AI – Technology should support people
Interview with the Director of the ELLIS initiative, Prof. Dr. Bernhard Schölkopf
Artificial intelligence has become the key technology for staying competitive. The European initiative ELLIS (European Laboratory for Learning and Intelligent Systems) is intended to bring more cutting-edge AI research to Europe. The basic concept of ELLIS is that AI researchers at universities work closely with basic researchers from industry. Siemens AG has contributed one of its experts, Prof. Dr. Volker Tresp, to this initiative as Co-Director. Prof. Dr. Bernhard Schölkopf from the Max Plank Institute serves as Director of the ELLIS initiative. Here’s an interview discussing his visions for AI.
How did you become interested in artificial intelligence?
Looking back on it, my journey to AI wasn’t really planned. As a physics student, I participated in a study group at the School of Theology. We read about and discussed constructivism and connectionism. And I once took part in a conference that happened to be held at the local Max Planck campus. There I learned about research into the neurosciences. I thought it was fascinating but too complicated. I was looking for something closer to mathematics. When I heard about a practicum program at Bell Labs that was being offered by the Academic Foundation and the possibility of working in the field of machine learning, I was thrilled.
What values do you think are especially important in research that should also be taken into account in industrial research?
The European academic tradition is based on a culture of open discussion and debate, and that’s one of the reasons why we want to strengthen European research and innovation in the field of learning AI. Because we’re talking about technologies that will have a transformative influence on society, it’s especially important that we deal equally with the opportunities, challenges, and risks of AI. The inclusion of ethical concerns in the early stages of development is crucial both in basic research and in industry, and many companies in Europe have already begun to take this seriously.
Society’s visions of the future of AI include either overblown expectations, such as transhumanistic utopias, or dystopian fears of ubiquitous surveillance and manipulation. What would you like to say to these people? Do you yourself have a utopian concept of the future?
Machine learning systems are currently being trained to solve specific problems. There’s nothing like an all-encompassing superintelligence that replaces people: That’s the stuff of science fiction. Like any other technological development, AI has the potential to be used for good or nefarious purposes. Given the right regulatory and political framework and the research conditions in open societies like I just mentioned, I’m confident that we’ll be able to design useful technologies that will have a positive influence. I’m no fan of transhumanistic utopias, but I see technologies being developed that have the potential to help solve many societal challenges.
What does basic research mean in the field of artificial intelligence? What results should this basic research be able to produce in the next five to 10 years?
Basic research is by definition free, independent, and driven by curiosity. Science deals with basic problems that humanity hasn’t yet solved, which is why it’s difficult to make predictions based on a specific timeline. But I hope that our work on training machine learning and obtaining a better understanding of causality will result in more diverse forms of artificial intelligence.
Fascinating Research & Technology
Insights into our research labs and technical backgrounds
How will the future use of AI – on a timeline of about 20 years – differ from today? Do you think that AI will noticeably change our lives over, say, the next two decades? If yes, how?
AI is slowly finding its way into all areas of our lives, and I expect this trend to become even more pronounced. There are countless examples of how our everyday lives are being changed: for instance, the way we drive. Over the next ten years, it’s conceivable that cars will mostly (though probably not exclusively) operate autonomously. Assistance systems like Alexa and Siri will develop in such a way that in ten years they’ll be able to engage in reasonable conversations. Beyond ordinary life, AI will be increasingly used in fields like healthcare, where intelligent systems will support physicians by offering diagnoses and recommending treatments. This is not to mention the use of AI to improve production and logistics in industry. A lot of this is already being developed to market maturity: for example, by startups in the Cyber Valley Start-up Network. The possibilities are endless.
Why are large companies participating in basic research at ELLIS?
European companies have a tremendous interest in developing technologies that will ensure their long-term competitiveness on the global markets. At the same time, the COVID-19 pandemic has quickly increased our dependence on IT solutions, thanks to teleworking and videoconferencing, and it’s apparent that European companies want to build a certain technological superiority in order to make their business models more robust.
Where do you see the greatest market potential for sustainable AI?
Machine learning could play a role in many areas: for example, in understanding the causal relationships of interventions in the Earth’s climate. There’s also a lot of potential for smaller applications, like energy-efficient control strategies in a number of sectors.
What are the greatest challenges for the ELLIS initiative? People have lots of often unconscious prejudices against others. They discriminate based on features like gender, skin color, sexual orientation, and weight. How can we stop AI from inheriting our prejudices?
Algorithms are only as good as the data used to train them. That’s why it’s so important that prejudices that can lead to discrimination be detected and eliminated early on, before the algorithms are used in practice. This is especially important in areas like criminal justice, medicine, and hiring and in other areas where decisions directly affect people. The field of human-centered machine learning has grown rapidly over the past few years, with scientists engaging more and more with the societal challenges, risks, and potential harm of machine learning systems. This also includes thinking about the power an algorithm should or shouldn’t have, with a focus on designing machine learning systems that are inclusive and nondiscriminatory. The ultimate goal is for the technology to support and help people.
Using artificial intelligence means leaving certain decisions to machines, and this will result in errors. Is there a concept for a machine error culture?
There’s broad consensus that machines should be used to support human decision-making and not to replace it. In some cases, machines are less likely to make mistakes than people because of the sheer amount of data they can process or the fact that they don’t get tired. Because the strengths and weaknesses of machines are different from those of people, they also tend to make different mistakes. That’s a good thing, because it means that we can reduce errors by combining the two.
The goal of ELLIS is to strengthen pioneering basic research into AI in Europe. Who are you trying to reach with ELLIS, and what do you offer researchers?
From the very beginning, the focus of ELLIS has been on the idea of a pan-European research network. We wanted to create a research environment that includes the best local talent in Europe and attracts superior researchers from around the world. By setting up 30 research units in 14 countries on the European continent and in Israel, we’ve created a network that combines leading institutions in machine learning and related fields and creates new opportunities for cross-border, interdisciplinary research. Our goal is twofold: We want to offer young researchers an opportunity to be based at one institution while having additional research stays within the network. And we want established scientists throughout Europe to have more opportunities to work together. The success of our PhD program shows that the concept is catching on worldwide. During the first central recruitment round last fall, more than 1,300 young people from over 70 countries applied, and over 60 new students will begin their studies at 51 institutions this fall.
How is collaboration at ELLIS organized, how do the European research teams work together?
ELLIS started as a grassroots initiative by people who’d already been collaborating long before the network was founded and who published many scientific works together. We’ve expanded the potential for collaboration with the addition of 14 research programs and 30 ELLIS units. The researchers now have regular opportunities to meet – for example, at workshops that we periodically organize – or take part in other ELLIS activities that include symposia, debates, brainstorming sessions, and so on. The groups are relatively small, consisting of about 20 people who actively exchange scientific information.
How do you see the current situation with regard to research talent in Europe? What’s still missing, and what’s going well?
Education in Europe is excellent, and we have a strong academic tradition. In terms of talent, I think Europe is on a par with the U.S. What we still need to work on is a culture of entrepreneurship and innovation. As I said before, we don’t just want to educate the most talented, we also want these people to stay in Europe – whether in science, changing over to industry, or founding their own companies. With the Cyber Valley in Baden-Wuerttemberg, we’ve already set up a growing local ecosystem for research and innovation, and ELLIS is a comparable concept on the European level. It’s still under development, but we’re well on our way to creating an environment that includes not only top universities and research institutes but also excellent opportunities for startups. Even German and European politicians have recognized this as strategically important and have promised to increase investments in AI research and innovation.
What do you think is the most important aspect of your role, and what do you particularly want to promote?
I have a shared vision with my colleagues in the ELLIS network: We want to develop interesting locations and ecosystems in Europe that also accommodate the top research laboratories in industry. This will create the conditions for more European AI champions to emerge. Do you remember when employees of IBM Germany famously founded SAP because they wanted to do their own thing? We want to see more of that!
What do you think of the idea of developing AI that can develop new AI?
If you asked ten different experts this same question, you’d probably get ten different answers! I don’t want to put a number on it, but I’m on the more conservative side. I think we’re still very far from creating machines that have the same capabilities as the human brain. When we look at what works well today, in most cases it’s about recognizing patterns in large amounts of data. General intelligence includes much more than that. I don’t expect to see general artificial intelligence in my lifetime; I’m now 53, and I hope to live a lot longer! But I think our attitude toward computers, what we expect from them and our ways of interacting with them, will change quickly in the coming years.
Susanne Gold, November 2021
Subscribe to our Newsletter
Stay up to date at all times: everything you need to know about electrification, automation, and digitalization.