I’ve spent much of the past year discussing the creation of artificial intelligence and large language models with robotics experts. It’s becoming increasingly clear that these kinds of technologies are destined to revolutionize the way robots communicate, learn, look, and program.
Therefore, a number of leading universities, research labs and companies are exploring the best methods to leverage these AI platforms. Well-funded Oregon-based startup Agility has been playing with technology for a while now using its bipedal robot, Digit.
Today, the company showcases some of these projects in a short video shared through its social channels.
“[W]We were curious to see what could be achieved by integrating this technology into Digit,” notes the company. “A physical embodiment of artificial intelligence created a demonstration space with a series of numbered towers of several heights, as well as three boxes with multiple defining characteristics. Digit received information about this environment, but received no specific information about its tasks, just natural language commands of varying complexity to see if it could execute them.”
In the video example, Digit is asked to pick up a box the color of “Darth Vader’s lightsaber” and carry it to the tallest tower. The process isn’t instantaneous, but rather slow and deliberate, as you’d expect from an early-stage demo. The robot, however, performs the task as described.
Agility notes, “Our innovation team developed this interactive demo to show how LLMs could make our robots more agile and faster to develop. The demo enables people to speak to Digit in natural language and ask it to do tasks, giving a glimpse into the future.”
Want the top robotics news in your inbox every week? Sign up for Actuator here.
Natural language communication was a key potential application of this technology, along with the ability to program systems through low-code and no-code technologies.
During my panel at Disrupt, Gill Pratt described how the Toyota Research Institute is using genetic AI to accelerate robotic learning:
We have figured out how to do something, which is to use modern genetic artificial intelligence techniques that allow human demonstration of both the position and the power to effectively teach a robot from just a few examples. The code does not change at all. What it is based on is something called diffusion policy. It’s work we did in collaboration with Columbia and MIT. We have taught 60 different skills so far.
Daniela Rus of MIT CSAIL also told me recently, “It turns out that genetic AI can be powerful enough to solve even motion planning problems. You can get much faster solutions and much more fluid and human solutions to review than with predictive model solutions. I think it’s very powerful, because the robots of the future will be much less robotic. They will be much more fluid and human in their movements.”
The potential applications here are wide and exciting — and Digit, as an advanced commercially available robotic system piloted in Amazon fulfillment centers and other real-world locations, appears to be a prime candidate. If robotics are going to work alongside humans, they will have to learn to listen to them as well.
