[MUSIC] What are we actually doing when we create new knowledge? What are we doing when we do science? Well, we use intelligence in order to discover something new, but what is intelligence? And working with artificial intelligence, we also happen to understand much more about this entire idea is about of intelligence. And that's nothing new, we often look at our technologies and then actually understand what something is about. For example, look at human flight, right? We we always thought something flying has to do with feathers, because the only things we usually saw flying in human history were birds. So from da Vinci to anybody else, we always thought, we need feathers and need to flap them, probably, in order to be able to fly. They had no idea what aerodynamics actually is. The brothers Wright, not too long ago, 100 and something years ago, the first time they were flying for 30 meters or 100 feet, they almost killed themselves. They had no idea how this stuff works. So they built the first flying machines, quite dangerous, without knowing what they were doing. Those two survived, many others didn't, we were gambling at that point. But then once we had these flying machines, we could take them apart and look inside, and start to understand what flying actually is. Well, what actually is flying? But it doesn't have to do with feathers, it has to do with the curvature of the wings. So if you see these huge airplanes, they have these little tiny wings, but they're curved and actually with suction, they're sucked up. Well, then we understood actually what this thing is about, right? And we could build all kinds of different flying machines, then. Flying machines that nature never came up with, helicopters, fighter jet planes, and rockets. And after the brothers Wright, only 60 years later, we were flying to the moon. Nature never invented something that could fly to the moon, at least not on this planet. As far as we know, evolution ever came up with that. But once we understood the principle of what it actually is, we were doing things that nature with this evolutionary tinkering never came up with. And we understood all kinds of flying techniques and approaches because we understood the framework of what aerodynamic actually is. So something similar is happening right now with intelligence. Nature came up with one solution to the intelligence problem, and that's what nature came up with. That's kind of like the birds and the feathers. Now, seeing how a machine learns, we actually start to understand the different kinds of intelligence. We're starting to have a glimpse of the bigger picture of what it actually is to create knowledge, what it is to learn, by these machines that we create that actually do exactly that. And these machines get so good at it that it actually becomes a little bit scary. And that brings us to discussion, if you open the newspaper nowadays, you see it. We don't have to go to the Terminators, and to how artificial intelligence will kill us off because we are the inferior species, no need to go there. Many people are worried that artificial intelligence is taking their jobs, for example. And even the biggest companies on planet Earth that do machine learning, artificial intelligence massively, they have a very healthy dose of respect towards it. Which led to one initiative, that many companies now open up their artificial intelligence. So if you want to do artificial intelligence, companies like Google and Amazon, they open up. Well, they open up an empty artificial brain, they don't tell you how they trained it. It's kind of like, the business value is still with that, for 20 years, they've been training this brain. So they have the data they trained it with, kind of like they have a university graduate, whereas what they give to you is kind of like a newborn baby. You can train it yourself and it's available, and during the course, we will play. We will play with some of the artificial intelligence that's opened up by big companies that do artificial intelligence. And we will be able to play with it in order to do some machine learning for ourselves, we have a lab of doing that. And we will have to learn that, we will have to cooperate with intelligence. Traditionally, we're always afraid of a new technology, because it might be taking our jobs, it might replace us. And that's a very old fear, and it's an old story. In the 1800s, the Luddites famously destroyed cotton and weaving machines because they were taking their jobs. They thought, my goodness, with these cotton and weaving machines, there won't be work anymore. Well, 200 years later, we still have work, we just don't use cotton weaving machines, we automated that. And there are many tasks in intelligence that, kind of like automating, we start to cooperate now with these machines. And that's something, if you go to one of the founders of computer science, John von Neumann. He famously said, the best we can do is to divide all processes into those things which can be done better by machines, and those which can be done better by humans, and then invent methods to pursue the two. So that was the idea since the beginning of computer science. And we have to have methods to cooperate, collaborate with our machines. Interestingly enough, that's also what we've been finding when we try to match ourselves with the machines. Of course, the first division would always be, my goodness, who is stronger, the human or the machine? Man against machine, right, think about chess. So in the late 1990s, the end of last century, we lost that battle in chess. We took our best chess player, Garry Kasparov, we went against a computer from IBM, Deep Blue, and we lost that one. We really did, and that's been several decades ago now, right, and we see AI always advancing. Interestingly, afterwards, Kasparov, he had a couple of choices of what to do. He could have gone home, studying more chess, or he could have gone home and buried himself in hiding in the woods and said, my goodness, the machines are too powerful, even more powerful. But actually, what he did is, he started to cooperate with these machines. He started to invent something called freestyle chess or centaur chess. Where actually, computers are allowed, and you play chess with the computers. What they found there is that the most successful teams were not the grandmasters, and also not the supercomputers. Interestingly enough, also not the supercomputers and the grandmasters. It was actually teams that were very good in cooperating with these machines. So let's say one IT guy and one medium chess player, but they were very good buddies, and they really knew what to do. And they cooperated with the machines very well, and they were beating the grandmasters with the supercomputers. So actually, the best we can do is get in touch with machine learning, separate these processes. What can machines do better in discovering knowledge, and what can we do better? And then cooperating with them, and we find out that this is the most powerful approach. Also to do science, and in that sense, do computational science, computational social science especially.
