Dovydas Matuliauskas (FWL Lithuania winner), Alessandro Grillini (FWL Groningen winner), and Julius Upmeier zu Belzen (FWL Adlershof winner) will participate in the global Falling Walls Lab Finale on 8 November in Berlin, and talked with us about their field of research and its importance.
Falling Walls: When did you decide to dedicate your research career to artificial intelligence and why?
Dovydas: I became particularly interested in the field of artificial intelligence in medicine when I was a 3rd year medical student (I am currently a 4th year medical student). I read articles about it and later co-founded a start-up named Ligence. Our team is developing machine learning algorithms to help physicians with heart ultrasound imaging. I find the field of artificial intelligence in medicine interesting because it is a recently emerging, still relatively underexplored one, and one which I believe will bring numerous benefits to healthcare in the very near future.
Alessandro: The first exposure I had to artificial intelligence was back in 2008-2009: My graduation project in high school was a support-vector-machine classifier to recognize road signs to enable a speed limiter. I have been coding since I was 14-15 years old and I love everything that revolves around numbers, but my primary research interests turned out to be the neural mechanisms underlying visual perception and eye-movements. I put artificial intelligence on hold for about 7 years until 2016, when I started developing a test to link eye-movement properties to neuro-degenerative diseases. As I was dealing with large amounts of data and complex non-linear models, artificial intelligence seemed to be the obvious answer to my problem. I am not an expert in this field by any means, but I am a passionate learner and I like to keep myself up to date.
Julius: In 2017 I was on Heidelberg’s team for the international genetically engineered machine competition (iGEM) and worked on our software project. We built a neural network that accurately predicts the function of a protein based on its amino acid sequence. Out of curiosity, I started to investigate what the network focuses on and how that relates to important properties in proteins, and I found astonishing connections. Eventually, we showed that one can use these analyses to learn more about where proteins bind other molecules, are catalytically active, and more. This idea, that we can learn something from our learning algorithms, is so fascinating to me that it is what I am working on.