Human + AI to Reimagine R&D with Smarter, Collaborative Machine Learning
Artificial intelligence has been making waves across sectors, from chatbots to self-driving cars. But scratch beneath the surface, and it becomes clear: much of today’s machine learning still stumbles when faced with real-world unpredictability.
Enter Professor Samuel Kaski, founding director of the new ELLIS Institute Finland, who’s spearheading a ground-breaking ERC-funded project aimed at solving a longstanding ML blind spot.
At the heart of this initiative lies a deceptively simple question: what if machine learning models could work with human experts to adapt and thrive in uncharted territory? It’s this question that Kaski and his team are setting out to answer over the next five years, with the goal of reengineering the design process behind R&D—and reshaping the way we think about AI.
“The basic tenet of machine learning is to apply a model trained on a learning data set. But that only works if the set is representative of the deployment setting—and that seldom holds, because life happens,” explains Kaski.
When an AI system is exposed to data that doesn’t quite resemble what it was trained on—think new variables, changing conditions, or novel goals—it falters. These so-called “out-of-distribution” failures are common, yet under-addressed.
Human Expertise in the Loop
Let’s face it: real-world R&D is messy. Developing something genuinely novel often involves stepping into the unknown, where previous data may offer little guidance. Kaski argues that the way forward isn’t simply gathering more data—which may be time-consuming, expensive, or downright impossible. Instead, he’s placing his bets on human expertise.
“Design processes involve a particularly important ingredient which I believe is necessary for introducing the missing link: domain expertise,” says Kaski. “Including the expert in the loop, while aiming to minimise their workload, may be a way around the out-of-distribution deployment problem.”
This new approach doesn’t seek to replace the scientist or engineer but rather to amplify their capabilities. Imagine a virtual laboratory, powered by simulation and AI, where researchers can iterate ideas, explore variables, and test designs in silico—all while retaining full control over the decision-making process.
“If we can make this loop more effective, with machine learning and a human expert at the core, the cumulative impact could be massive,” Kaski adds.
Building a Smarter Design-Build-Test-Learn Loop
Designing products, building prototypes, testing them out, and learning from the results—the so-called design-build-test-learn loop—has long been the engine behind R&D. But it’s often clunky, slow, and expensive. By integrating intelligent ML systems that actually understand a researcher’s goals and constraints, that loop could become exponentially more powerful.
The trick lies in crafting machine learning tools that don’t just crunch data but understand intent, adapt strategies, and suggest solutions. For that, AI needs to develop what psychologists refer to as a “theory of mind” – an understanding of another’s beliefs, desires, and goals.
In a scientific context, this means grasping a researcher’s tacit intentions, navigating ambiguous problem spaces, and learning from sparse feedback.
“For AI to become an effective team player with humans, it needs to develop what psychologists call theory of mind—in R&D, this would mean an understanding of scientists’ often tacit objectives,” Kaski explains.
Towards True AI4Science
The vision being brought to life through this project is one of AI4Science—where machine learning tools play an active, insightful role in scientific discovery. It’s not about replacing scientists, but about building co-pilots for them. That includes automating the parts of research that are repetitive or computationally intense, and guiding researchers toward the most promising hypotheses.
Kaski envisions AI agents that can reason about and optimise the behaviours of multiple human experts, learning not only from data but also from the collaborative dynamics that emerge in research teams.
“We think we can make AI agents that can reason about and optimise over the behaviours of multiple experts,” he says. “Some of the biggest problems society is facing today require truly interdisciplinary scientific teamwork. If we can engage humans and AI in the problem-solving loop, we can combine expertise, reasoning and optimal decision-making for ground-breaking outcomes.”
This collaborative, feedback-driven approach is a sharp departure from the one-size-fits-all AI solutions that dominate headlines today. Instead, it leans into nuance, ambiguity, and the kind of messy creativity that often defines successful R&D.
A New Hub for Human-Centred AI
This transformative work will anchor the new ELLIS Institute Finland—a bold addition to the European Laboratory for Learning and Intelligent Systems (ELLIS), built on a partnership between 13 Finnish universities. The institute isn’t just another research hub; it’s a playground for collaborative, interdisciplinary machine learning research that aims to move the needle in both theory and application.
The ELLIS network itself, launched in 2018, has been quietly reshaping the European AI landscape, promoting excellence and long-term AI research. With Finland’s addition, the mission expands further into the Nordic region, leveraging its strengths in digital infrastructure, education, and public-private collaboration.
The Finland institute is currently recruiting principal investigators, with an open call to AI and ML talent around the world. The goal is to build an inclusive, cross-domain research ecosystem that brings together academia and industry.
“I look forward to welcoming new colleagues to the ERC project and also to work in and collaborate with the ELLIS Institute,” says Kaski. “We welcome our excellent colleagues and aspiring new AI and ML talent globally to work with us in Finland.”
Changing the Way the World Designs
So, what does this mean for the future of design, engineering, and R&D at large? If the project succeeds, we could see:
-
Shorter R&D cycles with lower costs
-
Higher success rates in product development
-
Better-informed decisions in complex, uncertain environments
-
AI tools that act more like collaborators than calculators
This isn’t just academic theory. The implications span from pharmaceuticals to aerospace, smart cities to sustainable energy. In every corner of science and engineering, the ability to test and iterate with smarter tools could unlock faster paths to innovation.
And there’s a broader message here, too: for AI to be truly impactful, it must work with humans, not just for them. By putting collaboration at the centre, Kaski’s project is flipping the script on machine learning.
A Smarter Tomorrow Starts Here
As industries around the globe race to digitise, automate, and optimise, the need for robust, flexible, and human-aligned machine learning has never been greater. The European Research Council’s backing of this project signals not only confidence in its vision but recognition of its urgency.
It’s not about flashy AI demonstrations or viral gimmicks. It’s about building systems that can deal with real-world complexity, that understand people, and that help solve the grand challenges of our time.
With Finland stepping onto the centre stage of human-centric AI, this project could very well be the start of something transformative. And if Kaski’s vision comes to life, the future of R&D won’t just be smarter—it’ll be profoundly more human.
