Select Language:
Industry experts at the ongoing Asia Forum are discussing when humanoid robots will experience their “ChatGPT moment,” with projections ranging from two years to up to ten years. This milestone refers to the point when humanoid robots can consistently follow human instructions in unfamiliar settings. Achieving this would open the door to mass commercial use of these robots.
Expecting significant advancements within the next five to ten years, one developer of humanoid robots and embodied AI notes that the main challenge lies in limited model generalization, which currently hampers the robots’ ability to perform reliably across various environments.
Another industry leader is more optimistic, suggesting that the robotics field could reach this pivotal point once interaction data totals hundreds of millions of hours. Currently, about 100,000 hours of interaction data have been collected, but with new data gathering methods being explored, this could reach 10 million hours by 2027 and then rapidly expand into the hundreds of millions. This target could be met within the next two years.
In contrast, a chief scientist at a Chinese tech company’s robotics lab believes the breakthrough might still be a decade away. He points out that, since the rise of deep learning in 2012 and the surge of ChatGPT in 2022, a decade has passed. The success of ChatGPT was largely due to finding an affordable way to gather vast amounts of data, a process that the robotics industry still needs time to replicate on a large scale.
Despite differing timelines, all industry experts agree that humanoid robots are poised to grow their presence in industrial and specialized applications well before such a breakthrough in general-purpose robots is achieved. Industrial environments tend to be highly standardized, making it easier to train robots to perform key tasks once sufficient data and computing power are available, leading to scalable replication across thousands or even hundreds of thousands of similar roles.
However, deploying robots in household settings may be the “final step” for embodied intelligence due to the complexity and high capability demands. Whether this can occur within the next decade remains uncertain. Industrial and hazardous work environments, which are more uniform and predictable, are better suited for early adoption of humanoid robots, according to experts.
ChatGPT
Perplexity
Gemini AI
Grok AI
