Human beings are proof that intelligence is possible, but we don’t have a monopoly. Computers, although not on par with people for some tasks, compute. Even tiny chemical reactions perform computations.

Perhaps that’s where we should be looking for intelligence: embodied in the world around us, says SFI External Professor Jim Crutchfield.

“The paradigm we’ve come to associate with digital computing is very limited,” Crutchfield says, “and there might very well be other kinds of biological, chemical, and physical intelligence that transcend this digital, discrete computation paradigm.”

So he and SFI External Professor Alfred Hübler have set out to define, identify, and measure intelligence in physical systems.

But they have a strict ground rule: “No biology is allowed,” says Crutchfield. “No neurons, no cells, no human being, no subjective criteria like the Turing test.”

Instead, they’re looking at structures like self-repairing dendritic trees, self-replicating RNA, and metal-organic frameworks.

Furthermore, the intelligence they find must be rigorously measurable, he says. “Our physical intelligence model is: ‘We know it when we measure it.’”

Image caption: Ball bearings in oil form a tree-like structure when a current runs through it. Move one of the balls slightly and it will return to its original position. Scramble the whole structure and a new pattern emerges with the same number of endpoints and branching points.

Read the September issue of the SFI Update newsletter here.