• Print

Inspiring peopleThomas Gärtner

The best job in the world
Researcher in Data Driven Discovery
Thomas Gärtner
How would you explain your research?

In drug design you might test your molecules for particular properties, and you can do that in the laboratory, but that’s very expensive even with the most modern technologies. So with the machinery you have you test only a few molecules in the lab, and that lets you build a database. Then you can throw your algorithm at your database, and it sorts them and says “If you get to test 50 samples, test these first”. So you can prioritise, you can filter. You can actively pre-sort the data.

When I was in school I was inspired by mathematics, but also interested in becoming a clocksmith
What inspired you to pursue this area?

You know, it’s kind of funny. I think when I was in school I was inspired by mathematics, but also interested in becoming a clocksmith. So I wanted to find something that satisfied both interests, and that was computer science. So I ended up in what was one of the first courses on machine learning or data science that probably existed at the time, and I was really fascinated by it. There, especially, I had a solution and a problem, but they didn’t fit together.

How will your research affect the average person?

Where I hope it will impact is the drug design process. I’ve started collaborating with Jonathan Hirst from Chemistry, and we’re trying to set up something in connection with GSK (GlaxoSmithKline) and other people.

With the Horizon Institute I’m looking into data-driven personalised experiences. In the past I’ve looked into difficulty adjustment in games. For example, in chess you have a multitude of different difficulty settings, but it’s a problem for the player – they have to figure out which is the right one for me. You want to adapt the difficulty of the algorithm to the skill level of the player. We’re now actually looking at designing levels, racetracks, that are interesting for people.

What is the greatest moment of your career so far?

It’s a difficult question because there are so many. It’s like, when you prove something you have different puzzle pieces and at some point they fit together, and you turn around and look at them and you’re like “Oh yeah, that was easy!”. That moment is amazing. You get to meet lots of fantastic people. Brilliant people. All the time, it’s amazing. You can lie at the beach and discuss your research – isn’t that the best job in the world?

What advice would you give someone starting out?

Don’t always jump on the bandwagon. There are advantages to doing that, sure, but if you do it, be careful. Find your own topics that will also be distinct from the mass. Look for the challenges. It’s always the question of what is a hype – it’s very easy to get swept away in some stream of research. With kernel methods, it kind of became a hype after I got into it, but not too much.

You get to meet lots of fantastic people. Brilliant people. All the time, it’s amazing.
What’s the biggest challenge in your field?

I think it’s getting the theory and the algorithms closer together. The theory works nicely, and the applications are really important, and it will work some way, but we don’t understand why that is. Also actually making it more useful for users. A colleague said at some point it’s the “black art” of using machine-learning algorithms. Of course you need to put it in a tool box so it gets used, get away from the technical parameters to more understandable, more intuitive interactions with the algorithms.

Who would you most like to meet in your area?

I would love to meet the great mathematicians of the century, or before. Hilbert and Radon – they’re probably random choices, but lots of work in kernel methods goes back in some way to Hilbert, and Radon because we’ve just started working on some of the methods he introduced. 

John Tukey! That’s not the answer, but he’s closer to the field. He’s among the top candidates.

If you weren’t doing this, what would you be doing?

As I said, I was passionate about clocksmithing. The clocksmith I was working with, he was extremely good, people would sometimes send him antiques from New Zealand, all the way to Germany, a small village there. He had a couple of cars-worth of watches in his safe. Analysing a problem, seeing what doesn’t work and why, and then figuring out how to fix it – fantastic.

If you could jump forward 100 years, what’s the first thing you’d look up?

I’d want to look up the progress on a real strong AI. It’s also a moving target, every time we define a version of what’s AI, we get closer to it, and then we say “oh that’s just state of the art, that’s not AI”. So we redefine it – it would be interesting to find out what the goals are then.

Global Research Theme
Digital Futures

Research Priority Area
Data Driven Discovery

Read Thomas' full profile

Thomas Gärtner is the Professor of Data Science in the School of Computer Science. He is the acting editor of the Machine Learning journal, and is best known for his work on kernels (algorithms) for interpreting structured data.

World-class research at the University of Nottingham

University Park
+44 (0) 115 951 5151

Athena Swan Silver Award