If you haven’t seen any of the videos, humanoid robots are getting impressive:

That was Atlas 2 (by Boston Dynamics). Here’s the previous version showing off:

On a lark I asked Grok what the maximum lifting capacity of Atlas is, and though the information isn’t public, it guessed it could dead lift around 80 pounds. It suggested that the all-electric version, the Atlas 2, is stronger, and might be able to dead lift perhaps 100 pounds. (To be clear, Grok, being a large language model, is just producing text which is the most likely thing the people in its training data would have written, which works as a reasonable-enough summary of what actual people have written on the subject as long as nothing important depends on the accuracy of the facts. But since we’re discussing humanoid robots, I’m fine with just using Grok’s “guesses”.)

Also on a whim, I asked how far it can walk and Grok’s best guess is around three miles, perhaps a little more, before running out of power. On flat ground without obstacles.

The Atlas and Atlas 2 are generally the best bipedal robots available, at least as far as mobility goes. There is also Tesla’s Optimus robot, which is more designed around dexterously performing repetitive tasks than raw performance, but interestingly Grok guesstimated that it might be able to walk for 5-15 miles, with a lot of it-depends and caveats.

Actually, there are a lot of it-depends and caveats on both, because neither company has released official specifications. So in both cases, these might well be over-estimates.

To put this into perspective, my fifteen year old son has deadlifted 135 pounds and my best deadlift is 440 pounds.

I find the weakness of these robots a bit amusing because science fiction has tended to portray humanoid robots as effortlessly stronger than human beings to the point where we might as well be insects compared to them. Granted, the Boston Dynamics and Tesla robots are still relatively early on in technological development, but they are being developed in the early twenty first century, not the late eighteenth. These are already the beneficiaries of enormous technological progress in motors, pistons, advanced materials, and batteries. These things will all get better, but it’s going to take a long time and will probably cost even more. And I’m not sure how much greater efficiency there is to eek out of electric motors; they’re already quite efficient and there’s only one permanent magnet material known which is stronger than the neodymium-based magnets used in the best motors of today. That’s iron nitride, and at it’s strongest it’s only about twice as strong a magnet as neodymium. Which would only result in motors that are a few percent more energy efficient; the main benefit is that they can be smaller and lighter. And right now it’s unclear that iron nitride can be gotten to work in quantities large enough to see with the naked eye. Don’t get me wrong, humanoid robots will continue to improve and they may well get quite a lot stronger.

But they can get four times stronger and still be only about as strong as me. And there are a great many people much stronger than I am.