Humanoid AI Will Reach GPT-3.5 Level in 3-5 Years

Humanoid robots are poised to achieve a level of general task competence comparable to OpenAI's GPT-3.5 model within the next three to five years, according to Yao Maoqing, co-founder of Chinese robotics firm AGIBOT. Yao made this prediction during the Fortune Leaders Forum in Macau on September 8, referencing the significant impact of GPT-3.5, the model that powered ChatGPT. This AI service demonstrated the ability to perform common, everyday tasks with an 80% to 90% success rate, marking a pivotal moment for AI accessibility and utility beyond specialized applications. The World Humanoid Robot Games, held in Beijing last August, showcased the current capabilities of humanoid robots through athletic demonstrations such as sprinting, weightlifting, and kickboxing, though success was not always consistent. AGIBOT is a prominent player in the burgeoning Chinese market for humanoid and quadruped robots. The company reported shipping 9,700 units in the first six months of the year, establishing it as the leading seller of humanoid robots according to Counterpoint Research data released in late August. The firm is also reportedly exploring a potential Initial Public Offering (IPO) on the Hong Kong Stock Exchange. Beyond headline-grabbing performances like robot dancing and boxing, the industry is actively seeking practical, real-world applications for its technology. Keenon Robotics, based in Shanghai, has developed robots designed to automate services within hotels. Wan Bin, Keenon's chief operating officer, highlighted a specific implementation where five of their robots are employed in a Shanghai hotel to perform duties such as guest reception, room service delivery, room cleaning, and restaurant floor management, a scenario he anticipates becoming increasingly prevalent. Other manufacturers are focusing their development efforts on industrial settings. Jianxin Pang, vice president and vice dean of research at UBTech, stated that the company has made substantial investments in developing robots for the industrial sector, with an initial focus on common scenarios that offer significant market potential. The advancement of embodied AI, which integrates AI models with physical robotic systems, is seen as the key driver for achieving this next phase of humanoid robot capability, moving them from demonstration to widespread practical use.