The quest to build the perfect humanoid robot is heating up, as Figure – a startup currently operating in stealth mode – is developing a versatile bipedal robot it plans to pilot in 2024. Speaking on condition of anonymity, a source close company recently confirmed operations, funding, high profile hires and elements of its overall startup roadmap.
TechCrunch has also seen a pitch deck that gives more details on the figure plans, including a preview of renders of the robot it is working to develop. Currently, the Bay Area startup’s efforts most closely match those detailed by Elon Musk with Tesla’s upcoming Optimus robot. It’s effectively something of a holy grail among roboticists, a humanoid robot that could perform many daily tasks, from manual labor to caring for the elderly. It’s also an almost impossible target.
As a species, we tend to gravitate toward things that look like us. Bipedal robots are much easier to throw – which is a big part of why they dominate science fiction so much. Historically, however, most robots have determined that purpose-built robots are the path of least resistance. The best form factor for a specific job is the rule of thumb, and more often than not it doesn’t involve recreating the human form in its entirety. This is – among other factors ($$$) – why a vacuum cleaner shaped like a hockey puck is the most popular consumer robot to date.
So why are so many people trying to build the perfect humanoid robot in 2022? Boston Dynamics has featured Atlas parkour moves for several years now. Xiaomi paraded a bipedal robot called CyberOne on stage at a recent event, following in the footsteps of companies like Toyota. Tesla’s Optimus is perhaps the most notable of recent vintage – which, after a Spandex-suited knockoff, is set to appear in public later this month.
When I spoke with the team behind NASA’s humanoid robot a few years ago, they suggested a simple premise. We built our buildings and cities for humans, so something that looks and moves like us would be best equipped to navigate them.
That’s the philosophy behind Figure, the brainchild of Brett Adcock, founder of the Vettery Hiring Marketplace and, more recently, Archer. The first was acquired by Swiss recruiting firm The Adecco Group for $100 million in 2018. The second builds an eVTOL. United Airlines recently put down a $10 million deposit to buy 100 flying taxis.
But Figure hopes to succeed where countless extremely well-funded and intelligent people have failed. Along with filling the CEO role, Adcock is also launching the Bay Area-based company with $100 million in funding to get things started.
That money was partly used to hire key names including researcher Jerry Pratt as CTO, former Boston Dynamics/Apple/Arrival engineer Michael Rose and Google/Boston Dynamics roboticist Gabe Nelson as scientist chief. Additional hires from Apple, Tesla, Google X and the Toyota Research Institute round out a workforce of around 30 people (with more hires on the way).
TechCrunch has viewed renderings of an early version of the robot. The system seems more in line with the robot that Telsa is developing, rather than Boston Dynamics’ massive Atlas. The smaller, slimmer frame (human-sized, but on the shorter side) would be electric, rather than the hydraulics that power other robotic systems. It’s an ambitious project, and will likely take several years to get off the ground from a company that was only founded earlier this year. Admittedly, Figure has one of the most formidable recruiting teams I’ve seen from a young robotics startup, as well as the funding to get things started.
Given the scope of the product, however, the company has additional seed funding. Figure seeks to take as much of a comprehensive approach to developing hardware and software in-house as possible. It’s work that will likely come with a wave of patent applications over the next few years.
Some aspects of the work, including battery and SLAM navigation, are transferable from the autonomous vehicle category, hence the hiring of the non-robotics division of Lucid and Tesla. The company is also expected to significantly expand its AI team.
Figure will be aiming for a prototype reveal in 2023 (I anticipate that means going out of stealth late this year or early next year), with piloting beginning in limited quantities the following year. These applications will relate to warehouse work, retail, etc. While it seems likely that early iterations of the product could cost upwards of $100,000, scaling the product could bring it down to around a third of that amount. That’s still a high number (especially for non-industrial use) – as such, it looks like the company will adopt a robotics-as-a-service (RaaS) rental model to make the system more accessible over the course of of the robot’s potential decade. long life.
The company recently registered Figure.ai, which currently features the bespectacled disembodied head of a cat floating in space, while firing lasers out of its eyes. Not much to do there.
The figure did not offer comment on the robot or the chat. We’ll be sure to keep updating the story as we learn more.