this post was submitted on 24 Apr 2025
4 points (75.0% liked)
Futurology
2518 readers
138 users here now
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Humanoid robots, like all technologies, will be adopted on an s-curve. First, there will be just a few of them, and then rapidly they will be everywhere, as their adoption heads for market saturation.
Are humanoid robots ready for their s-curve take off phase? Seeing Xpeng's IRON humanoid in action might make you think they are. Xpeng say they expect to start mass-producing these next year, and say they are investing $13.8 billion to scale production.
IRON's specs look impressive. Xpeng says it operates at 3,000 TOPS of processing power with their Turing AI chip. For reference, Microsoft's baseline for an AI PC is 40 TOPS (Trillions of Operations Per Second).
I don't think there won't be mass adoption that soon. From what I've seen, fine motor skills are lacking, there isn't much training data to brute force it and other learning doesn't translate that well. The demos that show more are slow or remote controlled. Most not remotely controlled robots seem at the level of a trained chimp, so they can go and fetch stuff, or store things. Give them a peeler, a pot, some vegetables and expecting them to prepare dinner is way too far for a robot. Making coffee or tea is kinda the highlight, and it's pretty expensive for a coffee machine.
Robots can be useful for places that need some waiters, but that seems early take-off phase. But once fine motor skills and learning by watching a few videos are sorted out, then they'll take off. That's being worked on. Hell, it was announced this week they got it working in the lab. Once that goes into production models, things will move rapidly.