SCIoI: Current DFG Cluster funding concludes – Research moves forward

By: Maria Ott

On 22 May 2025, the German Research Foundation (DFG) announced its decision not to extend funding for Science of Intelligence as part of its Excellence Initiative. While the decision is a temporary setback, it underscores a broader truth: transformative research that transcends disciplinary boundaries often challenges established structures. We still look at the future with optimism and strongly believe that what we have achieved together has laid an important base for intelligence research.

What I’ve come to love about SCIoI, after the years in the communications team, is the science itself and the way our findings come into being. The way an idea moves between disciplines and comes back changed. The beautiful complexity of trying to pin down what intelligence is, knowing that the answer will never come from one field alone. Writing about this kind of work has always been a way of thinking with the science, following its turns, its tensions, its breakthroughs. And, in many ways, reshaping how I understand the world.

If this reads like a love letter, that’s because it is. SCIoI is a rare thing: a truly interdisciplinary effort that treats the science of intelligence not as a competition between fields, but as a common pursuit. And over the past few years, we’ve started to see the promise of that approach come to life.

What we set out to accomplish

We set out to understand intelligence—not just artificial or biological, but the deeper principles that make any system, living or synthetic, truly intelligent. We asked: What do a cockatoo, a robot, and a reasoning human being have in common when they act intelligently? And how can we harness those shared principles to build more thoughtful, adaptable, and ethically aware technologies?

That question became the beating heart of Science of Intelligence (SCIoI)—a research initiative in Berlin, funded by the DFG as a Cluster of Excellence. The cluster quickly evolved into a collective act of curiosity, a space where philosophers met engineers, roboticists learned from behavioral biologists, and psychologists built experiments alongside computer scientists.

Breaking disciplinary boundaries

When we launched SCIoI, the world of intelligence research was scattered. Psychologists were studying reasoning and memory, roboticists were fine-tuning motor control, and computer scientists were training AI systems to beat us at games. Each field was generating insights—but they weren’t speaking to each other. The result was a fragmented view of what intelligence really is.

SCIoI set out to change that. Our core belief is that intelligence isn’t just the sum of its parts—it’s also in the interactions between them. You can’t understand it by dissecting functions like perception, planning, or language in isolation. You have to look at how these functions come together to produce behavior, which is adaptable, goal-directed, and grounded in the complexities of life as it’s actually lived.

In other words, we’re interested in intelligence as it reveals itself through action in the real world. In how systems that sense, decide, and act perceive, respond, cooperate, learn, and adapt in messy, unpredictable environments. That’s where intelligence reveals itself. And that’s where we’ve chosen to look: In the kind of situations where, for instance, a toddler can navigate a living room more robustly than a billion-dollar robot.

This insight isn’t new. It echoes what AI pioneers have long called Moravec’s paradox: that what’s hard for humans (like chess) is often easy for computers, and what’s easy for humans (like walking) remains maddeningly hard for machines.

But instead of treating this paradox as a stumbling block, we made it our compass. We turned our focus to the foundational behaviors that biological intelligence performs with grace: grasping, navigating, interacting, adapting. And we asked how we might build systems that can do the same, not through brute-force calculation, but through embodied, flexible, and context-sensitive behavior.

From understanding to creating

Our method is simple, in theory: understanding intelligence by trying to create it. So instead of asking what intelligence is in the abstract, we ask what it looks like in action. And then we ask how we might build it.

In practice, that means building robots and simulations inspired by real organisms—not just copying their outputs, but learning from their strategies. When biologists uncovered how humans rely on contact interactions to grasp objects, roboticists in SCIoI built machines that did the same. The result was not only a better robotic hand, but also a new hypothesis about how humans manipulate tools, one that could now be tested in biology.

This looping process (back and forth between biology and engineering) became the methodological engine of the cluster. We called it the “synthetic approach” or the “analytic-synthetic loop.” It’s a kind of scientific empathy: building to understand, understanding to build.

Along the way, we confronted the deep complexities of intelligence. Its non-decomposability, its embodiment, its dependence on context and environment. We studied how agents, from insects to humans to collectives, reduce the overwhelming complexity of the real world through smart priors and clever shortcuts. And we began to trace the outlines of something bigger: a set of emerging principles of intelligence.

We found principles to be something more like blueprints: abstract, reusable templates that help us understand how intelligent behavior can arise across different systems, species, and scales. These principles don’t prescribe the fine details of how a mouse navigates, how a fish swarm moves, or how a robot learns, but they capture what these systems might share at a deeper level. Think of them as a periodic table for intelligence research. In a field as complex and wide-ranging as this, they offer something invaluable, a way to focus our questions. A way to see through the noise. A way to move forward.

Each principle distills recurring structure, guiding us toward mechanisms that are plausible, and already supported by comparative evidence across twelve disciplines and a remarkable range of species. Their strength lies in their mechanistic depth: From social insects to human collectives, from neural systems to robotic swarms, the same core structures appear again and again. And these recurring patterns can be used: With our principles we can both explain biological intelligence and construct artificial systems that reflect it. One of them, formalized through our AICON (Active interCONnect) framework, has already earned recognition with a Best Paper Award at ICRA 2025 proving that this is more than theory. It’s a working language for building intelligence.

A place like no other

What we set out to do is not simply to answer questions, but to build a new kind of scientific culture. One that can hold space for uncertainty, for different kinds of expertise, and for thoughtful, in-depth exploration in a world of quick answers.

Throughout the past seven years, we created a place where behavior, not just data or code, becomes the unit of analysis. Where synthetic artifacts like robots, simulations, and computational models become bridges between disciplines. Where we can wrestle not only with what intelligence is, but with what it ought to be.

We haven’t solved intelligence (yet). But we’ve changed how it’s studied. We’ve created a common language where there was once Babel. And we’ve shown that understanding intelligence — real, embodied, situated intelligence — is not just a matter of algorithms or neurons, but of collaboration, curiosity, and openness.

What now?

Our approach deliberately moved beyond conventional academic frameworks. It challenged assumptions. It created friction. It also created something beautiful: a vibrant, cross-disciplinary culture where novel insights could emerge from unexpected encounters.

None of this would have been possible without the people who brought it to life: researchers, staff, and collaborators who poured their energy, dedication, and creativity into this project over the years. The spirit of that work remains alive and visible in everything we’ve achieved: While the funding may have ended, the science and the vision most certainly have not. We believe we’ve created something special, and we are certain that what made it special will endure.

The search continues—across disciplines, across systems, and across minds. We’re not done yet.