Humanoid Robots in the Limelight: OpenAI and Figure’s Latest Advancements Unveiled by Douglas Campbell

Humanoid Robots

Humanoid Robots open AI is trying to design a better walking, talking humanoid robot. It was only a few years ago that it was nearly impossible to look at autonomous, human-shaped bipedal Humanoid Robots without laughing. Notably, two-legged Humanoid Robots that were designed to partake in a series of high-profile Pentagon challenges often used to stumble and fall on their faces while trying to navigate obstacle courses. A bot that was designed by Tesla, Inc. traversed a thin line between disappointment and controversy, years after it handheld a video taped of a machine dancing in a skin-tight bodysuit.
Nonetheless, in spite of those gaffs, robotics firms pressed on and now several believe their walking machines could work alongside human manufacturing workers in only a few short years. Figure, one of the more prominent companies in the humanoid robot space, this week told Pop Sci it raised $675 million in funding from some of the tech industry’s biggest players including Microsoft, Nvidia and Amazon founder Jeff Bezos. The company also announced it has struck a new agreement with generative AI giant, Open AI to “develop next generation AI models for humanoid robots.” The partnership marks one of the most significant examples yet of an AI software company working to integrate its tools into physical Humanoid Robots.
According to Figure Founder and CEO Brett Adcock, the partnership that the company has had with Open AI is a “huge milestone for robotics.” He has expressed hope that, ultimately, the firm’s partnership with Open AI will culminate in a robot capable of working with other humans to complete their tasks and holding the latter intelligent conversations. This arrangement will allow Figure to work with the creators of the world’s most popular LLARGE language models. According to Adcock, this should allow a robot to have more “semantic” understanding of the world, thereby making it more useful in work contexts.
He said, “I think it’s getting more clear that this [humanoid robotics] are becoming more and more an engineering problem than it is a research problem,” he said. “Actually being able to build a humanoid [robot] and put it into the world of useful work is actually starting to be possible.
Founded in 2021, Figure is in the process of building a 5 ‘6 130-pound bipedal “general purpose” robot that it says lifts objects about 45 pounds and walks 2.7 miles per hour. The company believes its Humanoid Robots could help address potential labor shortages in manufacturing jobs and generally “enable the automation of difficult, unsafe, or tedious tasks,” whatever those tasks may be. While it is not clear just how good current humanoid robots are at actually performing those sorts of tasks, Figure recently released a video of its Figure 01 model slowly walking towards a stack of creates, picking one up with its two hands, and tossing it onto a conveyor belt. The company says the robot did the entire job on its own.
Humanoid-style robots have had a long line of backers; those who point out that being bi-pedal makes them much more capable of climbing stairs, and making their way up uneven, or even unpredictable ground, over their more typical, wheeled or tracked alternatives. However, the technology behind the robots has come a long way from the slips and stumbles of yesteryear. When the publication Wired spoke with Figure Chief Technology Officer Jerry Pratt last year about the company’s robots, he asserted that their machines could complete the Pentagon’s test course in only a quarter of the time it took machines to do the same thing in 2015, in no small part due to the improvements made in computer vision technology. Other bipedal robots like Boston Dynamics’ Atlas can already do backflips and has a pretty worryingly good throwing arm.
As per Figure states, the new “collaboration agreement in place is going to take Open AI’s research, and combine it with the experience that [Figure has] in robotics hardware and software”. If things go according to the plan, the partnership will result in the development of a “robot that can process and reason from language”. This feature will allow robots to recognize the language and act according to the information they receive. As a result, robots will be able to better interact with a human warehouse worker and receive verbal commands. Adcock says that he is looking forward to the integration of language model processors into robots.
Adcock claimed that in the long run, people interacting with the Figure should be able to talk with the robot in plain language. Then, the robot will create a list of tasks and complete them by itself. Moreover, the cooperation with the Open AI will help the Figure robot self-correct and learn from its past mistreated tasks quicker. Adcock adds that the Figure is able to speak already and can manage by taking a picture with its cameras and than describes what it “sees” in front of it. It can also tell you what happened in a certain area over a period of time.
In a statement sent to Pop Sci, Open AI VP of Product and Partnerships Peter Welinder said, “We’ve always planned to come back to robotics and we see a path with Figure to explore what humanoid robots can achieve when powered by highly capable multimodal models.” Open AI and Figure are not the only companies trying to figure out how to use language models in human-looking robots. Last year, Walter Isaacson, a biographer of Elon Musk, wrote an article for Time, citing sources saying the Tesla CEO was considering how to integrate his newly improved Optimus humanoid robot and its accompanying training computers, Dojo, to create what some researchers call artificial general intelligence, a machine that could perform at many tasks above human level.
A number of tech giants have placed a big bet on Figure in a brewing humanoid robot race for supremacy. The $675 million in funding it disclosed this week was reportedly over $150 more than it had initially sought, according to Bloomberg . The company says it’s planning to use that capital to scale up its AI training, robotic manufacturing, and add new engineers. Figure currently has 80 employees. On that note, the support from Open AI, in addition to its gargantuan new wave of funding, might help speed up Figure’s previously announced timeline for making its product commercially available .
There is another humanoid robotics company with serious investments from Open AI, 1X Technologies AS, that has recently raised $100 million. Oregon-based Agility Robotics recently demonstrated how its Humanoid Robots could perform a variety of simple warehouse tasks autonomously. According to the same report, the company is now already testing machines in Amazon warehouses . Meanwhile, Figure reported a partnership with BMW, in the framework of which, the humanoid robot will be brought to the South Carolina BMW plant.
So, how much do you think companies like Amazon, Alphabet, Softbank, Huawei, and Alibaba are spending on building these companies? After raising over $1.5 billion last year, one company, Mega Robo, reportedly passed the $1.1 billion valuation. Wow! Do you think those companies and investors know something we don’t? There is another aspect you probably haven’t realized yet. The stocks of the companies previously mentioned have also started to tick up. Is all this just coincidence? I think not.
However, there is still much more to go before any of those big ideas start looking like a reality. These still very much in progress technology are only just beginning to be trialed and tested within the larger manufacturing buildings. The most impressive of these Humanoid Robots , such as the Boston dynamics product that can do a little jig, are still incredibly expensive to produce. It is also not quite clear yet whether these robots could ever mimic humans tasked with several different things to the same effect.
In general, it is not quite clear yet what specific tasks they are really good for solving. Both Elon Musk and Figure himself mentioned the possibility of dealing with assignments that are too dangerous, dirty, or simply unpleasant for human beings . But again, I do not really understand what they are talking about. For example, BMW has previously told Pop Sci that it is still “investigating concepts” when asked how the company plans to use Figure Humanoid Robots. Adcock went so far as to say that the Figure robot can be used to move sheet metal or perform other tasks in the body shop . Adcock also mentioned that Figure has five primary use cases for which the robot was designed, keeping them a secret from the public.
What to do with these Humanoid Robots after making them isn’t exactly a unique problem of the Figure, and in an interview with Pop Sci , Ding Zhao called that problem of use-cases a “billion-dollar question.” “Generally speaking, we are still exploring the capabilities of humanoid robots, how effectively we can collect data and train them, and how we can ensure their safety when they interact with the physical world.” added Zhao who argued that robots which are expected to work alongside humans will have to invest a lot in safety as well, which can even be equivalent to, if not exceed the costs of development.
The Humanoid Robots will need to get better, too, especially in messy, real work environments that aren’t as predictable and structured as robot training labs. Adcock said the robot’s speed at tasks and payloads will need to increase . But all of that, he said, can be tackled with powerful AI models like the one Open AI is building.
“We think we can solve a lot of this with AI systems,” Adcock said. “We really believe here that the future of general purpose Humanoid Robots is through learning, through AI learning.”