Technical University of Denmark: DTU continues mapping of AI’s working method
While the use of artificial intelligence is gaining ground everywhere, we still don’t know exactly how artificial intelligence works. But Professor Søren Hauberg from the research section Cognitive Systems at DTU Compute is on to it. And with a grant of almost DKK 4 million, he and his team can continue their work for the next three years.
Together with three other researchers, Søren Hauberg has received a so-called additional grant for the Villum Young Investigator (YIP) programme.
The VILLUM FOUNDATION writes in its explanatory statement that the four researchers – three of whom come from DTU – ‘have already proven their ability and determination as research leaders as they have established careers at a top international level. We hope that YIP+ will provide them with the stability they need to reach the next level’.
Søren Hauberg is a happy professor:
“It is so amazing to receive this grant from the VILLUM FOUNDATION. Firstly, because the committee, which is made up of some of the country’s most talented researchers, has looked at my work and said, ‘it’s going pretty well’. One can get a bit moved by that. And secondly, because it allows me to keep the going full steam ahead on a project that is wildly exciting.”
“The project is therefore very much basic research, and it is simply fantastic that we have private foundations that see value in supporting this kind of work.”
Søren Hauberg, Professor, DTU Compute
Basic research is the prerequisite
In the research project ‘Learning from learned representations’, Søren Hauberg develops geometric tools to better understand artificial intelligence:
“In all its simplicity, the project is about putting units on the intermediate calculations that are made in artificial intelligence. It is in many ways quite basic. When we look behind the scenes at an artificially intelligent system, the analyzed data is transformed into a series of numbers, which is immediately difficult for us humans to make heads and tails of. This makes it difficult to understand what the system is actually doing. This project builds tools to put units on the numbers found behind the scenes. This is based on the logic that it is, after all, easier to understand ‘100 cm’, ‘100 kr.’, etc., rather than just ‘100’.”
“When we are then able to put units on the intermediate calculations carried out by the artificial intelligence, we won’t suddenly be able understand everything it does. But I would venture to say that it is a prerequisite for us to ever come to understand it. The project is therefore very much basic research, and it is simply fantastic that we have private foundations that see value in supporting this kind of work,” says Søren Hauberg.
Important to look into the brain of AI
AI is gaining greater ground in society day by day, and therefore as a society you also need to know what is going on, Søren Hauberg believes.
“We have to deal with the obvious fact that artificial intelligence is going to play a bigger and bigger role in society. Both in systems we citizens directly interact with, but also in decision-making processes that take place without us necessarily being aware of it. This creates amazing opportunities, but also quite obvious challenges, as we often do not understand how the artificial intelligence reasons. My work is part of a long tough move towards creating understanding. This will create an elemental foundation for us to one day open that understanding.”
The grant from the VILLUM FOUNDATION makes it possible for Søren to employ a PhD student and a postdoc.