Professor Samuel Kaski, director of the newly launched Ellis Institute Finland and professor at the University of Aalto and University of Manchester, has been awarded a 2.5 million euro senior European Research Council (ERC) grant.
This is the only ERC Advanced Grant awarded to Finland in this cycle, coming as the country strengthens its position as a global AI hub.
Ellis Institute Finland is the latest node of the European Institute for Learning and Intelligent Systems (Ellis) Network, and has all 13 Finland universities supported by EUR 40 million from the Ministry of Education and Culture. It is supported by Peter Sarlin, co-founder and CEO of AMD Silo AI.
Check out our interview with Peter Sarin earlier this year.
Highly competitive funding supports groundbreaking research to transform R&D by integrating advanced machine learning with human expertise.
At the heart of the project is a new machine learning paradigm inspired by the progression of real-world research. This is through an iterative design build test learn (DBTL) cycle.
As shown in the diagram above, the correct loop represents the real-world DBTL cycle that governs experimental science. But in parallel, Kaski's approach incorporates an expert redesign loop. This is an internal loop driven by simulation and human insights.
This structure allows for continuous querying, relearning, and adaptation, and overcomes the major limitations of today's machine learning models. This is that we were unable to generalize the narrow external training domain.
By embedding human knowledge directly into the simulation loop and allowing models to be actively relearn, the project targets important bottlenecks in AI deployments.
result? Not only does it work in the lab, but it also moves into the complexities of real-world R&D – enabling faster discovery, more adaptive systems, and true collaboration between humans and AI.
With Kaski's ERC project and the momentum behind Ellis Finland, the country is now leading the push to bring out top-notch talent and develop Europe's generalizable, collaborative and robust machine learning systems that work in the real world.
Lead image: Professor Samuel Kaski, director of the Ellis Institute in Finland.
