What I call the “scaling law” of embodied intelligence is just the beginning. Stephanie Zhan I’m very optimistic about this. I’m curious, what are you personally most excited to see when we get to that tipping point? What industry, application, or use do you think will completely change the world of robotics today? Jim Fan Yes. Actually, there are a few reasons why we’re choosing humanoid robots as our main research direction. One reason is that the world is designed around the human form, and all of our restaurants, factories, hospitals, and our equipment and tools are designed for humans and their hand shapes. So, in principle, a sufficiently advanced humanoid robot hardware should be able to perform any task that an ordinary human can perform. While current humanoid hardware is not at that level yet, I believe that in the next two to three years, the humanoid hardware ecosystem will mature.
hardware, the question will be how to provide australia phone numbers these humanoid robots with AI brains. Once we have a universal basic model that can accept instructions in any language and perform any task that humans can do, we can unlock a lot of economic value. For example, we can have robots in our homes to help us with housework, laundry, dishes, cooking, or taking care of our pets. We can also use them in restaurants, hospitals, and factories to help with various human tasks. I hope that will happen in the next ten years. But as I mentioned earlier, this is not just a technical issue, it also involves many non-technical issues, and I'm looking forward to it. Sonia Huang Are there any other reasons why you decided to focus on humanoid robots? Jim Fan Yes.
There are also more practical reasons, such as training plans. There is a lot of data about people online, and the content in videos is mostly human-centered, with people doing everyday tasks or recreational activities. The humanoid form is closest to the human form, which means that the models we train with this data can be more easily transferred to the humanoid form than to other forms. For example, how many videos are there of robotic hands and grippers? It's rare. But we can see a lot of videos of people working with five-fingered hands. So training humanoid robots could be easier. Once we have these basic models, we can further specialize them and apply them to more specific robot forms like robotic arms. So we're looking for comprehensive solutions first. Stephanie Zhan Are you currently focused exclusively on humanoid robots, or will you also be working on robotic arms or robotic dogs? Jim Fan The Groot project is currently focused on humanoid robots.