Personally, I feel that the prospects of Robots industry are not optimistic. It’s not about robots themselves; the robot industry can only maintain its current popularity if there are advancements in AI or AGI technology. In other words, whether the robot industry thrives or not depends not on robot technology, but on AI technology. The current surge in the robot industry, represented by humanoid robots, is abnormal and irrational. Moreover, considering that robots are competing with humans, this industry might already be a red ocean. The emergence of humanoid robots hasn’t solved the biggest challenge in large-scale robot deployment: true (real-world) intelligence. In traditional robotics, the common requirement for robots is that if I tell a robot a location, it should be able to go there on its own and perform predefined tasks. For example, in the case of an AGV (Automated Guided Vehicle) used in factory transport: engineers define a task for the AGV robot: go to the stacking area to pick up items, go to the delivery area to drop off items, and then return to the stacking area to pick up more items. 1. Localization: In this task, just like humans, the robot first needs to know where the stacking area, delivery area, and the robot’s current position are, so it can determine the direction it should go and estimate how far it needs to travel. 2. Planning: Once the position is known, the robot needs to figure out how to proceed. For example, it needs to know that it should move forward at approximately 1m/s, travel about 100m, and then turn left to go 50m in order to reach the delivery area. 3. Locomotion: Once the planned path is established, the robot needs to move according to this plan. For example, a wheeled robot must be able to drive its wheels and follow the planned speed and path, a task known as locomotion. 4. Skill: Actions like picking up and dropping off items can be considered the robot's skills. For robotic arms, this is known as manipulation. Robots have developed to the point where solutions for localization, planning, and locomotion are relatively mature. Technologies like SLAM, graph search, trajectory optimization, and model-based control can complete these tasks in a fairly reliable way. In recent years, with the development of neural networks, end-to-end algorithms have also become a research focus. However, the problems or tasks addressed by these solutions are still primarily these three: localization, planning, and locomotion. As for robot skills (skill/manipulation), basic actions like grasping can now be mostly achieved, such as predefined programs to pick up a cup, etc. Specific algorithms include model-based control, imitation learning, visual serving, and so on. However, for some skills—like a robot playing soccer or grabbing a bowl—it remains a research hotspot and a challenge for real-world implementation. This is the reason why robots are struggling to scale up on the consumer side: their level of intelligence is not sufficient for them to "understand" their environment in the same way humans do. They can only perform simple "recognition" of the scene. Therefore, on the one hand, robots need to minimize their interactions with the environment; on the other hand, because robots cannot understand the environment, the tasks within the environment, or even their own actions, they find it difficult to automatically select the appropriate skill to complete different tasks, meaning they lack generalization across tasks. These two points significantly limit the possible application scenarios for robots."