University of Bremen Prof receives coveted ERC Advanced Grant from the EU

AI achieves impressive results, but the ability to plan actions with foresight is still lacking. Professor Michael Beetz from the Institute for Artificial Intelligence is working on the basics for this. For this he receives the European ERC Advanced Grant, which is endowed with 2.5 million euros.
“I am very happy for Michael Beetz, for his team and for the University of Bremen about this high award,” says University Rector Professor Jutta Günther about the decision of the European Research Council. “This once again confirms the outstanding scientific importance of research into artificial intelligence at our university. I warmly congratulate Michael Beetz on behalf of the entire rectorate,” said Rector Jutta Günther.

Artificial intelligence is not really intelligent yet
In recent years, applications of artificial intelligence have achieved numerous spectacular successes – from beating the best Go players in the world to passing a law exam to creating amazingly good text and pictures. One problem with this: With all these activities, the AI ​​still does not understand what it is doing and cannot explain its procedure. Another problem: AI quickly reaches its limits as soon as concrete action is required. For example, if an AI-based robot has learned to pour a glass of juice from a bottle into a glass, it is – in contrast to humans – not able to spontaneously transfer the ability to pour to other vessels and substances. He has to learn almost completely
“Answering the question of what enables our brain to carry out many different tasks on the basis of vague information and usually successfully on the first try is one of the greatest open scientific challenges,” emphasizes Professor Michael Beetz. “In this project we want to investigate whether the inner imagination of actions and their observation with the “mind’s eye” is an important key to a better understanding of this ability.
Novel technologies used in modern computer games, in combination with methods from artificial intelligence and cognition research, now give us the opportunity to examine this hypothesis in a new and promising way.”

Model for foresighted action
People can often successfully transfer learned skills to new situations on the first try, because they have a number of gifts that allow planning and testing an action in their heads – mostly lightning fast and unconsciously. We understand why we didn’t spill the juice and we have intuitions about the physical properties of the bucket and the batter. We can also visualize in our mind’s eye what happens when we grab hold of the bucket or let go of it at certain points. In order to also enable AI-based robots to use this form of planning, Professor Beetz wants to develop a model ( knowledge representation and reasing framework) that allows them to analyze their actions and their context. This model should also help them to obtain the information they need to complete a task themselves – for example the best way to grab a bucket of water.

To person
Professor Michael Beetz heads the Institute for Artificial Intelligence at the University of Bremen and has also been coordinating the Collaborative Research Center EASE (SFB 1320 Everyday Activity Science and Engineering) there since 2017). The SFB is working on equipping robots with the skills of humans when it comes to coping with everyday tasks. In view of demographic change, this is urgently needed, for example, to enable people with physical disabilities to live independently in their own homes for longer in the future. Michael Beetz first studied computer science at the University of Kaiserslautern before earning his doctorate at Yale University in the USA. In 2019 he received an honorary doctorate from the Swedish University of Örebro for many years of cooperation and outstanding international research. Michael Beetz is also intensively committed to open researcha – making scientific results publicly available in order to promote cooperation.

The ERC Advanced Grant
The European Research Council ( ERC ) awards its Advanced Grant to already recognized top scientists. Groundbreaking basic research will be funded with a maximum of 2.5 million euros for a period of five years. Funding through an ERC Advanced Grant is intended to enable research work that is associated with increased uncertainty, but also with particularly high opportunities for society.