In a breakthrough that could revolutionize robotics, researchers at Northwestern University have developed an artificial intelligence that can design completely novel robots purely from scratch.

Dubbed "instant evolution," the system requires only a simple text prompt to spit out a working robot design in seconds, with no need for the typical trial-and-error tweaking of evolutionary algorithms.

"We told the AI we wanted a robot that could walk across land. Then we simply pressed a button, and presto! It generated a blueprint for a robot in the blink of an eye that looks nothing like any animal that has ever walked the Earth," said lead researcher Dr. Sam Kriegman, an engineer at Northwestern. "I call this process instant evolution."

Starting with nothing but those vague instructions, the AI designed a tiny triskelion-shaped robot riddled with holes and sporting three uneven, finned legs. Though rough-looking, a real-world version built from its blueprint managed to shuffle along at nearly half the pace of the average human stride.

Remarkably, the AI arrived at the tripedal form independently, rediscovering the efficiency of legged locomotion rather than mimicking nature. Beyond that, its creation looks entirely alien compared to any known animal.

According to Kriegman, this demonstrates how artificial intelligence can surpass human biases and limitations to explore structural possibilities that we would never think to try.

The research team believes their approach could be a game-changer for robotics. By freeing autonomous design from the constraints of human engineering and natural evolution, AIs like this one may rapidly find new solutions to daunting mobility challenges.

"When humans design robots, we tend to make them look like familiar things," Kriegman explained. "But AI can create new possibilities and paths forward that humans have never even considered. It could help us think and dream differently, and solve some of our most difficult problems in ways we'd never imagine."

To create the walking bot, Kriegman's group gave their algorithm a simple physics simulator and let it churn through generations of designs in seconds. Each iteration was tested and tweaked until it could stably walk, with the whole process taking just half a minute on an ordinary laptop.

This is a fraction of the time it would take even the fastest evolutionary algorithm running on a supercomputer, showing the vast creative potential unlocked by AI.

Unlike living things, the simulated robots could explore wildly creative body plans free of any legacy anatomy. And indeed, the final design was something no engineer would think to try.

"It's interesting because we didn't tell the AI the robot should have legs," noted Kriegman. "But it rediscovered that legs are a good way to move around on land. Legged locomotion is, in fact, the most efficient form of terrestrial movement."

To validate the AI’s designs, the team 3D-printed physical models of the silicone robots and brought them to bipedal life by pumping air into them. The bots skittered along at a respectable pace, proving the AI’s effectiveness.

However, key details like the holes remain mysterious. "We don't really know what they do, but we know they're important," Kriegman said. "When we take them away, the robot can't walk as well."

This suggests entirely new engineering principles await discovery within the creative space AIs can explore. Kriegman believes today’s result is only a small taste of what instant evolution systems might soon dream up.

The same process could be directed towards designing robots or machines for other environments like swimming, flying or space travel. Beyond mechanics, artificial intelligence may also find fresh solutions to challenges in fields like electronics or chemistry.

"Instant evolution bypasses the slow traffic jams of Darwinian evolution," said Kriegman, "without falling back on human biases." By thinking outside the evolutionary box, AI promises to accelerate innovation beyond what either nature or our imaginations can keep pace with.