Sebastian Risi

Project title

INNATE: Adaptive Machines for Industrial Automation

What is your project about?

Low-cost industrial robot arms working collaboratively with humans for assembly or pick-and place operations are revolutionizing many industries. However, while these robots are already used in warehouses and factory floors, they only work well in structured environments with known objects and struggle to handle products with a high degree of variability. While current AI algorithms can teach machines these skills in the lab, they do not transfer well to the real world because of their inability to deal with unexpected situations. Here we will develop a new class of algorithms based on insights from evolution and learning in biological organisms, which will significantly extend the usefulness of robots in manufacturing and warehouse automation.

How did you become interested in your particular field of research?

I became very interested in the research field of AI when I attended my first class on artificial neural networks (which are loosely inspired by biological brains) as an undergrad student. I found it fascinating that you could teach these systems new skills without having to program them by just providing examples of how they should respond in certain situations. This fascination then led me to pursue a PhD and to study how we can combine different ideas from biology, such as natural selection and neural plasticity, to artificially evolve artificial neural networks instead of having to design their architectures myself.

What are the scientific challenges and perspectives in your project?

Artificial Intelligence (AI) methods are becoming part of our daily lives and can now outperform humans in many domains. However, these systems still pale in comparison to even simple biological intelligence, which can learn and adapt to unforeseen experiences. Current machine learning systems can only deal with situations they have been trained for in advance; they are unable to adapt quickly to unexpected events. A central question in this work is if we can we give machines innate skills inspired by biological organisms that would allow them to very quickly deal with new situations, such as handling never-before-seen objects.

What is your estimate of the impact, which your project may have to society in the long term?

Currently we view machines as entities with fixed behaviors that are designed and programmed to perform a certain job. The potential impact of learning machines could be huge. Instead of programming machines, machines in the future could be taught through example, accumulating an ever-increasing set of skills. These advances will be incredibly useful for many human endeavors, including research itself. Machines that automatically adapt to the needs of their users will fundamentally change our views on machines and revolutionize many industries. Additionally, artificial neural networks have been shown to provide valuable insights into how the brain might process information and this project could shed light on the long-standing question of nature vs. nurture.

Which impact do you expect the Sapere Aude programme will have on your career as a researcher?

The Sapere Aude programme is a very important step forward in my career. It will allow me to consolidate and expand my research group and give me the opportunity to gain further experience in mentoring young scientists. Receiving a Sapere Aude grant gives me the ability to perform high-risk but potentially ground-breaking research, whose outcomes can lead to significant societal benefits and important innovations in AI. It will also allow me to further strengthen my international collaborations with leading researchers in the USA, France and Germany.  

Background and personal life

When I'm not working, which should maybe happen more often, I enjoy playing the guitar or table tennis (if you don't find me in my office you might find me in ITU's bike cellar at the table tennis table).