Stories#065 Meike’s machine learning

#065 Meike’s machine learning

The story of Alain’s venturous line-up is a story of Meike’s machine learning

Twent Alain Leloux studied technical business administration at UT. Alain likes to pioneer. In 1994, he graduated in Artificial Intelligence. This, unknowingly, made him one of the first experts worldwide. What followed was a long career in IT. As a professional, Alain coached about 200 start-up companies. Nowadays, he invests in high tech startups. Today he interviews Meike Nauta, a role model for women in technology, from Brabant. With her research, Meike brings the human touch and logic back to artificial intelligence. A conversation about stereotypes, innovation, and ambition.

Click for Dutch version

Tuesday 15 February 2022

From assumptions to insight

Alain: 'Hey Meike, from a UT standpoint, you have a remarkable last name, but you’re not related?' 

Meike: 'I get that question a lot, but no, I'm not related to Professor Bram Nauta. I’m from Nuenen, a village near Eindhoven. So I’m from the south. In 2012 I chose to study Business & IT and came to Twente.' 

Alain: 'Why did you choose Twente and not Eindhoven?'

Meike: ‘Good question. And I'm not proud of it, but I initially was quite prejudiced against Twente. Thought it was a bit rustic here. With clogs and tractors, that sort of thing. But when I started looking into bachelor’s degrees in the Netherlands, I found Business & IT the most interesting. So I went to the orientation days in Twente. And believe it or not: the first thing we saw on campus was a tractor! I said to my parents: we’re turning around. Fortunately, we didn’t do that, because the orientation was great. After that day, I knew that this is what I want to do.’ 

Alain: 'Were there any other girls who opted for Business and IT?' 

Meike: ‘It was just me and one other girl, out of the 28 students that enrolled. So we were by far in the minority. It’s becoming more balanced, thankfully. But above all I’d like to tell the girls in high school: don't let that stop you! A technical degree is very broad and especially a lot of fun. So: don't hesitate, choose maths and choose technology. Then the world will be your oyster!’

“A technical degree is very broad and especially a lot of fun. So: don't hesitate, choose maths and choose technology”
Meike Nauta

Alain: 'And did you like the study programme right away?'

Meike: 'Well, I had a hard time with programming in the beginning. It was completely new to me. To think in steps that you want to pass on to the computer. But soon I got the hang of it. Nowadays, I program every day and it has become routine. I’m still fascinated by the combination of IT and the application to improve processes and organizations. I did my master's degree at Computer Science, specialising in data science.’ 

Alain: 'And then you got the hang of it, because you also did your PhD at UT?' 

Meike: ‘That's right! Even though I've always said I didn't want to stay in that academic bubble. I want to make a difference in society, in practice. During my studies I discovered that this too was a stereotypical prejudice of mine. Because as a PhD student you really don't sit in your ivory tower between the books. You can conduct very relevant research that has a real impact.’

“As a PhD student you really don't sit in your ivory tower between the books. You can conduct very relevant research that has a real impact”
Meike Nauta

Alain: 'You’re researching explainable machine learning and artificial intelligence. What can we, as ordinary citizens, expect from this? Will it make the world a better place, or should we be cautious?'

Meike: ‘It's funny that you say that. I’ve noticed that there are often tales in the media about artificial intelligence taking over the world. Well, I work with it on a daily base and can reassure you: these systems can only perform specific, predetermined tasks. And although these systems seem very smart, they can still make stupid mistakes. At the same time, it’s true that chess computers are winning over humans these days. Even from the grandmasters.’

Alan: ‘Really?! I didn’t know that. Back in the ‘90s, we were convinced that humans would always win. But yes, an unimaginable amount of computing power and data has been added in recent decades.' 

Meike: ‘Exactly. But most models need a lot of data to see connections. And because we don't guide people enough in what models must learn, they can also pick up wrong connections. An example. We’re developing a model that can read X-rays and find out whether there’s a bone fracture. That model scored well and was rarely wrong. But guess what? Most pictures are taken in an X-ray room. But some were taken in the emergency department. People lie there on a bed with a protruding metal rim, which was visible in the photo. The model had taught itself: if edge, then fracture. So the results were correct, but the underlying reasoning wasn’t. That’s why explainable machine learning is very important – then you can build in that control.’ 

Alain: 'And where do you see the most possibilities?' 

Meike: 'In hospitals and in healthcare, for example. I’ve been working in Germany for nine months now at the Institute for Artificial Intelligence in Medicine at the University of Essen. There I’m researching how we can use AI in diagnosis and reporting. The models score well. But hospitals and doctors are still reluctant. Explainable AI can hopefully take some of this reluctance away by offering insight into why a machine learning model predicts something. Ultimately, it will probably always be the doctor’s responsibility. AI must become a tool to improve and achieve efficiency gains.’

Alain: ‘If you were given a bag of money for research right now, what would you do?’

Meike: ‘Fun question! With explainable machine learning, we translate the concepts and connections of AI back into concepts and connections that we as humans find logical. I think that this can be done differently by developing new models that reason as we humans would: explainability by design. Then we don't have to decipher what the underlying reasoning of the model is afterwards. Basically, AI with the human touch!'

Alan: ‘Interesting! Tell me, what are your ambitions after your PhD?'

Meike: ‘I’m still orientating. I may continue to work at a university, but I don't rule out a corporate job or a venture into entrepreneurship. There aren’t that many people specialised in explainable machine learning. So we’re looking for ways to safeguard this knowledge.' 

Alain: 'Suppose you go your own way, can UT ask you as an advisor?'

Meike: ‘Definitely! I’ve always enjoyed studying and working here. Since 2012, even as a southerner, I have felt completely at home in the East. So I will always stay connected to UT.’

Alain: ‘Great. What else would you like to pass on to UT?' 

Meike: 'Well, that's something that I also tell myself: to be less modest and show that you are proud of your work. That’s what other universities do. Meanwhile, we’re doing unique things with regard to AI. I wish the rest of the Netherlands and the world would see that too. I’m also aware of my role in the field of artificial intelligence. A public interview like this doesn't really suit me. But I want young women to read this piece and think: hey, I want to do that! And that this way, we can give stereotyping a nudge in the right direction.’

ALAIN LE LOUX MSC (1971)

studied Computer Science and Industrial Engineering & Management at UT. He also obtained a Master of Business Administration (MBA) from Business School Netherlands. He worked for PinkRoccade and was CEO of several start-ups. Now he is mainly active as a start-up investor. For 11 years he was a coach at Novel-T, the UT platform that drives entrepreneurship.

Meike Nauta (1994)

is a PhD candidate in the Data Science group of UT. Her research focuses on explainable artificial intelligence and deep learning. In other words: making artificial decision-making transparent and comprehensible to fathom its correctness, fairness and reliability. In 2018, Meike completed her master's degree in Computer Science at UT cum laude, including a 10 for her prize-winning thesis.