Boston Dynamics and the Toyota Research Institute (TRI) on Wednesday announced plans to bring AI-based robotic intelligence to the electric Atlas humanoid robot. The companies will tap into the work done to date at TRI around large behavior models (LBMs), which work in a manner not very different from the better known large language models (LLMs) underpinning platforms such as ChatGPT.

TechCrunch visited the Bay Area campus of TRI in September to take a closer look at the work of the institute on robot learning. In a research revealed at last year's Disrupt conference, the institute head, Gill Pratt, explained how the lab has been able to get robots to 90 percent accuracy in performing household tasks, such as flipping pancakes, through overnight training.

"In machine learning, up until quite recently there was a tradeoff, where it works, but you need millions of training cases," Pratt explained at the time. "When you're doing physical things, you don't have time for that many, and the machine will break down before you get to 10,000. Now it seems that we need dozens. The reason for the dozens is that we need to have some diversity in the training cases. But sometimes, it's less.

In hardware, Boston Dynamics would be a good match for TRI. Spot_maker has already done its share on the software and AI side to power its own systems, but the manner of work required to teach robots to do complex tasks with full autonomy is another beast altogether.

"There's never been a more exciting time for the robotics industry, and we look forward to working with TRI to accelerate the development of general-purpose humanoids," writes Boston Dynamics CEO Robert Playter in a statement. "This partnership is an example of two companies with a strong research-and-development foundation coming together to work on many complex challenges and build useful robots that solve real-world problems."

Boston Dynamics unveiled its design for the electric Atlas in April, as it finally put to rest the humanoid's larger, hydraulic namesake. While we've seen very little of the robot since then, in August, TechCrunch managed to get its hands on a short video of the robot doing pushups. Like Atlas' first video, the quick pushup demo was a good demonstration of the robot's remarkable strength.

Boston Dynamics' closest competitors in the space of humanoid robots — Agility, Figure, and Tesla — have all largely chosen to build out their AI teams in house. The Boston Dynamics-TRI deal is of particular interest because the companies involved, Hyundai and Toyota, are direct competitors in the automotive space.

Meanwhile, Boston Dynamics has its internal spinoff research outfit, The AI Institute, previously known as The Boston Dynamics AI Institute. Although it's headed by Boston Dynamics founder and former CEO Marc Raibert, the institute technically keeps arms' length from Boston Dynamics, or so it seems. It is significantly younger and still in the process of building out its team. TRI has lost interest in the hardware side of the equation.

All this leads to a real general-purpose machine: namely, a system which is firstly essentially capable of learning and doing everything a person can do-and presumably, much more. While we have indeed seen robot hardware getting closer to something which would be able to achieve this level of sophistication, something that approaches general intelligence is an altogether much tougher nut to crack.

Certainly, the availability of SDK for systems has helped significantly increase the breadth of tasks that can be executed by robots such as Boston Dynamics' Spot, but true artificial general intelligence is farther away-if we ever get there.