The humanoid workforce has been running late
Daniela Rus is one of the top AI experts in the world. She spoke to a room full at Boston’s robotics exhibition on Thursday. Rus’s talk included a section where she shattered the myth that humanoids have already proven useful to manufacturing companies and warehouses all over the world.
This might surprise you. Investors have been clamoring for AI to make robot training faster. Figure AI is a startup building humanoids for homes and industries. They are looking to raise $1.5 billion in funding. Amazon and car manufacturers have also been conducting commercial humanoids experiments. Bank of America believes that robots will be more widely adopted in the near future. The number of people who are able to do this is billions Humanoids will be at work in 2050.
Rus, and others that I talked to at the Expo suggest this hype doesn’t make sense.
She said that humanoids are “mostly not intelligent.” Rus played a clip of her speaking with a humanoid who followed the instructions she gave to water nearby plants and pick up an watering can. Impressive. When she asked the robot to “water her friend”, it did not take into account that people do not need to be watered like plants, and proceeded to douse him. She said that these robots lacked common sense.
Pras Velagapudi is the Chief Technology Officer of Agility Robotics. He detailed the physical challenges the company faces. A humanoid that is strong will need a large battery and lots of power. It will run for less time without charging the stronger and heavier you make it. This robot is complex to make.
Humanoids that are impressive don’t necessarily overcome the core limitations, but instead display impressive features, such as nimble robot hands or the ability of interacting with humans via a language-based model. These capabilities are not necessarily applicable to jobs which humanoids will be replacing (it is more efficient to give a robot a list of instructions than speak to them).
It’s not that fleets of robots won’t join us in the workplace, it is just that adoption will be slow, specific to industries, and likely drawn-out. This is related to my last post: For people who view AI as a normal technology and not a dystopian or utopian one, all of this makes sense. It’s important to note that the technology which is successful in a lab will be very different than that adopted by commercial enterprises.
This all sets up what happened last week with one of robotics’ biggest names. Figure AI’s humanoids have raised a lot of money, with Brett Adcock claiming on X that it was “the most sought after private stock in secondary market” in March. Its work for BMW is the one that has been the most widely publicized, as Adcock showed videos of Figure robots moving parts for BMW and claimed that this partnership only took 12 months to start. Adcock and Figure don’t respond to most media inquiries and they rarely attend robot shows.
The month of April is a time for rejoicing. Fortune Published an article that quoted a BMW spokesperson, alleging the partnership between Figure and BMW involves less robots on a smaller scale. Adcock wrote on LinkedIn, “Figure’s legal counsel will aggressively pursue any available legal remedies, including, but not restricted to defamation suits, in order to correct the publication’s blatantly misstatements.” Fortune A representative from Adcock and Figure refused to comment on the parts that were incorrect. Adcock’s statement is lacking in details, according to the representative.
This conflict, despite its specifics in Figure, is indicative of our current tech moment. Venture capitalists are ecstatic, fueled by statements like that of Nvidia’s CEO Jensen Huang who said that “physical AI is the future”. They believe that robots will be able to do most manual work in the near future.
To achieve this, we must overcome countless obstacles. Safety regulations will be needed for humanoids to work alongside humans that do not yet exist. It’s possible that deploying such robots in an industry like the automotive sector will not translate to other industries. AI may solve many problems on the way. Roboticists are right to be skeptical of these things.
From what I have seen, roboticists are usually patient. It took over 50 years for the first robot arm to be produced, while the Roomba was launched more than 10 years after it’s conception. The patience of venture capitalists is not well-known.
Bank of America’s latest prediction of widespread adoption of humanoid robots was met by enthusiasm from investors, but skepticism among roboticists. Aaron Prather is a director of the ASTM robotic standards organisation. He said that these projections are “wildly out-of-base.”
Humanoid hype, as we have discussed before, is a vicious cycle. One video that’s slick raises expectations among investors. This incentivizes other companies to create even more slick videos. It’s difficult for anyone, like a tech reporter to peel back the curtains and see how humanoids will impact the workplace. But I’ll do my darndest.