Computational science – expanding the range of thinking.

What connects Aristotle to modern computing, why the digital age soon comes to an end and what will follow next. A visit to Petros Koumoutsakos. 

We meet Petros Koumoutsakos in his office on the top floor under the roof of an old former residential building. The small flats of this five-story house were turned into offices and seminar rooms to accommodate the Chair of Computational Science of ETH Zurich. Climbing up the steep stairways, we pass colourful pictures painted on the walls. Portraits of Aristotle, Newton and Bayes kindly smile at us on the third floor and a pod of dolphins welcomes us on the fourth. When we finally reach the fifth floor, slightly breathless, we first want to know more about these paintings. After a warm welcome, Petros Koumoutsakos tells us that these pictures have been painted by a former «artist in residence» in his lab, a young graffiti artist originally educated in Byzantine painting at the School of Fine Arts at the National Technical University of Athens. While we sit down at the table, he explains what these pictures have to do with computational science: «I would call computing an augmentation of the human brain that extends the range of our thinking. But the human brain is always at the centre. Modern computational science is rooted in a long tradition of human thinking, based on the findings and conclusions of Aristotle, Newton and Bayes. That is why I have their portraits painted on the wall.» And with a smile he continues: «The four causes (material, formal, efficient and final cause) of Aristotle are at the core of the definition of multiscale modelling. 

«Computational science is a domain of knowledge,
aiming at understanding nature
and solving complex problems
with the help of computers and mathematics.»

But Aristotle just provided a framework for thinking. Isaac Newton was the first who put down principles in mathematics quantifying natural phenomena by physical principles and mathematical formula. However, today we find that Newton’s deterministic approach does not take into account uncertainties and the inherent stochasticity of physical systems. So here, the English mathematician and Presbyterian minister Thomas Bayes comes in. He stated that every model has its parameters depending on data. Today, with the massive availability of data and supercomputers, we revisit physical laws. Data-driven, Bayesian uncertainty quantification is a key research field in our lab. Bayes’ theory requires the computation of integrals in high dimensional spaces which we can accomplish thanks to the available efficient machines.»

A powerful tool 

Computational science, the fusion of models, algorithms, data and high-performance computer, has opened completely new horizons in tackling scientific and even societal problems. It allows us to solve complex problems in relatively little time and to perform «what if?» studies and optimisation. What Petros Koumoutsakos fascinates most about computational science is the translation of highly difficult problems into algorithms enabling the computers to output facts-based reliable answers. The analysis of the models, simulations, data and visualisations help scientists to understand physical phenomena or suggest feasible solutions to engineers. A key topic he and his group are focusing on are the problems of fluid mechanics of collective behaviour. 

«Modern computational science is rooted
in a long tradition of human thinking
based on the findings
of Aristotle, Newton and Bayes.»

Fluid mechanics are fundamental to collective behaviour in nature and technology, as fluids pervade complex systems at every scale, ranging from schools of fish and flocking birds to bacterial colonies and nanoparticles for drug delivery. Little is known about the role of fluid mechanics in such applications. Is schooling the result of vortex dynamics synthesised by individual fish wakes or the result of behavioural traits? Is fish schooling energetically favourable? How does blood affect the collective transport of nanoparticles in cancer therapy? To all these crucial questions Petros Koumoutsakos and his team want to find the answers through computational methods that resolve the interaction of fluids and multiple, deforming bodies across several scales from macro to nano. 

Understanding fluid mechanics

Therefore, in 2013, Petros Koumoutsakos applied for an ERC Grant on «Fluid Mechanics in Collective Behaviour: Multiscale Modelling and Applications» which he received in 2014. Since the project has started, he and his group have achieved quite a number of results. Based on data and existing equations, they have developed the first ever physically accurate simulation of fishes swimming together. «These simulations allow us to understand the fluid forces that fish must overcome in order to swim the way they do and how they are sensing and reacting to their environment,» he tells us while showing the impressive flow visualisation on a large screen. This simulation is much more than a highly sophisticated software based on computational science methods. It has a direct impact on applications. «What we learn and abstract from the fish swimming could be used to put together wind turbines in a windfarm or fly drones in an optimal swarm formation, which could drastically reduce the consumption of energy and improve efficiency,» Petros Koumoutsakos explains. To develop a computer model to understand how nanoparticles flow through blood vessels is another topic he is focusing on within his ERC project. Cancer might be treated by nanoparticles transported through the body of a patient and delivered precisely into the tumour. But there are quite a couple of questions still unanswered. Blood cells have a soft structure, while nanoparticles are quite rigid. How will they interact with each other? Similar questions arise for the detection of other cells, such as circulating tumour cells in the blood stream. There is one circulating tumour cell per billion of red blood. How can we detect it? 

«What we learn from the fish swimming
could be used to put together wind turbines
in a windfarm or fly drones
in an optimal swarm formation.»

Hence, similar to the school of fish, nanoparticles and tumour cells in a blood stream are collective fluid mechanics phenomena, which Petros Koumoutsakos and his group want to understand and create reliable computer simulations thereof that will help their prediction. 

On the eve of a new age of computing

But computational science is more than solving problems by computing and computers. It also deals with key problems of computers themselves. During the last 50 years, according to Moore’s law, the number of transistors in a dense integrated circuit has doubled every two years. But this expansion has come to an end due to physical limits. You simply cannot pack additional numbers of processors onto the same area. Moreover, the speed of the machines cannot be exceeded any further as there will not be sufficient energy in the near future to run the expanding numbers of computers. They also may effect enormous environmental problems, as cooling the big machines pollutes the environment even more than air traffic. It seems that we soon reach the end of the digital age of computing as we know it. So, what will be the next technology, we ask Petros Koumoutsakos. He names three options to replace digital computing by 2040: Quantum computing, neuromorphic computing (processing information similar to the human brain) and reducing the accuracy of the existing digital computing. For example, instead of up to 25 digits, computing could be restricted to two digits to save energy. Today, scientists and engineers all over the world are thinking of ways to reinvent computing, searching for new methods and systems – and Petros Koumoutsakos is one of them. Together with 50 colleagues from ETH Zurich, the EPFL (École polytechnique fédérale de Lausanne), the University of Zurich and the USI (Università della Svizzera italiana), he is currently preparing a project proposal for the NCCR on emerging and sustainable computing. 

«Cooling the big computers
pollutes the environment even more
than air traffic.»

«Future computing algorithms will merge the classical way of deterministic physical principles and new ways of integrating data in our predictions and thinking processes, bringing Bayes and Newton together. This will have another gigantic impact on our society. Fusing these ideas with new computing architectures and hopefully making a contribution to some societally relevant applications – this is how I see my work for the next ten years,» Petros Koumoutsakos tells us at the end of our visit. 

«You start by thinking, you end by thinking
and the computer is just an augmentation
of our human capabilities.»

On the stairways down from the fifth floor, passing the portraits of Aristotle, Newton and Bayes that continue to smile at us from the walls, we remember what Petros Koumoutsakos told us before we left his office: «Even in the new computer age you start by thinking, you end by thinking and the computer is just an augmentation of our human capabilities.»

Interview with Petros Koumoutsakos