“The RAI Institute opens up unique opportunities for both researchers and students”
The Robotics and AI Institute (RAI) in Zurich Oerlikon celebrated the opening of its new centre on 15 September. Marco Hutter, ETH professor of robotic systems and director of the RAI Institute Zurich, sat down with us to talk about the new opportunities that the research centre opens up and how it came about.
Marco Hutter, you’ve reduced your workload as a professor to 50% and are now working 50% as head of the Robotics and AI Institute in Zurich. How did that come about?
I was contacted by Marc Raibert, the founder of Boston Dynamics and father of modern legged locomotion. He was leaving the company and was planning a new research centre for robotics and AI in Boston (the RAI Institute), funded by Hyundai. He wanted me to be involved. I agreed – on the condition that we set something up in Zurich.
Weren’t you tempted to move to Boston?
No, Switzerland is where my heart is. And I think developing something major here in the field of robotics and AI is a huge opportunity for Zurich and for Switzerland. So during my sabbatical two years ago I went to the US and planned the development of the RAI office in Zurich.
Why did Marc Raibert accept the proposal?
At the RAI Institute we want to bring together the best talent from research at the interface between robotics and artificial intelligence, and that is where ETH plays a leading role. Our research group has also made a mark specifically in the field of legged robotics and the application of reinforcement learning for locomotion. Boston has MIT and Zurich has ETH – so it’s ideal for a research institute in robotics to be based at these two top locations.
And what appealed to you about this new job?
The institute opens up new possibilities for our research, but also career opportunities for young researchers at ETH and from all over the world. It combines the advantages of research at a university with those at an industrial company. We can hire experienced researchers on permanent contracts with competitive salaries and therefore carry out complex and long-term research projects. Robotics is famously very hardware-intensive and requires a great deal of engineering. At the RAI Institute we have a large team and the necessary infrastructure.
“The RAI Institute allows us to pool everything on one site, from the idea stage to implementation.”Marco Hutter
How big is the team?
The institute as a whole has over 250 employees globally, in Zurich there’s around 30 of us, and that number will soon rise to 50. The two locations work together closely, particularly in infrastructure. So we have a software group that takes care of the infrastructure for machine learning-based approaches, and a hardware group that builds and maintains the robots. This is also what we do in the lab at ETH, only on a smaller scale. At ETH we can only develop robots up to a certain level of maturity. In the best case we can launch a start-up, like ANYbotics for example, that further develops the robots, which we can later reuse as enhanced systems for research purposes. The RAI Institute allows us to pool everything on one site, from the idea stage to implementation. Thanks to a framework agreement that we’ve signed with ETH, there will be unique opportunities for both researchers and for students.
If researchers at the RAI Institute have more options, aren’t you competing with your own professorship?
Those who want to do a conventional doctorate are best off at ETH. But what comes next? Many of our top doctoral candidates who want to stay in robotics and AI research move to the US. In recent years we’ve successfully countered this by launching start-ups, which are also very attractive in terms of job profiles. But those who want to do research have to pursue an academic career or leave Switzerland. The RAI Institute now offers new opportunities for graduates in Zurich. To me, there’s no competition, rather the two are complementary.
But your capacity as a professor has been reduced by 50%. Doesn’t that have an impact on the number of students you can supervise?
Over the past year not much has changed for ETH – I’ve just worked more [laughs]. My ETH lab is now ten years old and we’ve developed structures that work well. I have a deputy who is doing an outstanding job as a senior researcher. We have some excellent postdocs and a reputation as a group that allows us to attract people of a certain calibre. In this respect, the involvement in the RAI shouldn’t have any adverse effects on ETH. If anything, the project will advance robotics research at ETH. On the one hand, it will offer students internship places and the chance to have their Master’s and doctoral theses supervised by experienced researchers. On the other, the RAI will offer collaboration opportunities, not only for my lab, but also for other professorships at ETH.
In what form?
In addition to the framework agreement with ETH, the RAI Institute has also signed a donation agreement with the ETH Foundation USA. A research framework agreement allows direct collaboration between ETH and the RAI on selected topics, while the donation will also fund new research projects at ETH in the field of robotics. This creates a network that promotes innovation and further bolsters ETH as a leading institution in Europe.
Your dual role as professor and head of the Zurich office of the RAI Institute also raises some governance questions.
We specifically addressed this issue with ETH and the RAI at a very early stage to create the right framework. We also have very good and clear separation in terms of research and premises. Everything that happens at the RAI and that I and the engineers employed in Oerlikon create, belongs to the RAI Institute. On the other hand, there’s the traditional research that takes place at ETH. Everything that I do there belongs to the university.
Is there not a risk that research at ETH is guided by the interests of the RAI Institute?
Not at all, quite the opposite in fact. At the RAI I can build up resources in areas that interest me from a research perspective. What’s more, the cooperation with ETH and the open source contributions from the RAI will allow ETH researchers to take advantage of that, too.
But how is collaboration between your professorship and RAI regulated?
We’ve signed a framework agreement with the RAI Institute that allows us to launch specific projects between ETH and the RAI. This is particularly exciting in terms of my dual role as professor at ETH Zurich and employee at RAI Zurich as we’ve concluded a joint lab agreement in this area. This is less about a physical laboratory, and more about a legal structure that primarily serves to set out clear rules on intellectual property issues. Both parties – ETH Zurich and RAI Zurich – are entitled to publish results and findings from selected joint projects, either together or independently. They also both retain the option of commercially exploiting these results. All the research findings that are generated under this agreement and under my leadership will be jointly held by ETH Zurich and the RAI Zurich. This creates a transparent basis for innovative collaboration, while bolstering ETH’s position as a leading research institution in the field of robotics.
But if you now have to decide which idea to implement where, what criteria play a role?
The question is always in what context an idea originated, and about the right people and the right environment in which to realise a project.
In terms of industrial partnerships in the field of robotics, there is always the question around the military use of research findings. How do you handle this if you carry out research in collaboration with an American company?
The RAI Institute is geared towards scientific research, with a clear focus on civilian progress and innovation. This focus is particularly important as it allows transparent cooperation between ETH and other partners. Particularly at a time when the separation of civilian and military research is a growing topic of debate, we are consciously setting an example of responsible science. So, for example, the RAI Institute also has a dedicated research group that looks into ethics and the societal impact of robotics.
You’ve been working at RAI Zurich for a year now and the offices were inaugurated yesterday. What reactions have you received so far?
The space is great and the researchers seem really happy. There has been a lot of very positive feedback from both ETH and the department on the RAI set-up in general. We have to explore new paths in robotics and AI to satisfy the high level of student interest in this area while maintaining educational excellence. Currently, a small proportion of robotics professors are shouldering a disproportionately high number of Bachelor’s and Master’s theses and semester projects. So if the experienced researchers at RAI are able to provide additional support, that will help.
Has there been a response from outside ETH?
I’m noticing that Zurich as a location is becoming more attractive. All the companies that are locating to the city are creating a positive spiral – and RAI is definitely part of that. This also improves the situation for our start-ups, which have been able to raise venture capital more easily in recent years.