Unitree humanoid robots have great price point and dexterity, but like all other humanoid robots, it's really struggling with the "smarts".
I don't have high hopes for useful humanoid robots because we can barely managed even self driving cars. If something that have very predictable goals and objectives (Go from A to B, follow traffic rules) can be so difficult. A robot that must interpret vague/fuzzy commands will be that much more difficult.
I think the key problem with self-driving is that it has interact in an extremely complex environment.
IMO, the first robots will be similar to Roombas, they'll be working in an environment that's pretty controlled, with limited interaction with humans, or interaction with humans that will work with the machine rather than demand the machine work around them. IMO, some of the use cases for robots are actually in less developed countries.
1. Highly repetitive work for subsistence or small farmers.
2. Basic functionality in people's households.
3. Factory labor.
Humanoid or semi-humanoid robots can fulfill that middle-layer that balances that a dedicated machine can do versus the ultimate multi-functionality of a human laborer. Though on that last point, having been a warehouse manager I have to tell you, I think even dumb robots are better than the guys I had to work with sometimes.
We will eventually have hardware that's cheap enough and an AI model that's small/efficient enough to run on local robot hardware. When that "ChatGPT" moment is hit, China will be the one with the actual economics to make such robots economically viable.
