SCIoI presents candidates for universal principles of intelligence

There are wooden cubes on Professor Dr. Oliver Brock‘s desk, sitting comfortably next to plastic molecules from a chemistry kit, coloring pencils, and model figures: a mishmash of objects to fidget with while the brain ponders the nature of thought – and quite a symbolic constellation at that.

Brock is an Alexander von Humboldt professor of robotics at TU Berlin, and one of the initiators of the Science of Intelligence Cluster of Excellence (SCIoI). Launched on the initiative of the applicant universities TU Berlin and HU Berlin, the project has been bringing together more than 130 researchers from over 20 different institutes in Germany and Europe since 2019 to find out how intelligent behavior comes about.

After almost seven years of work, answers have emerged that could advance science and enable new, revolutionary applications in artificial intelligence and robotics.

The researchers have so far identified eight candidates for “principles of intelligence” – characteristics and structures that repeatedly appear when a spark of intelligence flashes in natural or artificial systems (see right-hand page). There is a ninth principle on the horizon, and it is wholly possible that more will be added in the future and those already identified will change. “We now have a blueprint for defining, understanding, and creating intelligence,” Brock says. “And that’s a breakthrough if you ask me.”

A breakthrough in what had escaped researchers until now: Though they could describe what intelligent behavior looks like, they could not explain how it comes about. Brock believes that this was due, at least in part, to the nature of the research: “It was a mistake to believe that we could divide the notion of intelligence into different research areas.” Aspects such as perception, information processing, thinking, acting, and reacting are too closely interwoven to be studied independently of one another.

Twelve disciplines

Hence, SCIoI pools researchers from twelve scientific disciplines, among them computer science, biology, psychology, behavioral science, and philosophy. And this broad spectrum also extends to the project level: while one investigates the ability of schools of fish to react in unison to a threat in the blink of an eye, another is exploring the signals that create trust between people and machines. At other times, it’s about social intelligence in people in virtual worlds, or the ability of cockatoos and robots to open a lockbox – a kind of safe that can only be cracked through observation, trial and error, and a touch of clever thinking.

Several principles of intelligence can be found in the latter, well-known experiment. It involves, among other things, registering signals on multiple channels (hearing, sight, and touch) and processing them at different times in order to respond to the environment. One of the most important findings from the SCIoI project is that intelligence does not necessarily reside in a single organ or element like the brain, but can also emerge from the way components of a system interact. This principle is evident, for example, in the collective intelligence of schools of fish or bee colonies.

“None of these principles should come as a complete surprise to us,” says Brock, “because they represent something inherent in all types of intelligence.” For him, the great value of the SCIoI project lies in understanding which properties intelligent systems should have – because this could be the key to making artificial systems more robust, adaptable, and powerful. “If I were to throw all the principles into a pot and give them a stir, what I get wouldn’t automatically be something intelligent,” says Brock. But he recommends that anyone developing intelligent systems incorporate as many of these principles as possible as it drastically increases the likelihood of success.

Brock can already see how this approach is paying off – for example, when artificial hands learn to grasp objects without conscious control. This falls under the principle that neither signal processing nor responses to it have to take place in the brain. Such insights are expected to open up completely new possibilities in robotics. “For the first time, we can now solve problems that others have previously failed at.”

That alone is a win for science, but the economy could also benefit. Brock is confident that the findings from the SCIoI project are paving the way for Germany and Europe to make significant advances in artificial intelligence. Thanks to the work of the SCIol cluster, there is now “a new opportunity to build intelligent technology” and potentially overtake the USA and China.

The Cluster of Excellence is receiving funding from the German Research Foundation (DFG) until the end of 2027, but the unique spirit of the project will inspire its members far beyond that: “We would all say that the way we do research has completely changed,” Brock says.

Eight candidates for the principles of intelligence

Intelligent systems…

• consist of actively networked components with interconnections that exchange information flexibly.
• solve problems on multiple levels, for example through neural planning in the brain and mechanical solutions in the body.
• act through a dynamic coupling of agent and environment.
• adapt their representations of the world flexibly to the environment, task, and goal.
• integrate information simultaneously via multiple sensory channels.
• process information on several timescales simultaneously in order to adapt to changes in the environment.
• master complex tasks by breaking them down into simpler, low-dimensional problems and combining the solutions step by step.
• adapt their structure to short-term requirements and long-term environmental changes.

The candidates can change, be supplemented, or discarded with new findings.