Boston Dynamics: The Relentless Robotics Pioneer
In a recent interview on the All-In podcast, Sergey Brin was asked about Google’s decision to divest from its former robotics companies. He gave a somewhat indeterminate answer along the lines of “…I think that was just a little too early to be perfectly honest”. Suggesting that in light of the current breakthroughs in LLM’s, its hard to forecast AI’s future and predict at what rate this technology will evolve in the coming decade.
The field of robotics has been around for quiet some time however talks about autonomous robots got a lot of attention lately, particularly since the term ‘AGI’ became mainstream, & discussions on how it will impact the workforce.
Humans however are not new to robots (if one might say), as various industries have been incorporating them into manufacturing plants since the 1960’s. The engineering dynamics of building a robot capable of movements mimicking a human or animal have always been a challenge for roboticists.
Boston Dynamics - founded in 1992 by an MIT professor named Marc Raibert, evolved from humble beginnings at MIT as a "Leg Laboratory", doing research on legged systems simulation software to eventually innovating in control algorithms & building many prototypes, with two robots now commercially available: Spot & Stretch.
As talks of AGI taking over humanity in six months are wayning down (following the launch & hype of Chat GPT)... let's not forget that Moravec’s paradox is still a major challenge in robotics engineering. As a recap; that notion highlights simple things like walking or doing repetitive tasks a hard challenge for robots, while extremely hard scenarios for humans (like solving advanced math) being easy. Over the years roboticists have combined various training methods to make the robots more efficient, from reinforcement learning to gathering neuroscience insights.
Although the company at first focused on R&D, often working with the U.S. military on battlefield prototypes, Google (Alphabet) eventually tried to push its lead in robotics by purchasing Boston Dynamics in 2013.
It sold it to Softbank in 2017, & subsequently a lot of shares were sold to the Hyundai Motor Group in 2021.
Recently, the Toyota Research Institute (TRI) partnered with Hyundai to collaborate on R&D. TRI will help Boston Dynamics to further develop Atlas, with its knowledge of large behaviour models & reinforcement learning (like Diffusion Policy).
Historically car manufacturing plants & warehouses have been progressive at welcoming robotics innovations.
Boston Dynamics is positioning itself as a pioneer in those industries (despite its recent layoffs) & over time managed to commercialize two advanced mobile robots: Spot & Stretch.
These robots already help alleviate the growing demand for labour shortage at warehouses & manufacturing plants, as well helping with public safety.
Spot is a robot that resembles a dog & has been used in cases from inspecting large facilities, providing assistance in emergency rescue missions, even helping law enforcement scanning areas for breach (& analyze suspicious objects).
Stretch is a robot that handles freight cargo & helps alleviate tedious tasks like hand bombing a semi-trailer (where temperatures can reach health hazard levels on both extremes).
Despite the general public fantasizing about humanoid robots taking over society, a majority of engineers concur that one of the main "limiting" factors in robot design is the capacity to carry an energy efficient battery package. A robot shaped in the form of a human isn't necessarily the ideal design for that.
Nevertheless Atlas is a humanoid looking robot project that BD is not giving up on. Although not yet commercially available - it recently got an upgrade by becoming fully electric. Engineers replaced outdated hydraulic parts with advanced electric components (like electric actuators).
Boston Dynamics is by far the most advanced Robotics company in the world, and Atlas is the most sophisticated human looking robot ever built. Despite not necessarily having as much battery carrying capacity, Atlas already exceeds the range of motions that a natural human body can do, thus promising productivity leverage once deployed to work in a human staffed environment.