Physicist Giovanni Volpe and his research group are at the forefront of a whole new research field that combines active matter with artificial intelligence.
“The beauty of it all is that we don’t know what we are going to discover,” he says and laughs.
The possibilities seem almost breath-taking. By combining machine learning with active matter, a whole new world opens up.
“This is really fascinating. The two areas are both interesting in themselves, but by bringing them together we will be able to use artificial intelligence to solve physical as well as medical problems,” says Giovanni Volpe, senior lecturer in physics at the University of Gothenburg.
“Our vision for this project is to design smart, artificial particles that are self-propelling, sense their environment and can cooperate in groups,” he says.
So what might these artificial “organisms” be able to do in the future? Well, this project could be about anything from creating robots that can cooperate to clean up a disaster area to designing targeted, microscopic particles that can break down oil spills, decontaminate soil or deliver a drug molecule to just the right cell in the body for example.
“Another aspect is that if we manage to create microscopic systems that can perform intelligent tasks on their own, we will save a lot of energy. Energy consumption is actually the biggest problem and the biggest challenge at the moment; it costs such huge sums of money to run AI algorithms.”
But there is some way to go before all this becomes a reality.
Recently, Giovanni Volpe and his research group were awarded a grant of SEK 37 million from the Knut and Alice Wallenberg Foundation for the project entitled Active matter goes smart.
“It’s great that the Foundation is investing in basic research – generally it’s applied research that is awarded these grants. But basic research is so important. Many inventions would never have seen the light of day without basic research, such as the transistor, solar cells and optical fibres,” he says.
So what Giovanni Volpe and his colleagues will come up with is still only in the realms of possibility. What is clear however is that the research group will be using nature as their inspiration in designing smart particles.
A first step will be to study collective behaviours in bacteria and plankton for example, and to try to understand how their way of swimming in schools – to find food or avoid being eaten – functions. These living organisms can sense their environment and respond to it.
The second step will be to try to copy this collective flock behaviour and transfer it to artificial particles. The challenge is to make the particles sufficiently smart.
“Bacteria are a lot smarter than the particles we can manufacture in the lab today,” says Giovanni Volpe. Artificial particles are generally just a simple mechanism for motility. In our new projects, we now want to make them so smart that they are also able to analyse and navigate their surroundings by sensing its temperature or chemical composition for example.
Combining research in active matter with research in machine learning is totally new. Giovanni Volpe has his own research roots in active matter but two years ago he had his eyes opened to the possibilities of artificial intelligence and began to study the subject. This has meant that he and his research group are at the forefront of this emerging research field. In September 2019, he was involved in arranging the first research conference ever in active matter and artificial intelligence in Lausanne, Switzerland.
“There was a great deal of interest in the conference. Researchers came from all over the world,” says Giovanni Volpe. This research field is growing, and it’s growing rapidly.
He believes that it is important to bring together the authorities in the field to share experiences. But there is also an element of competition and secrecy.
“I expect a bit of competition between us researchers from all over the world. It’s fun and it’s what motivates me.”
About the Active matter goes smart project
Project leader: Giovanni Volpe, senior lecturer in physics at the University of Gothenburg.
Research group: 10 researchers.
The project in brief: To create artificial, “smart” particles that can be controlled and that can analyse their environment, respond to it and cooperate within it.
Collaborators: Bernhard Mehlig and Kristian Gustavsson from the University of Gothenburg, Mikael Käll and Ruggero Verre from Chalmers University of Technology, and Joakim Stenhammar from Lund University.
What’s the latest: Giovanni Volpe will also be arranging a conference on the theme Applications of machine learning on photonics, active matter and neuroscience in San Diego on 23-27 August, 2020.
Senior lecturer in physics at the University of Gothenburg