If you were impressed by Tesla's humanoid

3 points by robomartin ↗ HN
Bipedal locomotion research from over twenty years ago:

https://www.youtube.com/watch?v=e2Q2Lx8O6Cg

Here's a more modern variant:

https://www.youtube.com/watch?v=rhu2xNIpgDE&t=26s

9 comments

[ 1.6 ms ] story [ 30.8 ms ] thread
Yes, because Tesla is working on a robot that costs can be bought.

Lots of people say the Honda robot did this before. Now try buying one, you can't because they did not make more then the demo units.

Others will point to the Boston Dynamics robots, great those are available to buy. Oops, forgot the price!

Tesla is aimming for a 20K robot. However, even if it is 30K, 40K, even 50K it will still be way cheaper that the other machines people point at.

> Now try buying one, you can't because they did not make more then the demo units.

I mean, yes, but try buying a Semi or a Cybertruck. My guess is the 20k robot will join that illustrous group.

I'm sure you can buy one.

Getting delivery is another thing of course.

> Tesla is aimming for a 20K

Yeah, well, good number to use for marketing. Outside of that domain, this is an absolute fantasy.

Think of it this way: A humanoid is easily one order of magnitude, if not two, more complex than a motorcycle. It's certainly at least two orders of magnitude more difficult and capital intensive to design. A new motorcycle sells for $15K to $20K. Using that as a ROM (rough order of magnitude), you'd expect a reasonably capable humanoid to go for more than $150K at a minimum.

My guess is you are looking at $250K to $500K as the baseline range for a working humanoid for the next ten to twenty years. I can't see them becoming capable enough to achieve mass production scales for a decade or two. At that point you might be able to start approaching manufacturing scales in the thousands of units per month.

When you can open the box, turn it it on, tell it to go clean the house, take the dog for a walk and cook dinner, well, at that point we are talking millions of units per year, tens of millions maybe. That's a long way out.

I really just don't understand the hype over humanoids, I understand the assumption that humanoids can do the same tasks as humans do but at this level of technology it's just too soon to have them do anything useful.

We yet have to achieve proper autonomy for simple tasks using robotic arms like picking items. I still have to prevent my roomba from avoiding socks. Those are machines built specifically for that single purpose and yet they are failing at it.

I just don't see how humanoids can be of any use at this stage apart from generating hype for Tesla or for some sci-fi enthusiast.

> I just don't see how humanoids can be of any use at this stage

Today, outside research environments, they are useless. We are still worried about them falling on their face, much less take the dog for a walk and cook you a couple of eggs. Generally useful humanoids are probably decades away.

I'd love to get my hands on one. My first use would be to sort, catalog, and shelve my books.
Pay a high school kid $15 per hour to do this for you. Nice little gig for a kid and you get to save tens a massive amount of money and aggravation when compared to a humanoid robot, which, today, is far more likely to be a total failure for your task. In addition to that, it isn't going to take 400 hours of your time to train and supervise.
It's the 400 hours I'm looking forward to.