The German Research Center for Artificial Intelligence GmbH (DFKI) has been researching artificial intelligence (AI) for humans for over 30 years and is oriented towards social relevance and scientific excellence. Diplomatisches Magazin spoke to the CEO of the German Research Centre Antonio Krüger, Professor of Computer Science.
DM: Professor Krüger, DFKI does basic research and application- oriented developments in its 25 research departments. Your own research department is concerned with “cognitive assistants.” What do you mean by that, and what does it mean in practice?
Prof. Dr Krüger: If you want to summarize the work in my research department conceptually, it is about the AI companion that supports people in a variety of their life situations. This could be assistance in the emergency room, shopping or climbing, or human-robot collaboration in care or in production – an area, by the way, in which we are cooperating very closely with the Czech Republic. AI should enable adequate human-machine interaction and provide relevant recommendations that correspond to the respective goal, not only qualitatively and conceptually but also multilingually and interculturally. In doing so, AI employs all the advantages of the machine: Speed, fatigue-free, robustness, memory, factual knowledge. For me, the AI Companion is paradigmatic of the role AI can take in our daily lives. However, reasoning, the action-adapted selection of the right action, the final decision remain tasks of humans.
DM: You are at the helm of one of the largest independent AI research institutes. Nevertheless, the question is: Can you keep up given the enormous investments by large companies, especially in the U.S. and China? Does Germany, an industrialized country, play in the premier league worldwide?
Prof. Dr Krüger: The big five U.S. Internet companies are five times as valuable as all the companies in the DAX 40 put together. So if you mean the league of platform successes, then Germany is first class in the enterprise software sector, but not in hyperscalers. Germany also does not benefit from the AI billion market of advertising placement of search engines and social networks. But as an “Industrieland”, Germany is a leader in industrial AI for advanced mechanical engineering, which is about automation through robotization, and in the process industry, which uses AI to combine enterprise software and factory control. When I think of Industry 4.0, a term, by the way, that DFKI helped to coin in 2011, and when I see our work in the area of human-robot collaboration, there is no doubt that Germany is a champion.
DM: Is there a coordinated approach or division of labor at the European level when it comes to AI research?
Prof. Dr Krüger: I would not call the European approach a division of labor at the state level but cooperative and networked in the research area. The EU coordinates and agrees on the framework. But DFKI and, for example, Inria, the French national institute for information technology, are working together very intensively in Trusted AI. This is about comprehensible, explainable, and reliable AI that can guarantee its functionality so that AI systems are certifiable and can really be used in critical decision-making situations. This requires joint European initiatives, which we at DFKI are instrumental in advancing, e.g., through the AI network CLAIRE with more than 430 AI institutions from 37 nations. In our view, the European path of human-centered AI is a globally unique selling proposition on which we should build.
DM: Fifty years ago, Professor Steinbuch formulated that computers could only be “partial intelligence amplifiers.” Will that change? Will AI be able to replicate domains of the human brain such as creativity, empathy, humor, or emotions?
Prof. Dr Krüger: Today, the computer is still qualitatively a partial intelligence amplifier, just as the pliers are a force amplifier or the car is a mobility amplifier. In individual domains, AI now exceeds human capability, and that is what is to be achieved, for a tool is, after all, intended to enhance human capabilities in specific applications to advance human potential. But in human-centered AI, humans set the purposes.
The fact is that the computer is supposed to recognize emotion in human behavior and emulate it in system actions so that human-machine communication can succeed as humans expect. This includes a machine understanding of humor. However, emulated emotions are only simulated in Affective Computing to decode human actions or achieve goals better. But the machine, which has only sensors but no sensory experiences, does not produce them based on passion or suffering. Keyword creativity: AI makes the computer technically creative so that it can produce results, but this does not correspond to the idea of human creativity as an idea-guided process. AI can produce texts and generate content, but the system does not pursue any goals or content plans and has no will. Creativity is also in the eye of the beholder. In this sense, AI systems can already be very creative and used as creative tools.
DM: The EU Commission has made proposals to regulate AI. Critics fear that risks such as autonomous weapons systems or public space surveillance will not be sufficiently contained. Are such and other ethical issues also the subject of research at DFKI?
Prof. Dr Krüger: It is right to regulate facial recognition in public spaces strongly, and it is, of course, important to internationally outlaw the use of AI in the context of autonomous weapons. We have taken a clear position on this. The European Commission invited us to work on ethical guidelines for AI as part of the High-Level Expert Group on Artificial Intelligence. When I look at the AI Act that the European Parliament is currently debating, many of our recommendations or clarifications have been considered. Data protection and privacy are, in principle, prerequisites for the concept of personal freedom, which has been conceived and fought for in Europe for almost 400 years. These are precisely the cornerstones of human-centric AI. We mean AI that cognitively or physically supports humans as citizens of the state and the world in their journey to achieve their own goals.
Interview Matthias Ginsberg