Professor Raúl Rojas still remembers clearly how he was telephoned by a member of his university’s selection committee when more than 30 years ago he had applied for a scholarship in Europe. A mathematician from Mexico, he wanted to pursue research in the field of artificial intelligence, and was planning to go to Germany on a DAAD grant. "My professor couldn’t understand why I wanted to go to Germany. Why not Chicago? Or the MIT? That’s where things are really happening, he said." But Rojas stuck with his decision. "I simply found Germany more exciting, especially on account of its scientific tradition." These days DAAD alumnus Rojas teaches computer science at the Freie Universität (FU) Berlin and runs the Dahlem Center for Machine Learning and Robotics (DCMLR), one of the most important centres for AI research in Germany. He never regretted his decision: "When it comes to innovation in AI, Germany is among the elite." In his view, the objection that Europe has fallen behind in terms of machine learning is not particularly constructive. "AI is not simply machine learning; it is also the attempt to build rule-based systems that can argue logically. And in this area, Germany is still a world leader." Robotics is one focus area at the AI centre in Dahlem. Roja’s team made a name for themselves with smart football robots, and since 2006 they have been concentrating on research into driverless vehicles. For eight years now, software developed at the FU has been guiding self-driving cars through Berlin’s traffic. Another area on which the four professors at the DCMLR focus is the development of biomimetic robots that imitate animal behaviour. "We want to develop systems that function well, whose actions remain understandable and which conform to our ethical values", is how Rojas summarises the central research approach.
AI made in Germany
In putting it like this, the AI researcher from Berlin gives a pretty good description of the aims that the country’s Federal Government has set down in its national AI strategy. The goal, according to the strategy, is for Germany to develop its own approach: "AI made in Germany." And this involves a great deal more than simply dedicating the Science Year 2019 to AI. The Federal Ministry of Education and Research (BMBF) is making three billion euros available for AI research by 2025. The idea is to embark on a path that will see AI used in the right way. And this requires research: "Research not only drives innovations forward – from their early stages to market readiness. It also helps us to better understand technical developments from an ethical, social and legal perspective", explains the BMBF. Furthermore, 100 new professorships are to be created to ensure that the next generation of AI specialists can be properly trained. There are plans to set up a German AI observatory to monitor the impact on the employment market and not to lose sight of ethical aspects. At the European level, Germany is hoping to launch a cooperative venture with France that would involve establishing a "virtual centre". A research consortium is also planned – a network of technology-, domain- and application-oriented institutions. In addition, the BMBF intends to provide further support to existing AI centres of excellence and to interlink them to form a wider network.
Visits to the DFKI and Cyber Valley
One of the best known of these centres is the German Research Center for Artificial Research (DFKI); established in 1988, it now employs around 1,100 people, making it the world’s largest AI research institute. The DFKI has bases in Kaiserslautern, Saarbrücken and Bremen, a project office in Berlin, a laboratory in Lower Saxony and a branch in the Saarland. Researchers from 60 nations work on basic research and product development there. The DFKI conducts research into virtually every conceivable field of application in AI, including agents and simulated reality, Industry 4.0, machine translation and smart assistance systems. Thanks to its excellent international reputation – more than 140 AI professorships around the world are held by DFKI graduates – the institute attracts many talented researchers from abroad. One of them is Ahmad Kadi. After acquiring a BSc in computer science in Syria, he came to Germany in 2015 on a DAAD scholarship and did his MSc at the Technische Universität Kaiserslautern. "The conditions were ideal for me – the high quality of the education on offer coupled with the low tuition fees. " Kadi began working at the DFKI while still a student, and is now employed there as an engineer. He is working on an app that is designed to help farmers draw up their harvesting plans and work out when to use pesticides. It uses AI technology for conducting semantic searches. "We want the machine to understand the questions in just the same way as a human being would."
The Cyber Valley research alliance in the state of Baden-Württemberg is proof that Germany has no need to hide its light under a bushel when it comes to machine learning, either. Since 2016, a whole host of partners have been collaborating there: the Max Planck Institute for Intelligent Systems, the universities of Stuttgart and Tübingen, and the companies Amazon, BMW AG, Daimler AG, IAV GmbH, Porsche AG, Robert Bosch GmbH and ZF Friedrichshafen AG. Their objective is to create the ideal conditions for transferring technology from research to practice. Dr Tian Qiu heads one of ten research groups in Cyber Valley. His "Biomedical Microsystems" group, which was established at the University of Stuttgart in July 2019, researches sensors and the control of microrobots in medical applications. "Our work focuses on developing autonomous systems that can find their way through tissue on their own." So as to be able to feed the robots the medical data they need – data that is not available for legal reasons, however – the team works with an augmented reality environment of the kind used to train surgeons. It was also the centre’s high level of expertise in robotics that attracted Dr Charlotte Le Mouel from France to Cyber Valley, and specifically to the Max Planck Institute for Intelligent Systems. Holding a PhD in neuroscience she came to Tübingen as a postdoctoral researcher on a DAAD scholarship. Le Mouel wants to find out exactly how humans learn movements, and if this mechanism could also be used to program robots. "There is a research group at the Max Planck Institute that is using a very innovative approach to explore this", she explains. "What we are doing is not conventional machine learning in the sense of data calculations. We are trying to understand how intelligence is embodied."
The future of work, health and mobility – these are the specialist AI areas at the Munich School of Robotics and Machine Intelligence (MSRM), another important hot spot for AI research. Since it was founded in 2018, the MSRM has brought more than 50 professors from the Technical University of Munich (TUM) together in interdisciplinary teams. One focus is on human-machine interaction. "We are not interested in making machines more human", explains MSRM Director Professor Sami Haddadin. "The idea is rather that robots should serve as intelligent tools for humans. Collective AI allows them to adapt quickly to new tasks. " This is also the subject that DAAD alumnus Fernando Díaz Ledezma is researching for his doctoral thesis at the MSRM. "I want to explore possible ways of making robots even more adaptive in collaborative environments", explains Ledezma, a research associate at the Munich School of Robotics and Machine Intelligence at the TUM. “Above all, by teaching them to cope better and better on their own." What he liked about Munich was the way research here is conducted in close cooperation with industry. "This is ideal for me, as I can well imagine working in the R&D department of a large company in the future."
Supercomputers from Germany
Germany also boasts first-class research institutes in big data analytics – a field in which the USA and China are the world leaders. One is the Competence Center for Scalable Data Services and Solutions Dresden/Leipzig (ScaDS), where new infrastructure architectures are developed to process data more efficiently. "The current challenge is scalability", says Professor Wolfgang E. Nagel, director of the Centre for Information Services and High Performance Computing (ZIH) at Technische Universität Dresden and scientific coordinator at ScaDS Dresden/Leipzig. "How can we build systems that can operate with hundreds of processors rather than with just a few? This is what we are working on." And with considerable success: "In the spring, we were able to install a system with an input/output pipeline that has the same processing power as the two largest supercomputers in the USA", says Nagel. The challenge in his eyes is how to manage and make available the huge volumes of data for research. "Currently researchers come to us with their own small data sets and have to top them up with data from external providers so as to arrive at good results. "The tricky part is that the data then often become the property of the provider. "We are working on a national infrastructure for research data so as to avoid this happening in future. This should mean that we can obtain good results that allow us to make data available – in line with accepted access regulations – and to enable them to be analysed on a cross-disciplinary basis with only moderate effort." Meanwhile, progress is definitely being made at the political level: a cloud infrastructure named Gaia-X is to be established that will allow EU countries to network with one another. This is intended to form the basis for an open digital ecosystem that will keep Europe globally competitive even into the future. This is certainly good news for Germany as a centre for AI.
Author: Klaus Lüber
For more information about AI research in Germany take a look at our newsletter edition of February 2019.
This article was originally published in LETTER, the magazine for DAAD Alumni, issue 03/19.