HomeArticle

Well-known robotics experts are speaking out: The billions of dollars invested in humanoid robot startups are going down the drain.

AI前线2025-10-10 18:14
Robot experts question the route of humanoid robots, stating that there is insufficient tactile data and high safety risks.

Renowned robotics expert Rodney Brooks is the co-founder of iRobot and spent decades at the Massachusetts Institute of Technology (MIT). He is particularly skeptical of the technological approaches of companies like Tesla and the highly - watched AI robotics firm Figure. These companies attempt to teach robots the ability to perform flexible operations by having them watch videos of humans performing tasks. In a new article, he calls this method "pure fantasy."

What's the problem? The human hand has an extremely complex structure, with about 17,000 specialized tactile receptors. Currently, no robot can come close to this level. Although machine learning has revolutionized the fields of speech recognition and image processing, these breakthroughs are based on decades of accumulated existing technologies for capturing high - quality data. Brooks points out, "In the field of tactile data, we don't have such technological accumulation."

There are also safety issues. Full - sized walking humanoid robots need to consume a large amount of energy to stay upright. Once they fall, they can be extremely dangerous. From a physics perspective, if a robot is twice the size of an existing model, the harmful energy it generates when falling will be eight times that of the existing model.

Brooks predicts that 15 years from now, successful "humanoid" robots will actually be equipped with wheels, multiple robotic arms, and specialized sensors, and will completely abandon the human form. Meanwhile, he firmly believes that the billions of dollars invested today are just paying for expensive training experiments that will never achieve large - scale mass production.

This is not the first time Brooks has poured cold water on ambitious entrepreneurs and eager investors. Last year, he had an in - depth discussion with foreign media about why the hype of generative AI far exceeds its actual capabilities, and in some cases, it even increases the workload.

For example, this summer, the non - profit AI research organization METR said that they recruited 16 high - scoring developers from large open - source code libraries to measure the impact of AI tools on actual software development work. The research team assigned nearly 250 real - world development problems to these developers, asking them to solve the problems with and without AI tools respectively, and monitored their operation processes. The results showed that when using AI tools, the time it took for developers to complete tasks increased by 19%. Interestingly, however, the developers themselves thought that AI had improved their work efficiency by 20%.

For a long time, Brooks has also believed that AI is not a "threat to human survival," as many people, including Elon Musk, have claimed. As early as 2017, TechCrunch discussed this topic with Brooks at MIT. Although the technological landscape at that time was different from today's, it was not completely dissimilar.

At that time, Brooks said that he had begun to see more and more companies focusing on creating datasets for machine learning, and this trend has continued since then. Related to this, he also once proposed the view that although large technology companies have seemingly insurmountable advantages in data control in the long run, it does not mean that their victory in the robotics field is an "inevitable outcome." However, today's leading robotics companies still cannot escape the influence of these tech giants.

For example, the humanoid robot manufacturer Apptronik has raised nearly $450 million from investors, and Google is one of its investors. At the end of last year, the company also reached a cooperation with Google DeepMind's robotics team, aiming to "combine top - level artificial intelligence technology with cutting - edge hardware and embodied intelligence."

Similarly, Figure's investors also include Microsoft and the OpenAI Startup Fund. In February 2024, Figure partnered with OpenAI, planning to combine OpenAI's research results with its own "in - depth understanding of robot hardware and software." But just one year later, in March this year, the two sides officially parted ways. FigureAI said that its self - developed end - to - end AI for robots has achieved a "major breakthrough."

Earlier this month, Figure announced that it had received over $1 billion in committed capital in its latest round of financing, and said that the deal valued the company at a staggering $39 billion.

Reference Link

https://techcrunch.com/2025/09/26/famed-roboticist-says-humanoid-robot-bubble-is-doomed-to-burst/

This article is from the WeChat official account "AI Frontline," written by Hua Wei and published by 36Kr with authorization.