Caixin

Commentary: The Misguided Quest for a Human-Like Robot Hand

Published: Oct. 7, 2025  1:49 p.m.  GMT+8
00:00
00:00/00:00
Listen to this article 1x
A humanoid robot showcases its piano-playing skills. Photo: Xinhua
A humanoid robot showcases its piano-playing skills. Photo: Xinhua

From Tesla and Amazon to tech upstarts like RoboSense and Magiclab, companies are investing heavily to develop robotic hands as dexterous as our own — capable of folding clothes, opening a Coke, or even playing the piano. But is this really necessary?

As global tech giants and venture capitalists pour billions of dollars into humanoid robotics startups, Rodney Brooks, a legendary figure in the field, has issued a stern warning: you are wasting your money. According to TechCrunch, Brooks, a co-founder of iRobot and a longtime MIT scholar, cautioned investors funneling fortunes into such ventures.

loadingImg
You've accessed an article available only to subscribers
VIEW OPTIONS

Unlock exclusive discounts with a Caixin group subscription — ideal for teams and organizations.

Subscribe to both Caixin Global and The Wall Street Journal — for the price of one.

Disclaimer
This is an AI-generated English rendering of original reporting or commentary published by Caixin Media. In the event of any discrepancies, the Chinese version shall prevail.
Share this article
Open WeChat and scan the QR code
DIGEST HUB
Digest Hub Back
Explore the story in 30 seconds
  • Tech companies are investing billions in robotic hands, but experts like Rodney Brooks warn this may be wasteful due to challenges balancing dexterity, cost, and reliability.
  • Although human-level dexterity has niche value, most industrial tasks are better served by specialized tools rather than human-like hands.
  • Sustainable progress in robotics should prioritize practical needs and intelligent perception over imitating human anatomy.
AI generated, for reference only
Explore the story in 3 minutes

The global race to develop robotic hands as dexterous as those of humans is attracting heavy investment from major tech players such as Tesla, Amazon, RoboSense, and Magiclab. These companies envision robotic hands capable of tasks ranging from folding clothes to playing musical instruments. Yet, the fundamental question remains whether achieving human-like dexterity in robots is genuinely necessary or merely a technological obsession [para. 1].

Rodney Brooks, a pioneer in robotics and co-founder of iRobot, has publicly cautioned investors and tech firms about the risks and limitations of such pursuits. He describes the current frenzy as a misguided investment, particularly criticizing approaches that attempt to train robots’ fine motor skills by having them watch human task videos. Brooks labels this “pure fantasy thinking,” arguing that it underestimates the complexity of replicating human dexterity [para. 2][para. 3].

Despite these concerns, the market for humanoid robot dexterous hands is projected to expand rapidly. Sealand Securities estimates this sector could be worth 900 million yuan ($124 million) in 2025 and 37.6 billion yuan by 2030, with a compound annual growth rate of 110%. Tesla aims to mass-produce its Optimus Gen3 robot by 2025, noting that the development of its hands represents about half of the engineering workload. Morgan Stanley reports that the hands will account for 17% of the Gen2 robot's hardware costs [para. 4][para. 5].

China’s RoboSense, primarily known for its LiDAR technology, has swiftly developed two generations of dexterous robotic hands, achieving almost 22 degrees of freedom (comparable to Tesla’s), a 5kg payload, and tactile sensors for fine control. Similarly, Dexcelbot, a startup by a former Tencent Robotics X lab leader, has raised significant funding for a robotic hand with over 20 degrees of freedom and a record-breaking 30kg payload [para. 6][para. 7].

Brooks’ core critique highlights the “impossible triangle” of performance, cost, and reliability: enhancing one dimension typically compromises the others. Human-like dexterity requires breakthroughs in materials, actuators, sensors, and AI, yet current costs remain prohibitively high, and reliability suffers due to the complexity of the required mechanisms [para. 8][para. 9][para. 10][para. 11][para. 12].

The human hand’s unique integration of dexterity, strength, and tactile feedback, evolved over millions of years, cannot be readily reproduced by watching videos, Brooks argues. Robots trained this way lack true understanding of, and adaptation to, the unpredictable variables of the physical world [para. 13].

As a result, the wisdom of striving for human-like robot hands is questioned. Many industrial applications can be executed more efficiently and economically with specialized end-effectors rather than general-purpose dexterous hands. For many use cases, pursuing perfect human imitation is not cost-effective [para. 14][para. 15][para. 16].

Nonetheless, fields like medical surgery, scientific research, and rescue scenarios may benefit from near-human dexterity, but the return on investment and reliability remain critical considerations [para. 17].

Brooks and other experts advocate for a shift in focus from imitating human hands to designing specialized, optimized robotic tools for particular tasks. This approach promises greater commercial value and problem-solving capacity [para. 18][para. 19].

Future robotics progress is likely to come from advances in machine perception and cognition, enabling even simpler robotic hands to perform complex tasks intelligently [para. 20].

While some startups, like X Square Robot, continue to develop sophisticated dexterous hands—featuring 20 degrees of freedom and advanced tactile feedback—they recognize that true growth and market acceptance depend on balancing performance, reliability, and affordability [para. 21][para. 22][para. 23].

Ultimately, the quest for robot hands may reach its limits in imitation. The greater value lies in robots that collaborate with humans to solve real problems through innovative and specialized solutions, rather than merely replicating human abilities [para. 24][para. 25].

AI generated, for reference only
Who’s Who
Tesla
Tesla is heavily investing in developing robotic hands as dexterous as human hands. The company's Optimus Gen3 humanoid robot is set for mass production in 2025. Elon Musk has emphasized that the "hand and forearm" development is a core challenge, with the hands potentially accounting for half of the machine's development and 17% of total hardware cost for Optimus Gen2.
Amazon
Amazon is mentioned as a company investing heavily in developing robotic hands as dexterous as human hands. The article groups Amazon with other tech giants like Tesla, indicating its involvement in the competitive and financially intensive field of humanoid robotics development.
RoboSense
RoboSense, a leader in China's autonomous driving LiDAR sector, has branched into developing dexterous robotic hands. Within six months, they released two generations. Their second-generation hand boasts 20 degrees of freedom (similar to Tesla's Optimus), a 5kg payload, and 14 force sensors across its fingertips, pads, and palm, demonstrating precise control.
Magiclab
Magiclab is identified as one of the tech upstarts investing heavily in developing robotic hands as dexterous as human hands. These hands aim to perform tasks like folding clothes, opening a Coke, or playing the piano. The company is part of a broader trend where global tech giants and venture capitalists are pouring billions into humanoid robotics startups, despite warnings from industry figures like Rodney Brooks.
iRobot
iRobot was co-founded by Rodney Brooks, a legendary figure in robotics and a longtime MIT scholar. Brooks has warned against the heavy investment in humanoid robotics, particularly regarding attempts to train robots' fine motor skills with human video footage, calling it "pure fantasy thinking."
Figure
Figure is a company that Rodney Brooks specifically criticizes for its approach to training robots' fine motor skills by showing them videos of humans performing tasks. Brooks calls this practice "pure fantasy thinking," questioning the massive investments in such ventures.
Sealand Securities
Sealand Securities (西南证券) has provided estimations regarding the global market for humanoid robot dexterous hands. They project this market to reach 900 million yuan ($124 million) in 2025 and an impressive 37.6 billion yuan by 2030, indicating a five-year compound annual growth rate of 110%.
Morgan Stanley
A Morgan Stanley report projected that the dexterous hands on Tesla's Optimus Gen2 robot would constitute 17% of its total hardware cost.
Dexcelbot
Dexcelbot is a startup founded by Yang Sicheng, former head of Tencent's Robotics X lab's dexterous hand project. It focuses on developing robotic hand technology. Dexcelbot's robotic hand reportedly has over 20 degrees of freedom and an active payload capacity of 30kg, which is stated to be the highest among publicly announced high-DOF dexterous hands. The company has successfully secured an angel funding round worth tens of millions of yuan.
X Square Robot
X Square Robot, founded in December 2023 by Wang Qian, develops a high-DOF bionic mechanical hand called ARTIXON. This hand has 20 degrees of freedom, 15 drive units, and integrated high-precision tactile sensors. The company aims to create general-purpose robots with fine manipulation capabilities using a self-developed "general embodied intelligence large model," being one of the first in China to adopt a fully end-to-end approach to this goal.
AI generated, for reference only
What Happened When
December 2023:
X Square Robot was founded to focus on general embodied intelligence and dexterous manipulation.
late 2024:
RoboSense began developing dexterous robotic hands.
early 2025:
RoboSense demonstrated its dexterous hand at the U.S. Consumer Electronics Show.
By mid-2025:
RoboSense released two generations of its dexterous robotic hand.
AI generated, for reference only
Subscribe to unlock Digest Hub
SUBSCRIBE NOW
PODCAST
Caixin Deep Dive: Former Securities Regulator Yi Huiman’s Corruption Probe
00:00
00:00/00:00