 An intelligent machine needs great perceptive abilities. It needs great communicative skills. And if you look at the Hollywood vision from the 80s, you might know KIWT in Night Rider. A car with great sensing abilities, but also with great communicative skills. Talking to its driver, just like a human would to a human. And at the end, they got quite some good friends. But what does it actually take for a machine to be intelligent? And let me highlight just three aspects. It needs to be able to learn from data and then it needs to be able to achieve goals and ultimately reach brain-like or maybe even beyond brain-like intelligence. It's another example from the movies. You may know Johnny Five from the Short Circuit movie. And Johnny Five was going around and exploring the world by talking a lot to humans and constantly asking for more input from them. So he was interacting a lot with them and cooperatively learning with the humans about the world around. Now constantly asking may get on your nerves. And you might need more efficient ways to do that. By first having some known data with which the machine can learn about the world around and coming up with its first model of the world. But then you have the big unknown data that we've heard so much about these days. And the machine does not yet know what it is. And it asks itself, can I do by myself decide what this new data is? Or do you need help from the human? Or is it interesting enough at all to ask a human? So then it gets this helps. It learns from us what this new data means. And it can add it to its experience and enhance its model of the world. Now an intelligent system would then not only decide when to ask, but also whom to ask, just like a child would decide when to ask mommy, when to ask daddy about something. And that way around, you can these days reduce the number of questions an intelligent system has to ask to some 5% or even below to get a very rich representation of the world around. Now let us take this one step further and see how we can get the information in even more subtle ways. The movie Matrix, if you know it, had a rather sinister view on what machines will do in the future with us. They will use us as biobatteries. Let's share a more positive view on how they can benefit from us and say, rather than using us as these battery tanks that will exploit our knowledge and source from the crowd and source from us the knowledge. Taking this one step further, it could even directly scan your brain or access your brain waves and really learn a lot from us. So in other words, to build an intelligent machine, we will need human help in particular to tell the machine about the world and about the things. And how can we exploit that? Let me give you one example from my favorite application domain, and that is speech processing. You might have seen the movie Her, an intelligent machine embodied sort of only by voice, so only voice interaction, and the system and the user at the end of the movie or during the movie actually fall in love with each other. So the machine can send a lot from the voice of the user much more than just the words. In my ESC ground, I hear you, and in our startup or hearing, we have been doing that. We've been collecting a lot of information, sourcing the crowd about what is inside the speech. And the speech is out there. We've seen YouTube and other channels where you have a lot of speech, but we need the information for the machine what is inside. If you're doing that, you can sense a whole lot of things from the voice, such as the intoxication of somebody, the sleepiness of somebody, the cognitive load level at the very moment, also health related states such as autism or Parkinson's disease. You can sense the emotion, interest, height, personality, name it. But what you really need is loads of data, and you need human help to get information on what is inside in this data. Now taking this one step further, the machine learns from us, but can it even teach us? And by teaching, we reach a whole new level of insight into problems. If you take this new perspective of, how can I teach somebody else about something? Now Hollywood has also a vision on this. For example, in the movie Big Hero 6, you might have seen this medical robot who teaches a boy about moral values in life. It teaches a boy that revenge is not the right way to go. And how does this look in reality? In a European project I coordinated, we recently were trying to help autistic children by a computer system and an intelligent computer system to teach them in a playful way how to express emotion in a way comprehensible to others. So what this machine does is it uses its perceptive abilities to sense the body gesturing from a child, to sense the facial expression, to sense the voice, and from that, in a communicative way, tell the child in the best way how to better express emotions such that others would understand it. And it seems that intelligent machines in the future can really not only learn from us, but also teach us, leading to a loop of exchange between machines and humans, increasing intelligence of the machines. While we're lacking one component for a machine to really be intelligent, and that is emotional intelligence, maybe you've also seen Ex Machina, a movie inspired by my Imperial colleagues book, a machine that not only is able to have a human fall and laugh at her, but to exploit that human for her own goals. You can imagine how much social intelligence and emotional intelligence that takes to accomplish such a goal. And human mankind has had emotion as quite a survival factor in its history. It's my belief that in the future, all intelligent machines will have emotion as a survival key factor. And in fact, we are seeing this increasingly these days, this main project ended in 2010 and introduced artificial listeners that could sense your emotion from your gesturing, from your facial expression, from your voice, and then react in an emotionally appropriate and socially appropriate way. And new projects like the ARIA project at this very moment are picking up on this sort and taking emotional machines to the next level and lending them even more emotional abilities. So that means we have machines that learn from us, with us, teach us, and they have emotional capabilities, but we're lacking another ingredient. And this further ingredient is creativity. For a machine to be really intelligent, it needs to explore new ways, new solutions, and further explore a little bit what it can do and find new solutions to the goals it has to accomplish. Now, let me give you just two examples of what that can look like today. You have a machine exposed to visual art and to many of visual art pieces, finally training a deep neural network and being able to come up with its own visual art, like it or not, but certainly creative. Let me choose another example. Imagine after this talk, you would have an intelligent machine giving access to your MP3 collection, maybe in your pocket, on your player, and the machine comes up and surprises you with a whole new interpretation of what you have there. So let's have it pick one song, maybe of the classical type, the Canaan D. Pick a song that's used that, for example, a rock song, and only pick the vocals from that, the basket case. Do you have the time to listen to me why? And maybe pick the rhythm of some other song extracting only the drum beat to come up and surprise you with its whole new version of the songs. Do you have the time to listen to me why? The background is in everything I'll ever want. So machines can be creative to a point already, but it's my firm belief that we need much more computational creativity to have a machine really become intelligent in the future. So let me summarize a bit. We have the machine learn together with the human, go to the point where it teaches us, and from that learns even more. We have the machines enhanced by emotional intelligence leading from artificial intelligence to artificial emotional intelligence, and we give them creativity. So what remains for us to do at that point, the machines come up with their own version of the world knowledge here indicated by this pixel-esque image of the world. And what is needed for us to endure is that we remain in control. We have to know what is being trained in such learning algorithms such as deep neural networks. We have to ensure that it remains transparent what these artificial brains have learned for many ethical implications and just to make sure that we still know what these machines will be doing and thinking. Thank you very much.