The humanoid robotics scene has seemingly exploded in recent years. For a decade, it seemed that Boston Dynamics was the only company in humanoid robotics. However, in the last year or two, a new cohort of humanoid robotics startups has emerged. New players such as Figure, Agility Robotics, 1X Technologies, Tesla and more have entered the stage.

These companies promise that within the next year or two, commercial humanoid robots will become a reality and start working alongside humans in factories, warehouses, and other places.

The hype is real, and the sci-fi dream of humanoid robots walking among us seems to be just around the corner. But are these high hopes justified? Will humanoid robots meet the high expectations, or will they fail like many overhyped and overpromised technologies before them?

In this article, we will explore what humanoid robotics companies promise to deliver, what they need to do to succeed and answer the question of whether humanoid robots will take off.

Humanoid robots are the new hot thing

A number of trends have converged to make the humanoid robotics explosion possible. We now have smaller, more powerful, and more efficient electric motors powered by more energy-dense and less expensive batteries. Computers have become smaller and more powerful, and combined with better software and artificial intelligence algorithms, this has made the robots smarter and more capable. These robots can now learn on their own in simulated environments, repeatedly refining how to use their bodies. Additionally, rapid prototyping and manufacturing have become cheaper and faster, which in turn has made the development of humanoid robots more affordable and efficient.

We have also seen significant investments made in humanoid robotics companies. Earlier this year, Figure raised $675 million in Series B funding from major investors such as Microsoft, OpenAI Startup Fund, and Nvidia, bringing the company’s valuation to $2.6 billion. 1X raised $23.5 million in Series A funding, led by OpenAI, in 2023, followed by a $100 million Series B funding round this year. Other companies in this space are also well funded, with each raising millions or even tens of millions in funding.

Apart from newcomers, we also have established companies like Tesla and Boston Dynamics taking commercial humanoid robots seriously. Tesla is developing Optimus to work in its factories. Meanwhile, Boston Dynamics, the pioneers in robotics and humanoid robots, has retired the original Atlas robot in favour of a new, all-electric robot and is preparing to start selling humanoid robots.

With all of those resources, these companies have been building world-class teams of top talents in robotics, electronics, mechanical engineering, and artificial intelligence to develop their machines. The results are quite impressive. Figure has shown their robot using OpenAI’s GPT-4 to understand human language and take correct actions. 1X demonstrated its robots being able to do the same. Boston Dynamics showed how agile their robot is. The Chinese Unitree H1 robot runs and dances. Meanwhile, Tesla went from a guy dressed as a robot to a fully functioning humanoid robot in the span of two years.

Some of these robots are already being tested in real-life scenarios. Agility Robotics’ Digit is being trialled at Amazon, while the Figure 01 robot is being tested by BMW, and Mercedes is trialling Apptronik’s Apollo for "low skill, repetitive" tasks. Sanctuary AI, another humanoid robotics company, has partnered with European carmaker Magna to deliver their Phoenix robots to Magna's factories. Recently, another car company from China, Dongfeng Liuzhou Motor, has also started to incorporate humanoid robots in their factory.

All of this makes humanoid robots (and robotics in general) the new hot thing in tech. Jensen Huang, CEO of Nvidia and a master of riding tech waves, from crypto to AI, started to include humanoid robots in his keynote speeches this year, saying that the future of AI is robotics. The field is moving fast and every month brings a new breakthrough. Some even believe we are very close to robotics having its own ChatGPT moment.

But why a humanoid robot?

Before we explore deeper whether humanoid robots will take off, I think it is necessary to ask why we are building these machines in the first place. What is their purpose, and what problem are they solving?

The purpose of humanoid robots is simple—they are meant to replace expensive human labour with cheap robotic labour.

The companies building these robots promise that their machines will fit neatly into existing workspaces, use the same tools a human worker uses, and start working immediately, with no need to adjust the work environment to accommodate the robots. The robots will be intelligent enough so that all that is needed is to tell them what they need to do and will do that perfectly. The humanoid robotics companies hope their robots will address labour shortages and take over repetitive, low-skill jobs that are difficult or too expensive to automate using non-humanoid robots.

Additionally, humanoid robots promise to scale labour costs up and down easily, work for longer periods of time than humans can, and be easier to deal with than human employees. They are touted as the easiest way to introduce automation into workplaces. Furthermore, if the workplace is already somewhat automated, humanoid robots promise to connect the pockets of automation that are currently linked by human workers to create a more automated workplace.

And all of this is promised to be done by one type of general-purpose robot—not a group of single-purpose robots, but one, easy-to-deploy and train type of machine.

What humanoid robots need to succeed?

The success of any humanoid robot will depend on how much cheaper and how good of a job these machines are compared to human workers and traditional robots.

Humanoid robots are complex machines. They are a marvel of mechanical and electrical engineering combined with the best of what artificial intelligence has to offer. An enormous amount of research and development has gone into creating these machines, so it is reasonable to expect that they won’t be cheap initially.

Of all humanoid robotics companies, only one has revealed the price of their robot. Unitree H1 robot is priced at $150,000. For that price, you get a general-purpose humanoid robot that can walk and run but does not have hands. Unitree also offers a smaller, 1.27m tall G1 robot. The price for the G1 robot starts from $16,000, and Unitree asks prospective buyers to contact them "for the real price," which likely means the actual price will be higher than $16,000.

For comparison, let’s look at the prices of robot dogs. Boston Dynamics asks $75,000 for a base Spot robot dog. Adding extra upgrades and equipment can elevate that price tag to almost $200,000. Unitree does not provide the price tag for their most advanced robot dog, B2, but I found one place offering them for about $90,000.

It’s fair to assume that humanoid robots—robots that are bigger, more complex, and require more intelligence than robot dogs—will cost more than robots like Spot or the Unitree B2. Therefore, the realistic price for a more capable humanoid robot should be around $100,000 or higher. Then, you need to take into account the cost of support and maintenance, the cost of supervision, and other hidden costs associated with operating a humanoid robot.Benjie Holson has a good analysis of these costs over at General Robots showing that operating a humanoid robot might cost more than hiring a human.

We rely on what robotics companies show us to assess the capabilities of the current generation of humanoid robots. And they showed some impressive feats, such as a group of 1X robots cleaning an office. These robots are also proficient at manipulating objects, which, as Moravec’s Paradox has taught us, is a deceptively difficult problem to solve. Both Sanctuary AI and Tesla have showcased how good their robots are at picking objects and moving them to a new place.

These demos do a great job as engineering milestones and in showcasing what the technology is capable of. However, they fall short of portraying the practical usefulness of these robots.

A prospective humanoid robot operator will have to ask themselves what the best choice will be: hiring a human, getting a humanoid robot, or getting a more traditional, specialised robot.

If we take moving objects from one place to another as an example of a practical task a humanoid robot could do, then there are more efficient and cheaper ways to automate moving objects from one place to another. You can get a quite capable small robotic arm for $25,000-$30,000 (or maybe even less) and train it to move objects from one place to another. Or you can get a specialised machine that will move these objects faster than any robotic arm can.

If you need to connect pockets of automation, I’d look first at bringing them together. If that does not work, connecting them with a conveyor belt or an autonomous mobile robot like the ones Amazon uses might be a simpler, more cost-effective and efficient solution.

If your robot needs to load and unload trucks, then maybe something like Boston Dynamics’ Stretch will be a better option.

If you need a robot to operate a tool, then maybe turning a tool into a robot is a better solution.

If your objective is to improve the efficiency of a business, then humanoid robots are unlikely to help you reach that goal in the near future. You might be better off investing in traditional robotics or hiring humans. However, I can see a case for humanoid robots when traditional robots or humans are not available and a solution is needed now. In that situation, if the cost of inaction is greater than the losses in efficiency, then humanoid robots could be a good solution (assuming they can do the required job).

I’m confident that the capabilities of humanoid robots will continue to grow. The challenge for humanoid robotics companies will be to find that magical point where the capabilities justify the cost of operating a humanoid robot.

How will they work with humans?

One of the humanoid robots’ key selling points is their ability to slot into existing workflows alongside other robotic or human co-workers. To work alongside humans, humanoid robots need to be able to communicate effectively and be safe to work around.

Thanks to large language models, robots can now understand human language. They can take in a high-level command and do what is asked them of, like in this demo from Figure. However, spoken language is not the only way we communicate with each other. We also use non-verbal communication—body language and eye contact—to express our emotions and intent.

I don’t see many companies focusing much on non-verbal communication. They do the bare minimum by adding a head with some kind of display to show the robot’s face and to communicate non-verbally from a robot to a human. I haven’t seen any demos of humanoid robots understanding human facial expressions or gestures, such as using hands to signal the robot to come over or to stop. Robotics companies will need to address this issue of non-verbal communication to ensure their robots understand human intentions and are safe to work with.

Speaking of safety, that’s another critical aspect that every humanoid robotics company seems to be deprioritising at the moment. These robots are big, heavy machines made from metal, which could be dangerous to be around if something goes wrong. I haven’t seen any of those robots having an emergency stop button—a big red button that every industrial machine is required to have to shut down immediately in case of an emergency.

I understand that at this point in development, companies are focusing more on making their robots work, and safety is not at the top of the list. There are ways to safely introduce a humanoid robot into a workplace without fully solving the safety problem, such as by constraining the robot to work in a closed area (that’s what Agility Robotics does with Digit). However, if one of the promises of humanoid robots—being able to work in a human environment without any changes to it—is to be fulfilled, then these robots need to be safe to be around.

Even if humanoid robots can effectively communicate with people and be safe to be around, there is one more factor that could impact their adoption: how humans will react to their robotic coworkers.

If a humanoid robot is added as an augmentation, as a helper, then I think the dynamics of the human-robot relationship could be healthy and positive. However, the equation changes when a robot is seen as a replacement. In that case, a humanoid robot becomes a walking symbol of jobs lost due to automation and their adoption could be met with a strong negative response. The ongoing resistance and disdain against AI art could be a taste of things to come if the mass adoption of humanoid robots comes to be.

That resistance may not be a significant factor in the adoption of humanoid robots in factories or warehouses. However, if humanoid robots fail to find success there or if companies wish to expand, these robots will seek new applications in environments closer to humans. They might attempt to pitch their robots to work in hospitals, supermarkets, or other places in close proximity to people. In such cases, public perception of humanoid robots will matter more than their capabilities.

So, will humanoid robots take off?

My honest answer is, “I don’t know.” There are a number of unanswered questions and unknown variables that need to be just about right for humanoid robots to succeed.

Automation will still happen but the question is what shape the automation will take. Will it be in the shape of a general-purpose humanoid robot capable of doing everything a human can do? Or will the industry instead of an army of humanoid robots working in warehouses and factories, choose to become more like Amazon, and use an army of cheaper, simpler and easier-to-maintain specialised robots?

Overpromising is one of the biggest risks facing humanoid robotics companies. The flashy demos we see on YouTube might give a false impression of the true capabilities of humanoid robots.

The real test for these robots will be their interaction in real workplaces with real customers, who expect them to deliver value rather than just being there for exposure. This contact with reality will be harsh, and I believe the first wave of humanoid robots will not meet the high expectations set for them. Many humanoid robotics companies will not survive this initial reality check.

The critical question will then be what lessons the surviving companies learn and whether they can transform the dream of humanoid robots into a product that the market wants.

Thanks for reading. If you enjoyed this post, please click the ❤️ button or share it.

Share

Humanity Redefined sheds light on the bleeding edge of technology and how advancements in AI, robotics, and biotech can usher in abundance, expand humanity's horizons, and redefine what it means to be human.

A big thank you to my paid subscribers, to my Patrons: whmr, Florian, dux, Eric, Preppikoma and Andrew, and to everyone who supports my work on Ko-Fi. Thank you for the support!

My DMs are open to all subscribers. Feel free to drop me a message, share feedback, or just say "hi!"

Message Conrad Gray