Brooks, a co-founder of iRobot and a veteran of MIT, expresses doubt about the effectiveness of teaching robots dexterity through human task videos, a method employed by companies like Tesla and Figure. In a recent essay, he dismisses this approach as “pure fantasy thinking.”
The main issue lies in the complexity of human hands, which possess around 17,000 specialized touch receptors that robots cannot replicate. While machine learning has revolutionized speech and image recognition, the lack of a similar foundation for touch data hinders progress in this area, according to Brooks.
Additionally, safety concerns arise with full-sized humanoid robots, as they require substantial energy to maintain balance and pose a significant risk when falling. Brooks highlights the exponential increase in harmful energy with larger robot sizes, making them even more hazardous.
Looking ahead, Brooks foresees successful “humanoid” robots in the future as having wheels, multiple arms, and specialized sensors, deviating from the traditional human-like form. He firmly believes that the current influx of funding is primarily fueling costly training experiments that are unlikely to translate into mass production scalability.
