Robot dog learns, adapts like humans, Swedish AI startup IntuiCell says

By Supantha Mukherjee

STOCKHOLM (Reuters) - A Swedish AI startup company has created a robot dog named Luna that has a functional digital nervous system capable of learning and adapting like humans and many animals, the company, IntuiCell, said on Wednesday.

In one of the first-use cases of physical agentic AI, which can make decisions and take actions towards specific goals rather than just perform narrow tasks or generate content, the robot dog would be able to learn like a real dog. 

IntuiCell will hire a dog trainer - instead of integrating a generative AI model and vast datasets - to teach Luna to walk, based on how neurons interact and process information, it said.

"What we have built is the first software that allows any machine to learn like humans and animals do," CEO and co-founder Viktor Luthman said in an interview.

"There's no pre training, no offline simulations and no billion-dollar data centre in the background, but there's a nervous system that allows the machine to learn."    

Indeed, the robot dog's potential would lead to advances in human-like robots' capabilities in different environments, he said.

"The next step is exploring humanoid robotics, exploring autonomous robotics in unpredictable environments such as space exploration, deep sea exploration, or disaster response," Luthman said.

Intelligent machines could be sent to Mars, for example, to build future habitats for people, an environment where the robots cannot be pre-trained and would need to be able to solve problems as they arise, he said.    

While Luna can currently stand up on its feet, the robot dog would perceive, process and improve through direct interactions with the world. 

Formed as a spin-out from Lund University in Sweden in 2020, IntuiCell is funded by Sweden's Navigare Ventures, Norway's SNÖ Ventures and the European Union. 

(Reporting by Supantha Mukherjee in Stockholm; Editing by Bernadette Baum)