Despite billions in investment, the dream of lifelike humanoid robots is clashing with hard scientific limits—and a history of overpromising.
The Voice of Skepticism in a Robotic Gold Rush
Renowned roboticist Rodney Brooks, co-founder of iRobot and a longtime MIT researcher, is challenging the wave of enthusiasm—and capital—flooding humanoid robot startups. While companies like Tesla and Figure promise near-human capabilities, Brooks warns that this bubble is doomed to burst.
- Fantasy vs. Physics: Brooks criticizes efforts to train robots using video footage of humans doing tasks, calling it “pure fantasy thinking.”
- Dexterity Isn’t Just Visual: Unlike speech and vision, robotic dexterity lacks decades of foundational research and usable data.
The Missing Ingredient: Touch
One of Brooks’ central arguments is that human touch is vastly underestimated in robotic design. Our hands contain roughly 17,000 touch receptors, a level of nuance that no current robot matches.
- No Data Tradition: Speech and image recognition advanced because they were built on robust data infrastructures. Touch data has no such legacy.
- Machine Learning Limits: AI can’t replicate touch by watching videos—it requires tactile feedback, something robots simply don’t yet possess.
Safety Concerns: The Physics Problem
Beyond the lack of dexterity, Brooks highlights another concern: safety. Humanoid robots are inherently unstable and can be dangerous.
- Energy and Scale: As robots grow in size, their potential for harm increases exponentially. A robot twice as big can cause eight times the damage if it falls.
- Design Rethink Needed: Brooks argues future successful robots will likely have wheels, multiple arms, and specialized sensors, not human-like bodies.
Billions for What?
Despite his concerns, the money keeps flowing. Brooks believes these investments are funding experimental prototypes that may never scale.
- Unsustainable Models: Current efforts are more like academic training labs than commercial pathways.
- 15-Year Outlook: Brooks predicts that within 15 years, the most useful robots will look nothing like humans—and many of today’s startups may be long gone.
Generative AI: A Parallel Problem
Brooks also draws comparisons to generative AI, another tech darling with unfulfilled promises. He points to a recent METR study that found AI tools slowed down developers by 19%, even as they felt they were working faster.
- Perception vs. Reality: The illusion of productivity often outpaces actual gains.
- More Work, Not Less: Generative AI sometimes creates more cognitive load, especially in complex real-world scenarios.
Big Tech’s Robotic Gravitational Pull
Although Brooks once believed Big Tech wouldn’t dominate robotics, recent moves suggest otherwise.
- FigureAI and OpenAI: Their short-lived partnership showed how even top startups are drawn into Big Tech ecosystems.
- Apptronik and Google: With nearly $450 million raised and Google as a backer, Apptronik’s partnership with DeepMind suggests that robotics innovation is still heavily tied to AI giants.
The Real Future of Robots
Brooks envisions a future less about humanoids and more about practical, purpose-built machines. He emphasizes function over form, and safety over spectacle.
- A Post-Humanoid Era: Expect robots with wheels, not legs, and arms tuned to task, not mimicry.
- Hardware Matters: Breakthroughs will come from new sensors, better motors, and smarter design, not just bigger models or better training videos.
