Eindhoven University of Technology becomes new home for AI research

0

The opening of Neuron, the new home of Eindhoven Artificial Intelligence Systems Institute, is an ideal moment to look back at the flying start of EAISI since 2019 and to consider the institute’s research on AI: its current state of play and its future. We speak to managing director Patricia Jaspers, scientific director Wim Nuijten, and Carlo van de Weijer, the person who carries ultimate responsibility for this endeavor.

The transformation of the Laplace building from a “sorry sight” (Patricia Jaspers’s words when she first saw the dark building two years ago) to a “vibrant, scintillating learning environment” (Carlo van de Weijer’s wish) is at an intermediary stage, and the building is already a “bright, light, spacious building where the clamor of voices, hubbub and peace each have their own place” (Wim Nuijten’s description).

Clearly, all three of them are pleased with how Neuron is taking shape. That EAISI’s own premises have turned out so well, compensates for the patience that was needed in the face of delays in the building work. “It is more attractive than the artist’s impressions provided by the architect. And you can’t often say that,” they say.

TU/e’s going to do AI’
In 2019, EAISI, as the TU/e’s second institute besides ICMS (there are now four institutes), was created to shape cooperation between different departments. The main objective was to set Eindhoven on the map as the place for AI research, and the honor of heading up this endeavor went to Carlo van de Weijer. “We had to let the world know that if you’re a student or researcher interested in AI, you need to be in Eindhoven. In 2019 a great deal of research on artificial intelligence was already being done at TU/e, only back then we didn’t always call it AI.”

EAISI got off to a glittering start. TU/e made a promise to invest twenty million euros each year for the next five years. Van de Weijer: “When we were in the papers every day with: ‘TU/e’s going to do AI’, I decided to put the brakes on. In Brabant we don’t make promises we may not be able to keep. So first we needed to expand our staff, set up degree courses and secure external funding for research. And, please, could we have a building of our own?”

Expanding staff
Then Wim Nuijten was appointed as scientific director. “In 2019 there were 250 assistant profs, associate profs and professors working on AI. We had to find another fifty by 2024,” says the scientific director. “And – with the exception of two positions – we’ve already managed that. The TU/e AI community now comprises nearly nine hundred researchers, including PhD candidates, postdocs and EngD candidates. It was only right, too, that as TU/e we offered AI education. Since then the departments have set up two interdisciplinary master’s programs and we are looking to add a third.”

To be clear: the research is conducted in the departments. The roles fulfiled by EAISI are these: helping to bring the researchers together, writing proposals and securing funding. From the start, EAISI has collaborated with six departments; in recent years three more have also become involved.

Health, mobility, high tech systems
“We do research in three domains,” says Nuijten, “data and algorithms, engineering systems, and humans and ethics. Our strength lies in our ability to combine these fields. The development of a caregiving robot exemplifies this. Much of our research brings together two, often three, of these domains. As for the areas of application, we are focusing on health, mobility and industry. In the first place research must be valuable, but it should also be applicable. For us that means collaboration within the region takes a front seat. Bearing this in mind, EAISI’s natural partners are companies like Philips, NXP and ASML.

“As a rule, the millions we have secured go to PhD candidates, postdocs and EngD candidates,” says Van de Weijer. “We are now attracting between 25 and 30 million euros annually in external funding (from, among others, the Dutch Research Council (NWO, Brussels and industrial partners) to be divided among some thirty projects, on which one or more researchers are working. Take this as an example: TU/e people are collaborating with chip developer NXP to build radar chips that can assess traffic situations. With AI assistance they can tell whether an object is, say, a lamppost or a lanky student.”

Nuijten recalls a recent success: “Working with ICMS and EIRES, we were able to organize a doctoral position for a physics professor. She is researching AI in the context of materials being used in the energy transition. We wasted no time in sitting down to talk because this is something any institute would be interested in, and within a day we’d secured the funding.”

Implementation
After a decision like this, the managing director steps into the picture. “Then it’s over to Patricia to do the difficult work,” says Nuijten lightheartedly, while Van de Weijer hastens to add that whatever arises, she always knows how to manage it.

“She’s not been here long but Patricia’s our leader alright. As the managing director she makes sure that our institute complies with all the rules and does all the necessary reporting, and she doesn’t let the pace flag. It’s thanks to her that EAISI is able to keep up so well in the lightning-fast world of AI.”

Jaspers has been involved with EAISI for two years. Until taking on her current position, she was attached to her alma mater Maastricht University, in roles including that of researcher at the interface of ethics and health research. She brings relevant knowledge and experience to EAISI. The managing director describes her remit as “to make the connection between the institute and the departments and services”.

In practice this means she ensures that the strategic plans for research that Nuijten and Van de Weijer have in mind can actually go ahead. Funding in the right place and staff recruitment, to name just two of her concerns.

Manufacturing
“We are meeting a need. People who have studied AI have the choice of something like twenty jobs.” TU/e is neither the only nor the first university where AI work is being carried out, but every institution has its own individual character.

Van de Weijer: “Amsterdam is very active in the AI field but is concentrating primarily on the financial sector and on language interpretation. We are situated in a region of manufacturing. When machines are involved, the data and situations are entirely different, which means the AI is too. The consequences are also different. If chatGPT makes a mistake, you’re indignant or you laugh about it, but if a car steers left instead of right, it’s not funny. TU/e has traditionally been interested in machines and we’ve made the deliberate choice to take EASI down this route.”

Future
AI is going to impact everything we do, that much is clear. To illustrate how important it is not to lose sight of the ethical aspect, Nuijten gives an example involving caregiving robots. “It’s an engineering system that works with people and for people. In order to improve the models, you’re collecting personal data. Giving a glass of water or pulling on a support stocking are the kind of tasks we would let a robot do. But when it comes to administering medicines or acting in a panic situation, we are more cautious. We don’t yet trust a robot to do these things. We face a huge challenge in deciding how far we want AI to be making its own decisions, where and how in that decision-making process we want to keep the human element, and the pace at which we want to move towards whatever division of roles we eventually choose.”

Sometimes a robot or a machine is required to color outside the lines. Something a human does better. Look at a self-driving car. It performs perfectly well on the highway, a controlled environment. Here it can even communicate with other independent vehicles. This lends itself to coordinated action, such as platooning, where slow-moving traffic drives at a given constant speed behind a lead vehicle. Nuijten can imagine that at some stage in the future it would be wise to let AI take over the job of driving on highways, relieving humans of the task. “But don’t let vehicles loose on a busy four-way intersection without traffic signals in Bangkok, Varanasi or Eindhoven. That’s the kind of place that the rules get broken and we firmly believe that’s something you should leave to people.”

Putting the individual first
“Of course we have to think carefully about the dark side,” says Van de Weijer. “The shifting role people will play is being studied at Innovation Science. In fact, it’s been taken up as an issue across the whole of Europe. It’s important that we add our voice to these others.”

He means that it isn’t wise to leave new issues of morality to be decided by the US and China. “In those regions, values are more closely aligned with industry and the government. Here in Europe we value protecting the individual. That’s not the quickest route, but in our eyes it’s the only one.” He concludes: “We seriously think that the world is better off having Eindhoven participate in AI research. And ultimately, making the world a better place is EAISI’s most important mission.”

Building a position
The six universities affiliated to Eurotech all have an institute similar to EAISI. “We have similar DNA and a good click. We’ve taken on the presidency, together with Lausanne,” says Van de Weijer.

There’s something else, too, that illustrates that Eindhoven is on the AI map. “The potential dangers the development of AI could pose have prompted European politicians to draw up all kinds of new regulations. Some of these are clearly going to block innovation. Let’s not throw the baby out with the bathwater. We’ve got our ears pricked up and two weeks ago we went straight to Brussels to submit an amendment on behalf of all European universities to members of the European Parliament. I’m pretty proud of how we stuck our necks out to save everyone’s bacon. That’s TU/e putting itself on the AI map all over again.”

Back to the old nest
­Neuron has become a fantastic building, all three agree. Van de Weijer: “And to think that Wim and I used to walk over here as students, our floppy disks and punch cards in hand. In those days the building was in its first incarnation, as the computer center. I love that. And the architect has incorporated the punch card motive the awnings. That’s a nice touch.”

Jaspers is also impressed by the charm of the – now – smart building, although she has no emotional connection with punch cards. She started studying health sciences in Maastricht in 2000. “Neuron is a good place to work too. I’m happy to come to Eindhoven every day to be here.”

Anyone who has already taken a look around in Neuron, will have seen the abundance of empty tables and chairs standing ready. Neuron is preparing for the arrival of everyone at TU/e who is working in AI. They can meet each other here. “But …if they all come every day, there’ll be three people sitting on every chair,” predicts Van de Weijer. “We still have to produce a roster for who is welcome when, otherwise the building will be bursting at the seams.”

Neuron also hosts TU/e’s AI student teams. It currently houses HART, CORE, Serpentine and Fruit Punch.