By Isabelle Bousquette

CAMBRIDGE, Mass. -- Everyone is obsessed with humanoid robots right now, but the director of MIT's Computer Science and Artificial Intelligence Laboratory thinks tomorrow's intelligent physical machines could be something radically different.

Think soft and squishy robots, says Daniela Rus. Picture flexible robots, or even edible ones.

Her research group has built a robot out of sausage casing (and a small magnet) that could theoretically be eaten and then perform small-scale non-invasive surgeries, Rus said. Another project is a robotic sea turtle named Crush, designed to help monitor sea life, which uses silicone flippers to maneuver around delicate coral reefs.

Rus was a pioneer of this approach, known as " soft robotics." Now creative new uses of artificial intelligence are pushing her work to a new level.

"I really wanted to broaden our view of what a robot is," Rus said. "So if you have a mechanism that's made out of paper and that moves, is that a robot or not? If you have an origami flower that you attach to a motor, is that a robot or not? To me, it's a robot."

A robot rethink from beyond Silicon Valley

There's a growing amount of interest in robotics, which is frequently held up as the next frontier of world-changing technology by tech luminaries such as Nvidia Chief Executive Jensen Huang and Tesla's Elon Musk.

Advances in AI brainpower are helping push robots beyond factories and warehouses to homes, stores and even the local half-marathon. Gartner estimates that by 2030, 80% of humans will engage with AI-powered autonomous robots on a daily basis, up from less than 10% today.

Much of the interest currently centers around humanoids, or robots that have two arms, two legs and a head. Rus's research group is working on these too, developing robots that can pour and stir materials in the kitchen. But she also said that for robots to fulfill their full promise, they need to go beyond the human form.

The Romanian-born Rus has been reimagining those boundaries ever since completing her Ph.D. at Cornell in 1993. Now she's a major star in the field, said Michael Peter Kennedy, chair of the IEEE Awards Board. Earlier this year the engineering professional organization awarded Rus the Edison Medal. Alexander Graham Bell and Nikola Tesla were previous recipients.

"There's literally nobody in the world that knows more about this stuff than Daniela Rus," said Steve Crowe, chair of the Robotics Summit and Expo, when he introduced her for a keynote earlier this month.

Rus today works from the multicolored halls of the Computer Science and AI Lab, or CSAIL, where she was appointed director in 2012. The lab, a powerhouse of AI talent since the 1960s, now houses over 1,800 students, faculty, researchers and staff, a number of interactive AI-powered art installations and one police car hacked by MIT pranksters in the '90s.

Swim like a turtle, think like a worm

But research isn't limited to the lab.

On a sunny Thursday, the lifeguard at the MIT pool is nonchalant as CSAIL students Emily Sologuren and VeeVee Cai lower their robotic sea turtle into the water. Apparently, it happens all the time.

Crush, much like his namesake from millennial classic "Finding Nemo," was designed to swim in the ocean, helping monitor coral reefs and sea life with his camera-eyes. But he's not quite there yet. The turtle represents both the opportunities and pitfalls of a so-called soft robot.

For one, finding the right balance of hard and soft material is difficult. Too soft, and Crush could get carried away by the current; too hard, and he might damage the reefs. It's also tough to keep his non-waterproof electronic components from getting waterlogged, said Cai.

"It's very hard to put things underwater," he added.

That's especially true when it comes to large, expensive robots that, when deployed, could risk damaging themselves or the surrounding environments, said student Pascal Spino.

That's why Spino developed what are known as "bubble" robots. The small spheres, which include sensing equipment tucked inside a 3-D printed shell, can each be made for less than $200, he said. It's a critical advantage when it comes to exploring tight or delicate environments, like caves or shipwrecks.

The spheres use Lidar, the same technology that helps autonomous vehicles detect their surroundings, to navigate, he said. And four thrusters help it push itself around the water. "That's a self-driving car in the water," Rus said of the project.

In the future, there may be better shapes for this than just a sphere, she added.

"We have actually built an eel also, and we're experimenting."

Building robot brains is a challenge all its own. Generative AI has provided a lift here -- for example, helping the bots intuit objects they might not have been trained for. But these new models also bring some new problems.

Often they'll require a separate physical computer system. And they can make mistakes when it comes to perceiving the physical world, which -- depending on the robot they control -- may have harmful consequences.

Rus aims to solve both those problems with a new architecture known as liquid networks, systems modeled on the neural activity of worms known scientifically as C-elegans.

The resulting algorithms are compact enough to run directly on robots, or even smartphones, yet intuitive enough to interpret and adapt to complex physical environments. They can be trained on hundreds of GPUs, rather than tens of thousands, she said.

Rus and three other MIT researchers last year spun out a company called Liquid AI to apply the technology to real-world applications like self-driving cars.

Prompt your robot

Despite these advancements, building and designing effective robots is still a very slow, iterative process, Rus said. But that might be another area where AI can help.

Rus and her team have designed a special AI system trained on the laws of physics that suggests robot designs at a prompt. It's called "text to robot."

"So you can start with very simple language prompts such as 'make me a robot that can walk,' or 'make me a robot that can operate a drill,' or 'make me a robot that can make lemonade,'" Rus said.

In the robotics lab, AI helped design a three-fingered robotic hand that can operate a syringe. Rus said she can imagine a hospital having a base robotic arm with different attachments, each optimized for various medical tools.

With these new design methods, the possibilities for future robots are essentially limitless. Rus said sometimes it feels like there aren't enough hours in the day, or even years in a lifetime, to build them all.

So what does she really wish she'd focused on? Something that would give her more time to pursue all her ideas: Technology that can reverse the biological mechanisms of aging.

"There is an AI solution to that," she said.

Write to Isabelle Bousquette at isabelle.bousquette@wsj.com

(END) Dow Jones Newswires

May 16, 2025 07:00 ET (11:00 GMT)

Copyright (c) 2025 Dow Jones & Company, Inc.