Ute Schmid researches a topic that makes the headlines every day: artificial intelligence. The bidt director devotes a considerable amount of her time to explaining the technology – not only to her students, but also to politics and business, entrepreneurs and employees, teachers and children. Just recently, the scientist, who is also active in many committees and professional societies, was appointed to the AI Council of Bavaria.
What motivates the AI researcher is “the desire that we use AI for the benefit of all people” and “counter ignorance, dispel prejudices, show where the real problems are and what methods are needed to realise human-centred AI”.
Explaining what AI can and cannot do
Often, Ute Schmid has to clear up overly high expectations of the technology. For example, an AI system is only “intelligent” in a very narrow sense. If it has been trained to recognise traffic signs, for example, that is all it can do. “You could almost say AI systems are idiots,” Ute Schmid once said in an interview. One therefore also speaks of “weak AI”. Strong AI, on the other hand, as presented in science fiction portrayals, would have general abilities like humans, its own consciousness and its own will. Ute Schmid considers it rather unlikely that it could ever get that far.
Another typical misunderstanding: “Many people believe that machine learning always means that the systems are constantly learning, just as learning for us humans means lifelong learning. There are such approaches, but most machine learning systems don’t work that way.”
Ute Schmid chooses the example of systems that classify image data, which is often cited in the AI context, to illustrate this. “To do this, image data is first collected and annotated by humans, for example with class labels such as ‘there is a cat in this picture’. The system, which has learned to make decisions on this basis, doesn’t change once it’s been trained – so it can’t detect if the data changes in the meantime.”
Ute Schmid is a professor of applied computer science, especially cognitive systems, at the University of Bamberg. Her research focuses on “human-like” machine learning. She develops methods for learning rules from a few examples and is particularly interested here in relational concepts, comparable to human learning from experience in cognitively demanding areas such as mathematics. She also researches the explanations for why an AI system has decided in a certain way. Thus, her research lies at the interface of AI technology and its application by humans.
AI systems should encourage us and also challenge us. It's not enough for me that humans should only be there to nod off AI decisions.Prof. Dr. Ute Schmid To the profile
Ute Schmid, who also studied psychology, explicitly opposes the idea of fully automated AI in any field. She favours AI that is “interactive and explainable: The moment humans are no longer involved in the application, it is not only the absolute loss of control, but also of competence. AI systems should encourage us and also challenge us. It’s not enough for me that humans should only be there to nod off AI decisions.”
When it comes to explaining AI, she relies on cooperation with other disciplines, especially psychology, but also the humanities and social sciences. “Computer scientists have not learned to look at when an explanation is helpful. But of course it would be naïve to believe that it is enough for a system to provide any explanation and people will then automatically understand the AI system better and trust it – whether justified or not.”
Regardless of the research into such concrete methods for human-centred AI systems of the future, Ute Schmid relies on education and the most well-founded knowledge possible about the technology among the population. If it were up to her, everyone should be able to judge what AI can and cannot do.
In dialogue with the public
Ute Schmid has just been awarded the Rainer Markgraf Prize for her knowledge transfer between research and society. Even if her efforts cost her a lot of time – the scientist feels that the communication work also benefits her research work: “What really pleases me is that I gain insights into many areas of application and that exciting questions for my research also arise from this.”
“Super exciting” is, for example, the project she started at bidt together with computer scientist Professor Alexander Pretschner and law professor Eric Hilgendorf. As part of the research project, the team will develop models to enable learning in the interaction between humans and machines.
It is a pleasure to discuss and work in the interdisciplinary atmosphere at bidt.Prof. Dr. Ute Schmid To the profile
Ute Schmid has been a member of bidt’s Board of Directors since February 2020, and from 15 March 2021 she will take on more responsibility as a member of the bidt Board of Directors’ four-member Executive Committee. “One of my absolute highlights in this AI hype of the past few years is that I was appointed to the bidt Board of Directors. It is a pleasure to discuss and work in this interdisciplinary atmosphere,” says Ute Schmid.
In her view, the complexity and speed of technological development makes direct coordination between the disciplines necessary: “If, for example, social science research on new technologies takes place completely detached from those who develop these technologies, it can easily lead to misconceptions that can often be clarified in direct interdisciplinary exchange, on the other hand.”
Ute Schmid appreciates not only the professional exchange, but also the opportunity to implement the knowledge gained from it in the design of technology and thus also to contribute with her own research to the further development of artificial intelligence for the benefit of society.