Boston Dynamics' humanoid robot can now pick up and toss heavy objects at people, jump around on scaffolding, and land a perfect backflip (video)

8 Likes

… and land a perfect backflip

Silly human! That was a front flip with a half twist.

3 Likes

“At least when they murder me, I’ll have a large bank balance.”

5 Likes

That tracks; it’s how we got flips phones.

2 Likes

But is it a Pusher Robot, or is it a Shover Robot?

Is that true, though? Your average back yard critter probably has a brain with hundreds of millions of neurons, honed over hundreds of millions of years of evolution under life-or-death stakes to perform fast multi-sensory signal processing, integration, decision making and motor coordination. This is largely single-purpose neural circuitry that cannot be reprogrammed for abstract thought, so it’s not “smart”, but there’s a lot of processing power involved.

This actually brings up an interesting point. The Boston Dynamics development of autonomously operating robots has been running parallel to more abstract AI development concerned with understanding and producing text and images. The former we think of as very basic and common to all animal life, while the latter has up to now been the province of human “high intelligence”. But looking at the results so far, its not clear that the first task is easier than the second: I’d say we’ll have an AI that can pass a high school final exam sooner than we’ll have a robot that could do a high school student’s temp summer job at a construction site.

The trick is to understand that a lot things for which we employ our abstract-thinking big brains, like performing mathematical calculations, writing formulaic school essays or drawing portraits, are actually pretty simple when broken down into logical parts and considered abstractly. The big brains are needed to ensure our survival as individuals, communities and civilizations which then make “writing an essay” a potentially worthwhile endeavour, but the actual writing is easy. You could probably encode the essay-writing faculties of ChatGPT into a wet neural network that would fit in a squirrel’s brain - for no benefit to the squirrel, obviously.

5 Likes

When do they program them to flip burgers, that’s the biggest fear of my local Facebook group.

That’s a cool video, I hope I live long enough to see how it plays out.

You think the robots will take over within 15 - 20 years?

Or at least fix/build everything on the list my wife gave me for this year.

3 Likes

Hey, those squirrels doing the novel obstacle courses in Mark Rober’s back yard seem to do ok with new challenges. I would say that their little brains are more flexible than you give them credit for.

5 Likes

Oh, certainly. I meant it more in the sense that a squirrel can’t decide to give up on its acrobatic skills and use that freed-up brainpower to learn a language or something. (Nor can humans, of course - motor circuitry is always motor circuitry.)

3 Likes

Might be a Bender
image

1 Like

Boston Dynamics’ humanoid robot can now pick up and toss heavy objects at people, jump around on scaffolding, and land a perfect backflip

… for three minutes until the power runs out, or as long as the tether can reach and not be in the way.

As impressive as Boston Dynamics is, nobody has solved the portable power problem for these kinds of robots. The stunts they are doing require enormous amounts of juice and batteries are nowhere close to good enough to supply this level of activity for more than a couple of minutes. That’s why most BD demonstrations are on power tethers, and why the first Big Dog ran a gas engine to make electricity (and the heavy service ones still do).

So those of you worried about a robot attack, just wait a minute and they’ll all keel over. Then have yourself a 1200 calorie cheeseburger to celebrate the wonder of energy efficiency and conversion that is the animal body.

11 Likes

You seem to be assuming that humanity wouldn’t be wiped out within that first minute or two. Doesn’t take long for an angry robot to perforate a squishy meat bag, and they’re unlikely to give us advanced warning. :wink:

4 Likes

I’m not convinced robots could find all of humanity within the first minute or two. We have some very clever disguises, sometimes even by accident.

Metal Gear cardboard box

6 Likes

Heh, heh. Such a great movie.

1 Like

… I’m sure they’re working on that

6 Likes

Haven’t checked in with BB for a while, seems like the general vibe about technologic advance has changed here?
I’m still more afraid of humans.

5 Likes

Then you should be very afraid of humans with more powerful weapons at their disposal.

6 Likes

Yeah, good thing robots will never, ever be equipped with infrared sensors that can detect human bodies underneath a cardboard box, or AI that can detect unusual movement in objects that are large enough to harbor a human.

2 Likes

That’s why all robots, AIs, and MLs must be equipped with a Big red Switch, readily accessible to humans but inaccessible to robots, that cuts all motive power. Not a dry contact like on a ATX motherboard (that can ve ignored by clever robots), a real set of contacts on the incoming electrical power. Like the one on the back right side of the original IBM XT.

If I could edit the above down to one simple sentence, I would propose it as Law 0 of Asimov’s Three Laws. Sadly my verbal diarrhea prevents this.

2 Likes