Investors just gave one robotics arm company a major lift.
Where Does Kamala Harris Stand on Climate, Trade and Tech?
Robotics startup Standard Bots announced July 12 it had secured $63 million in Series B funding, led by General Catalyst with additional participation from Amazon Industrial Innovation Fund and Samsung Next.
The Glen Cove, New York-based startup, will use the newly announced funds to amp its research and development (R&D) work—both on a new, AI -powered robot and on a larger robotic arm for logistics —and to open a new manufacturing site, which will allow it to build its robot arms itself.
Evan Beard, the company’s co-founder and CEO, said Standard Bots already assembles the cobot arms it sells or leases to clients, but fully building its robot arms would be a major milestone.
“With our next facility, we’re really trying to make a Made in America arm where we have 12-foot-long stock pieces of metal going in the door and robots going out, so we’re really excited about that,” he said.
He went on to say that the majority of the individual parts will be primarily sourced from the United States. The company currently imports some of its parts, but Beard noted that nearly all of the imported pieces are of Standard Bots’ own design—the motors, controllers for the motors and more.
“Other than the chips, we make almost everything in the arm. For this new facility, we’re literally buying aluminum—and our plan is to buy the aluminum local and the steel local,” he said. “Our goal is a Made in America arm—that virtually everything, other than some of the chips and minor components [is sourced from the U.S.].”
Beard said he expects the facility, which will be about 15,000 square feet, to come online around October. The company’s goal is to start putting out its first Made in America robots by the end of the calendar year, and it plans to add a few positions to its production team to account for the additional work that will need to be done in the facility.
In addition to its new building, Standard Bots has some lofty goals ahead for the type of robots it will offer to clients; it already has two core models, which it calls RO1 and RO2, in production.
On top of the new way of training robots, the company has been gearing up to release another robotic arm, which it will call the RO3-Max, into the market. It will be larger than the previous models Standard Bots has created and will also have increased capabilities, like a two-meter reach and a 30-kilogram payload, Beard said.
CMA CGM Taps Google to Deploy AI Across Shipping, Ceva Logistics
The RO3-Max’s abilities will serve the logistics industry particularly well, he noted. In discussing the new arm’s potential with logistics companies, the company has placed a specific onus on ensuring the robot isn’t too heavy and can move efficiently and quickly in a logistics setting.
Beard said that some of the company’s other work will focus more heavily on in-factory assembly for goods that have always required humans’ touch to build out.
Those kinds of capabilities are only possible because of a new approach to training the robots with artificial intelligence , he said. The Series B round, in addition to helping the company open a new facility and release a new arm, will allow it to drastically change training measures with technology.
The company is actively training AI models that will allow people to demonstrate what they want the robots to do, which the robot will be able to learn from and replicate. The technology behind the new training process includes cameras for computer vision and foundation models, Beard said.
“It’s using a lot of the same underlying technology as a ChatGPT or Stable Diffusion, and it’s learning that policy, and it’s using the camera data in real time to make decisions. So, if the robot has dropped something or something’s assembled wrong, it can correct it in real time,” he explained.
Beard noted the company is working to build both its own foundation models and the applications that layer on top of those models, which can be a difficult feat—many companies have instead chosen to build on top of pre-existing foundation models, like OpenAI’s GPT-4 or Meta’s LLaMa.
One concern that has started to arise with the building and training of such models is that they often have a tendency to require a great deal of energy resources. Beard, though, said he doesn’t have major concerns about that, nor does the cost of building a foundation model stop the company.
“I think the potential is so great to do so much good for the world that the training cost is a fraction of the potential benefit,” he said. “I think there will be very few use cases, when we look back in five years, that this new way of programming doesn’t touch.”
Privacy-Centric Lawsuit Against Patagonia Shows AI's Murky Legal Footing
Most Popular
Comments / 0