Humanoid robots not future: Physical hurdles AI can't fix

An opinion piece by the Financial Times presents a critical analysis of the hype surrounding humanoid robots, particularly in light of advances in artificial intelligence (AI). The author discusses the example of Pepper, the humanoid robot introduced by SoftBank in 2014, which was heralded as a breakthrough in robotics but failed to live up to expectations.

The article argues that while AI has made significant strides in areas like machine learning and pattern recognition, the physical challenges of robotics—such as creating human-like mobility, sensory feedback, and energy efficiency—remain formidable.

The piece starts by highlighting the case of Pepper, a humanoid robot designed to interact autonomously with humans, which was widely promoted as an example of how robots could integrate into everyday life. Despite initial optimism, the robot failed to reach the mass adoption that its investors hoped for, with production halting in 2021. The author uses this failure to underscore the difference between the allure of humanoid robots and the reality of creating machines that can perform complex tasks like humans.

The article emphasizes the fundamental difficulty in building robots that replicate human-like behavior, focusing on the physical challenges. Human limbs are powered by muscles, but robotic limbs require motors, gears, and complex transmissions that add bulk, cost, and the risk of breakdowns. Additionally, the article notes the challenge of sensory feedback—robotic systems lack the ability to "feel" in the same way that humans do, relying on machine vision to make inferences about touch, taste, and smell, which are imperfect substitutes.

While AI advancements are improving tasks such as machine vision and data processing, the piece argues that AI alone cannot solve the physical issues inherent in robotics. For example, robots need energy sources like batteries, and current technologies cannot address the trade-offs between size, power, flexibility, and cost effectively. The author points out that even a revolution in AI does not automatically resolve these fundamental hardware challenges.

Rather than focusing on humanoid robots, the article suggests a more practical application of AI in enhancing existing machines, like self-driving cars and industrial robotic arms. The example of autonomous vehicles is particularly notable: AI can help optimize the sensors and algorithms that drive a car without requiring a complete overhaul of the vehicle’s physical infrastructure. Similarly, AI can gradually improve the capabilities of existing robots, such as factory robots and vacuum cleaners, making them safer, more versatile, and better at interacting with humans.

The article concludes by acknowledging that AI will slowly advance the field of robotics, but it will not lead to the creation of humanoid robots capable of doing complex household tasks like cleaning. Instead, it suggests that simpler, more controlled forms of autonomy, like robot delivery systems in hotels, are more likely to become widespread. The piece humorously points out that while AI may help robots write bad poetry, it is far from making them competent in tasks like cleaning toilets.

The overall tone of the opinion piece is cautionary. While it recognizes the significant potential of AI in robotics, it warns against overestimating how soon we will see truly autonomous humanoid robots capable of performing complex tasks in daily life. The piece advocates for a more grounded view of AI's impact, focusing on the incremental improvements it can bring to existing technologies rather than expecting a sudden breakthrough in humanoid robots. This analysis underscores the reality that while AI can make machines smarter, the hardware challenges required to bring truly autonomous robots to life remain a significant hurdle.

By Vafa Guliyeva