“I don’t see why machines can’t be creative”: An interview with Kristian Kersting
There’s an old saying in computer science circles that goes along the lines of “everything we don't understand yet is called artificial intelligence.” Kristian Kersting from Germany’s new Hessian Center for Artificial Intelligence says he wants more awareness around machine learning and algorithms to strengthen future thinking and debate on AI.
Back in 2018, the artwork Edmond de Belamy, which was created using an intelligent algorithm, was auctioned by Christies for 432,500 US dollars. Kristian Kersting says that while many at the time debated whether the machine or the artists involved owned the copyright, he and his colleagues were wondering ‘does this show that machines can be creative?’
“When we look at how we define creativity, we quickly realise how difficult it is,” admits Professor Kersting, one of the Founding Research Co-Directors of the new Hessian Center for Artificial Intelligence. Does creativity, for example, imply an own free will and initiative to start drawing or combine new ideas? And, what about consciousness?
“Many people say machines can't be conscious,” Kersting says. “But the problem for me is, unfortunately, we don't have a measurable definition of what consciousness is. That’s why I remain open to the idea and say: I think that's a great and interesting direction to explore.”
Kersting says that for him it’s always a bit like the Turing test — the imitation game that tests a machine's ability to exhibit intelligent behaviour indistinguishable from a human. If he can't tell the difference between machine and human, that's great progress.
“I find it very exciting that some artists now view computers as muses and inspiration. I think that’s really great. I don’t see why machines can’t be creative. Maybe I wouldn’t say we already have an artificial Jeff Koons — that will still need much more.”
“I like this childlike approach of combining, focusing on the positives and simply exploring,” he adds.
The AI-produced painting "Edmond de Belamy" sold for top dollar in 2018
| © Christie’s Images Limited 2018
However, an interdisciplinary approach is crucial argues Kersting. He sees AI as a type of team sport, to make the most of designing and using AI for the greater good. The new Hessian Center for Artificial Intelligence, for instance, is unique in Germany as it’s supported by 13 different universities across its local Hesse region. This also reflects the institution’s plans of a broad approach to research and solution-finding on some of the most important challenges of our time.
The centre’s research, which is still being mapped out, will also create an environment to tackle the intersection of AI and culture. Teams of scientists will be asking diverse questions such as ‘Can machines understand humour or make sense of historical texts’? Or ‘how can AI be used to piece back together century-old remains in archaeological projects?’
For Kersting the questions of who we are and what defines us lie at the heart of psychology, cognitive sciences, biology and also artificial intelligence and can be found all throughout history. German philosopher Gottfried Wilhelm Leibniz pondered whether it was possible to formalise thinking, for instance, and even the ancient Greeks thought about Talos, a robot-like creature. However, Kersting argues the difference today is that humans now have machines that can think and that start to develop commonsense, so we don’t just think about the issues, we create new options too.
“We cannot only ask how we can make intelligence more measurable and even observe it, but suddenly we can also simulate or produce it to a certain extent,” he says. “I can understand that this scares some people at first, because we are used to being the culmination of creation. People have the feeling: “‘Oh God, am I still special at all?’”
Kersting with his fellow Founding Research Co-Director at the Hessian Center for Artificial Intelligence, Professor Mira Mezini
| Photo: Credit Claus Völker, © TU Darmstadt
Kersting says that fears about artificial intelligence are particularly pronounced in Germany, while in China or the US people are more open about AI and embrace it, even if they have concerns.
“Delivery Hero or similar services for example have been around in America for a long time and people think about it this way: ‘Can the drone deliver to my window on the 15th floor? Why do I still have to run down to pick things up?’”
Kersting wonders whether his compatriots’ concerns are the result of a lack of knowledge and awareness around the issues. Or whether a perceived heightened awareness for privacy and data protection in Germany, and most of Europe, is to blame.
“I don’t think people have a general aversion to data collection,” Kersting says. “Things might look very different if data was publicly and transparently collected and also made accessible to others. What if there was a data collection system that was open and accessible to many companies and benefiting citizens. Why not? Collecting data might then not be that difficult after all.”
Kersting, whose Artificial Intelligence and Machine Learning Lab at the Technical University of Darmstadt won the German AI Award in 2019, recently published a book with his students to demystify the field of AI. He emphasises that people shouldn’t talk about the field without knowing what is possible and realistic. But, they shouldn’t leave it too late to learn about the technology either.
“In Germany we have the tendency to first discuss and determine everything before we start, that’s the German nature,” Kersting says. “I would like it if sometimes we could show a bit more creativity and ask childlike questions. A “let’s-just-try-and-see” approach. I don't think we can plan the future and determine every step in advance.”
TU Darmstadt, which will work in partnership with Hessian.ai, is one of Germany's leading technical universities
| Photo: Thomas Ott / TU Darmstadt
Often the debate around defining intelligence is too humanized, deplores Kersting. “Artificial intelligence, it's not always about people. Humans are a great example, because we don't know many other systems that are just as intelligent”.
Kersting uses the example of autonomous driving. “If we were to take people as role models here, we wouldn't achieve anything new. We would still have just as many traffic jams and as many accidents. The aim is not always about imitating humans, but rather to really understand and create intelligent behaviour — that can sometimes even be that of a fly,” he says.
Professor Kristian Kersting
| Photo: Katrin Binner / TU Darmstadt
The core question is how can computer sciences and AI help and improve the world, Kersting suggests. “There is some great work with AI out there on how to avoid overfishing and still have enough fish to feed. It's about giving politicians or fishermen a policy in a mathematically optimal sense. And if you follow those, you will still be able to fish in ten years’ time.”
“People have a lot of distorted views and if possible I would like to fix them so that we can live in a better world.”
Of course this shouldn’t imply that every AI application has to be good, Kersting admits. “For example I don't want an atomic bomb, let alone an atomic bomb controlled by an autonomous AI system. Nevertheless, I think radiation is important and good in medicine.”
“What we need is the discussion of what we want and what we don't want.”
Learn more about Kristian Kersting's views on the future of creative AI here.
“When we look at how we define creativity, we quickly realise how difficult it is,” admits Professor Kersting, one of the Founding Research Co-Directors of the new Hessian Center for Artificial Intelligence. Does creativity, for example, imply an own free will and initiative to start drawing or combine new ideas? And, what about consciousness?
“Many people say machines can't be conscious,” Kersting says. “But the problem for me is, unfortunately, we don't have a measurable definition of what consciousness is. That’s why I remain open to the idea and say: I think that's a great and interesting direction to explore.”
Kersting says that for him it’s always a bit like the Turing test — the imitation game that tests a machine's ability to exhibit intelligent behaviour indistinguishable from a human. If he can't tell the difference between machine and human, that's great progress.
“I find it very exciting that some artists now view computers as muses and inspiration. I think that’s really great. I don’t see why machines can’t be creative. Maybe I wouldn’t say we already have an artificial Jeff Koons — that will still need much more.”
“I like this childlike approach of combining, focusing on the positives and simply exploring,” he adds.
The AI-produced painting "Edmond de Belamy" sold for top dollar in 2018
| © Christie’s Images Limited 2018
Who are we? And what defines us?
However, an interdisciplinary approach is crucial argues Kersting. He sees AI as a type of team sport, to make the most of designing and using AI for the greater good. The new Hessian Center for Artificial Intelligence, for instance, is unique in Germany as it’s supported by 13 different universities across its local Hesse region. This also reflects the institution’s plans of a broad approach to research and solution-finding on some of the most important challenges of our time.
The centre’s research, which is still being mapped out, will also create an environment to tackle the intersection of AI and culture. Teams of scientists will be asking diverse questions such as ‘Can machines understand humour or make sense of historical texts’? Or ‘how can AI be used to piece back together century-old remains in archaeological projects?’
For Kersting the questions of who we are and what defines us lie at the heart of psychology, cognitive sciences, biology and also artificial intelligence and can be found all throughout history. German philosopher Gottfried Wilhelm Leibniz pondered whether it was possible to formalise thinking, for instance, and even the ancient Greeks thought about Talos, a robot-like creature. However, Kersting argues the difference today is that humans now have machines that can think and that start to develop commonsense, so we don’t just think about the issues, we create new options too.
“We cannot only ask how we can make intelligence more measurable and even observe it, but suddenly we can also simulate or produce it to a certain extent,” he says. “I can understand that this scares some people at first, because we are used to being the culmination of creation. People have the feeling: “‘Oh God, am I still special at all?’”
Kersting with his fellow Founding Research Co-Director at the Hessian Center for Artificial Intelligence, Professor Mira Mezini
| Photo: Credit Claus Völker, © TU Darmstadt
Lack of knowledge or fear for privacy?
Kersting says that fears about artificial intelligence are particularly pronounced in Germany, while in China or the US people are more open about AI and embrace it, even if they have concerns.
“Delivery Hero or similar services for example have been around in America for a long time and people think about it this way: ‘Can the drone deliver to my window on the 15th floor? Why do I still have to run down to pick things up?’”
Kersting wonders whether his compatriots’ concerns are the result of a lack of knowledge and awareness around the issues. Or whether a perceived heightened awareness for privacy and data protection in Germany, and most of Europe, is to blame.
“I don’t think people have a general aversion to data collection,” Kersting says. “Things might look very different if data was publicly and transparently collected and also made accessible to others. What if there was a data collection system that was open and accessible to many companies and benefiting citizens. Why not? Collecting data might then not be that difficult after all.”
Kersting, whose Artificial Intelligence and Machine Learning Lab at the Technical University of Darmstadt won the German AI Award in 2019, recently published a book with his students to demystify the field of AI. He emphasises that people shouldn’t talk about the field without knowing what is possible and realistic. But, they shouldn’t leave it too late to learn about the technology either.
“In Germany we have the tendency to first discuss and determine everything before we start, that’s the German nature,” Kersting says. “I would like it if sometimes we could show a bit more creativity and ask childlike questions. A “let’s-just-try-and-see” approach. I don't think we can plan the future and determine every step in advance.”
TU Darmstadt, which will work in partnership with Hessian.ai, is one of Germany's leading technical universities
| Photo: Thomas Ott / TU Darmstadt
It’s not always about people
Often the debate around defining intelligence is too humanized, deplores Kersting. “Artificial intelligence, it's not always about people. Humans are a great example, because we don't know many other systems that are just as intelligent”.
Kersting uses the example of autonomous driving. “If we were to take people as role models here, we wouldn't achieve anything new. We would still have just as many traffic jams and as many accidents. The aim is not always about imitating humans, but rather to really understand and create intelligent behaviour — that can sometimes even be that of a fly,” he says.
Professor Kristian Kersting
| Photo: Katrin Binner / TU Darmstadt
The core question is how can computer sciences and AI help and improve the world, Kersting suggests. “There is some great work with AI out there on how to avoid overfishing and still have enough fish to feed. It's about giving politicians or fishermen a policy in a mathematically optimal sense. And if you follow those, you will still be able to fish in ten years’ time.”“People have a lot of distorted views and if possible I would like to fix them so that we can live in a better world.”
Of course this shouldn’t imply that every AI application has to be good, Kersting admits. “For example I don't want an atomic bomb, let alone an atomic bomb controlled by an autonomous AI system. Nevertheless, I think radiation is important and good in medicine.”
“What we need is the discussion of what we want and what we don't want.”
Learn more about Kristian Kersting's views on the future of creative AI here.