Hacker News Reader

> you can realize they're just large puppets. We've had dancing puppets like Walt Disney's Small World.

This is hugely understating just how difficult it is for a biped/quadruped robot to move around in the world without continuously falling, getting stuck, etc. Like seriously, seriously underestimates it.

Watch this video to see spot walk over some terrain that even people struggle with: https://www.youtube.com/watch?v=k7s1sr4JdlI

Doing it without a tether for a reasonable timeframe (30+ minutes) is insane! Disney animatronics don't even come close to the complexity of these robots (even though what Disney has done is for sure impressive).

> robots still generally can't interact with humans or the environment in an unstructured way.

I think this is pretty short sighted, and you're going to have your mind changed quite rapidly in the next couple of years. This style of robot has definitely hit a threshold of price and usefulness, not unlike what happened for drones just ten years ago.

I could be wrong, but I hope I'm not! It really feels like things are moving at a crazy speed in the robotics field right now.

>> robots still generally can't interact with humans or the environment in an unstructured way.

> I think this is pretty short sighted, and you're going to have your mind changed quite rapidly in the next couple of years.

Oh, I'd love to see that change. That would be a change in reality, not perception. It wouldn't force me to change my belief that as things stand now, walking have been an ongoing failure and disappointment.. See:

Edit: Also, I should admit I'm discounting the serious engineering challenge of just getting robots to walk on uneven ground by itself. Doing seeming simple stuff like has where the progress of idk 30 years has appeared. But anything more than those ultra-simple things really wasn't happened. I'd still stick with 90% of impressive is puppeting.

https://en.wikipedia.org/wiki/DARPA_Robotics_Challenge

Most people think like you do. But then again most people are blindsided by revolutionary change.

The robotics field has been moving at a crazy speed for decades.

I think people, probably not yourself, underestimate the complexity of things like bipedal movement. This display was really quite impressive.

> Watch this video to see spot walk over some terrain that even people struggle with: https://www.youtube.com/watch?v=k7s1sr4JdlI

Where's the terrain humans struggle with? At the most extreme it climbs a small mound of rocks and gets stuck.

It's cool we can build something that can do that semi-autonomously but it's lightyears from being as capable as anything biological.

Indeed, people struggle with? Adam Savage (53 years old) navigates that pile of rocks without problems, while controlling the robot. Never even needing to use his hands.

The work Boston Dynamics does is extremely impressive, but they've been at it since 1992. Each year they make small and incremental changes. But its still a century away from a human, dog, or cat. Not a revolution, just slow and steady evolution of knowledge, software, and hardware.

You and I are around different people! Yes, that pile of rocks isn't too big of a deal for young and agile people, but try to imagine the 10th percentile of people walking over that. Older, not as fit. That rock pile would be a struggle.

I don't disagree that it's still far away from a human in good shape, but there are certainly a significant percentage of people who wouldn't be able to walk over that, or would slip a few times while doing it.

That sort of terrain is also no joke if you have to traverse it for large stretches, even for athletic people. It's very similar to mildly rough mountaineering areas and that will wear you down quickly!

Comparing the best robot to the worst human isn't very interesting.

>Part of the reason the robots evoke little excitement is because after a short look, you can realize they're just large puppets.

I think it's more like the Feynman example of how a scientist may look at a flower.

A person may see a full-size puppet when they see a choreographed robot, but all I can think about is how complicated the mechanisms -- software and hardware -- must be in order to dynamically balance a robot while maintaining whatever timing is called out for the choreography work; and the person-hours that such work must have consumed.

It's amazing to me they see a puppet. Do they call a car a puppet? It's controlled by a human. The robot spot is hardly what I'd call a puppet, for instance.

I got a dog a year ago and one of the things I'm most impressed with is how well he understands context and adapts to new situations.

Granted, the first time I took him to a rocky riverside all four paws fell into cracks between the boulders like some early Boston Dynamics prototype (he's from Texas and I guess never encountered terrain like that before) but he's a pro now. And he's mastered "soft human interaction" right out of the box (is just amazing with toddlers).

Seeing all this firsthand makes me appreciate how fine-tuned the product of evolution is and how much work must go into achieving basic behaviors we take for granted.

I adopted a then-8-week-old kitten less than two months ago, and it's been an eye-opening delight watching him adapt just to things in my home.

During his first week he managed to hop onto a window sill, and then fell off when he tried to turn around to walk along it the other way. Less than a couple weeks later he was effortlessly walking along the edge of a pillow stood on its side, a much narrower, unstable surface.

It's amazing to me that a tiny creature like that can learn to adapt in that way with so few days of life under his belt.

It is! Of course we animals are standing on the shoulders of hundreds of millions of years of evolution's fine-tuning. But to see it in action like this is incredible. Or a child from ages 2 or 3 to ages 4 or 5, is night a day. At 2 or 3, they're drowning risks in the tub. At 4 or 5 they're playing mario kart with you.

> It's amazing to me that a tiny creature like that can learn to adapt in that way with so few days of life under his belt.

I work at a raptor conservancy. The young birds can fly as soon as their wings/muscles are suitably developed, but learning to master the air takes a lot longer. E.g. they initially fail downwind landings on a gusty day.

> I watched radio controlled boats go around lakes in 1970s.

Tesla did this in 1898.

https://www.engadget.com/2014-01-19-nikola-teslas-remote-con...

It could also just be that it took 100 takes to shoot this little video. You might stop to watch the first 10 times, but the third day, it's become routine.

Great comment! So many, at least from my reckoning online, seem to assume complex AI is just a given and we'll have it soon enough, but the reality is that these things are mostly puppets and the kinds of brains you need to come up with a clever song and dance on your own as opposed to having one carefully programmed into you is night and day. I think there's a reason we don't have mini, less capable, robots like these in our homes. We have the bodies, sorta, but we just don't have the brains.

Dancing just isn't impressive and like you say, seems a lot like a higher tech version of Disneyworld. What I want to see is one of these, on its own accord, run into a burning building and save a child. Choreography just isn't impressive outside of the 'wow' factor. Without advanced AI brains, these bodies are almost useless shells and I imagine Google getting out of this space may have had something to do with that.

Exactly. Popular perception after seeing a video like this is to project capabilities into the robot that just aren't there. Talk to the lay public, people!

People assume this is actual AI, that it could just walk into an unknown new kitchen and do useful tasks like get dirty dishes in the dishwasher or other general stuff.

The other thing is a conflation of this kind of robotics with deep learning. Most of the work by Boston Dynamics uses no fancy machine learning. It is "just" (for experts it's not a "just", but a "just" for the public) electrical and mechanical and control engineering plus lots of specifically programmed behavior.

Now, it is impressive sure, but the humanoid form makes laypeople think there is more to it than there is.

Well, technically some AI programs are already pretty good at creating new songs and choreography. Google 'AI choreography' and 'AI songwriter'.

> We've had dancing puppets like Walt Disney's Small World.

With all due respect to Disney's Imagineers and their technical accomplishments, you can't really compare Disney animatronics that are literally bolted to the floor to what's being demonstrated here.

You can compare the animatronic presidents etc. to various industrial robots that are also fixed in place if you like, but the closest Disney equivalent to Atlas etc. are the 'stuntronics' first demonstrated in 2018: https://techcrunch.com/2018/06/28/disney-imagineering-has-cr...

A perfect puppet is a necessary but insufficient condition for creating the magical robots we were promised as kids. So it's progress.