, China has certainly been very generous with Cuba. I would say they've just received a new fleet of trains that have come from China on a very generous loan of $150 million that help connect the country. Many of the newer cars that you see in Havana are Chinese from the brand Gili. And much of the debt that China has forgiven over the past few years has been to Cuba. So it's unclear sort of what the long term strategy is. My sense is that it is very much a long term one because the returns are certainly not going to be immediate. But one also has to think that strategically, from an ideological perspective, Cuba is an important partner to China, given its proximity as well to the United States, a country that is certainly not enjoying optimal relations with at the moment. And you know, given the situation in Venezuela as well, that has traditionally been a benefactor to Cuba, given that it's no longer able to do that, China is going to be playing an increasingly important role in the island.

So you are, you are clearly familiar, more than familiar with Chinese foods and you're an expert on Cuba. What do you make of the Beijing? What was your favorite dish?

Is my favorite all time favorite eggplant dish. And the Beijing did it really, really well. And I tend to only exclusively consume it with a side of Gambian zejito, those dry fried zone green beans. And they did it very well as well. Despite the complexities of procuring these dishes, the food is legitimate. The sounds coming out of the kitchen are also legitimate. You can hear the chefs sort of yelling at each other and saying, we've run out of this, we've run out of that. Which I think is probably going to be a very common tale for them. But given the circumstances that they're working under, I'd say they did a pretty top notch job of getting the food to taste like it should.

Roseanne, thank you very much for joining us.

Thank you, Jason.

That's all for this episode of the Intelligence. If you like us, give us a rating on Apple Podcasts and you can subscribe to the economist@examiner.com radiooffer 12 issues for $12 or 12 pounds. See you back here tomorrow.