Professor Leo Kärkkäinen at the Department of Technical Physics has a long research career in the industry.
“I’ve done all sorts of things in my life. Early in my studies, I was fascinated with the Theory of Everything – in other words, the theory where all the known interactions – electroweak, strong force and gravitation – are quantized. With gravity, this hasn’t been achieved yet,” he says.
According to Kärkkäinen, it was already known that a universe consisting of mere gravitational waves is one with unlimited dimensions and with all locations next to each other. With weaker gravity, the universe turns out to be mostly one-dimensional.
“In Germany, colleagues and I thought that by adding simple matter, the universe would look more like our observations – one time and three directions or something like that, but nothing changed. That kind of put an end to my efforts to solve the biggest question.”
In his doctoral dissertation completed at the University of Helsinki, Kärkkäinen used supercomputers to study how post-Big Bang plasma, consisting of quarks and colour fields (gluons), cooled into protons and neutrons. This line of research took him to Germany, Arizona and Copenhagen.
“Pure understanding of things is fun, but when you have children of your own, you realise that the things you do could also have practical relevance. When the opportunity arose, I returned to Finland and started working at Nokia Research Centre,” Kärkkäinen says.
He spent a good 25 years at Nokia, first developing acoustics simulation tools to enhance the sound quality of phones. That requires not only physics, but also algorithms, which must be embedded in the device and its processors – in other words: signal processing, compression, echo removal and noise control. Collaboration was carried out with, for example, the University of Kuopio, to produce 3D audio. The simulation tools are still being used at Nokia today.
Acoustics has remained a favourite of Kärkkäinen, but the world has posed new challenges as well. The potential of nanotechnology was understood at Nokia, and the company was particularly interested in the use of graphene in electronics.
“These are matters of a longer time scale, but already now, graphene has proven to be quite a miracle material – it’s even superconducting. Back then, we were interested in flexible electronics for flexible devices.”
“Nowadays, intelligence is introduced everywhere with the help of embedded systems. Examples include cars and medical systems – for instance, a blood pressure monitor is no longer a purely mechanical device. Using it as a purely mechanical device requires professional skills, but nowadays, as an embedded system, it is easy to use. However, ease of use is not to be taken for granted.”
User interface design research examines how people react to user interface function
“User interface management must be easy and interactive. Simple things such as these are the most complex to build. Ease of use requires intelligence, and a computer will look intelligent only when it is programmed properly,” Kärkkäinen says.
“As a result, my research and focus shifted to artificial intelligence, segmentation of satellite images and robotics, as well as to medical applications as the leader of the Predictive Health Analytics team, as Nokia sought to expand to new areas of business.”
Kärkkäinen’s work with his team in the medical field continued as a Bell Labs Fellow at Bell Labs, and as a part-time Professor of Practice at Aalto University. At Bell Labs, artificial intelligence was, of course, also used to optimise 5G.
The use of artificial intelligence comes with ethical questions, too
Nowadays, artificial intelligence and machine learning add their own twist to embedded systems. They require a very specific type of computation, new types of microcircuits and efficient processors.
The use of artificial intelligence also involves the need for regulation, which was included in Kärkkäinen’s role with his membership in the European Commission’s High Level Expert Group on AI. The group produced guidelines for building trustworthy artificial intelligence.
“This topic pertains to artificial intelligence that resembles human capabilities. For instance, it is easy to build a self-driving car for empty roads: artificial intelligence sees the road and knows how to drive. Second, the behaviour of a pedestrian entering the road, for example, must be predicted. Using an enormous number of observations, we then need to build an anticipatory system so that the car can plan a reasonable route,” Kärkkäinen explains.
The third stage in the design is ethics. The most difficult question, however, is not whether or not people can be run over, although that is essential.