Midea's high-profile entry into the service robotics track with the launch of its new brand “WISHUG”

Tang Jian said that this is due to compared to industrial robots, household robots are faced with a more complex scene, which means that AI algorithms and models need to have a more powerful generalization and robustness, in order to quickly adapt to different application scenarios.

“To take a counterintuitive example, which action is more difficult, going to the refrigerator to get bread or assembling a car? For a human, the former is a very simple task, but for a robot, it is more difficult than assembling a car.”

Tang Jian further explains, “Although the steps of assembling a car are complex and may require hundreds or even thousands of steps, the operation process and objects are very single and fixed. On the other hand, home service robots face thousands of households, each with very different refrigerators and drawer doors. At present, there is basically no robot that can smoothly complete the action of fetching bread from the refrigerator.”

Can it become a breakthrough in AI industrialization?

Another major difficulty for household robots lies in human-robot interaction. “There are also challenges in how robots can communicate with people in a smooth and free manner, which involves voice interaction, as well as the understanding of natural language.” Tang Jian said.

Overall, the bottleneck faced by service robots for large-scale application is the general challenge of AI landing at the moment.

“If a family service robot, which can move very autonomously in the home, complete all kinds of household chores through mechanical arms and dexterous hands, and have a comfortable conversation with the user, it means that it has realized the leap from weak AI to strong AI. This is the direction that AI academia and industry are currently working on.”

Although the challenge is not small, Tang Jian believes that from the perspective of scale landing, products such as home service robots are expected to become the first breakthrough in the industrialization of AI.

“Simply using AI algorithms for commercialization is less sustainable, and the commercial value it brings is not enough to support the cost of human input. In my opinion, the right way to land AI should be a combination punch, i.e., software, hardware, and a cloud platform, to form a complete set of products or solutions, like products like home service robots, so as to create greater user value and commercial value, in order to achieve sustainable and high-speed development.”