The whole video is worth a watch, but here are some key points I took away from it:
- Robots from Boston Dynamics have already been left in the dust as they rely on antiquated coding and software structures. These robots can’t learn in the same way computers/robots relying on modern neural nets can learn.
- Robots from Tesla and startups like Figure AI have quickly surpassed those outdated robots and are already humanoids much more similar to humans in movement, fluidity, the breadth of tasks they can conduct — and overall usability. And they’re just getting started!
- Large language models are making it much easier to train and improve these advanced robots.
- Improvements in cost and capability of electric motors, batteries, and other hardware are key to making humanoid robots more viable beyond super niche applications. A usable and practical humanoid wasn’t close to viability 10 or even 5 years ago. Ongoing experience curves in these fields help more and more to make such robots viable.
- The first commercially useful applications of such robots will be in factories.
- With Tesla’s manufacturing needs and expertise, ability to quickly train large numbers of robots, and experience in integrating advanced hardware and software, no other company can genuinely outcompete Tesla in terms of scaling up testing and eventual utilization of such robots.
- Tesla is probably already training robots on some production lines and could well start using them for commercial purposes (to save money) in about a year.
- Tesla can then move to selling Optimus bots to other manufacturing businesses at a scale no other company will be able to touch.
https://youtu.be/T2hARSyv8OE
and
( Comment: This is simply fascinating. My years of making space robots (Canadarm) in the 1975-1984 was state-of-the-art at the time. But this advancement is simply staggering in just a few years. Art Hunter)