You’ve probably already seen the video of the stupid Russian AIbot falling flat on its face in its debut, but here it is again.
This is all a waste of money, time, and effort. Before you reply with that “what use is a baby?” line, there’s no there there — these aren’t aware, thinking machines, they all need pre-programmed tasks to follow in their slow, clumsy way. You wouldn’t want one with full autonomy, anyway, given how erratically AI performs in simple text tasks.
Near the end of the above video, they talk briefly about industrial automation, which is good, I’m all for it. Those are repetitive tasks in which you do give the machine a set of programmed operations, and you gotta wonder…what is the “humanoid” part for, anyway? Wouldn’t it be smarter to have just an arm, with specialized grasping elements?
This is just another example of hyping up AI, because a bunch of billionaires make more money by selling server farms and stolen information, but they need flashy stuff to convince the rubes to buy into it.
Also, some of these robots aren’t even independently controlled by AI — they’ve got a guy behind a screen diddling switches and joysticks, in which case they should just cut out the middle android.
Did it learn to walk by imitating drunk Russians?
Whenever I see people building fully humanoid robots, I keep remembering Anthony Boucher’s story, “Q.U.R.”, about a world where robots are all humanoid and are running into all sorts of cost and maintenance headaches. Then, two robot engineers realize that by stripping down robots to just the parts that they need to do their job, they can make them cheaper, more reliable, and actually more versatile.
When I first read this story, I thought it was unlikely to ever actually happen, because it is blindingly obvious that building a full-up duplicate of a human body with all those joints and actuators to do a simple job like sweeping the floor makes no sense when something like a Roomba could do the same thing at a fraction of the cost. But now, with guys like Musk getting all wound up pushing for full-humanoid robots to “do everything”, I’m not so sure.
Silvrhalide @ 1. Yes, I saw this on the BBC and immediately thought ‘Vodka’.
I’m okay with roboticists working on the difficult problem of bipedal locomotion just to figure out how it can be done. Just don’t embarrass yourself by showing a project before it’s fully baked and ready to demo. Unfortunately, the people with money want their robot butler now, Now, NOW!
Even if you’re building a general purpose robot, there’s no point to making it bipedal. A setup with four wheels each on the end of a leg with robust and versatile joints allows for far more rapid motion on a flat surface and much greater stability on pretty much any surface that a bipedal design could handle, all whilst reducing the computational load of keeping balanced on two legs as it’s so much easier to simply keep the centre of mass firmly inside the shape delineated by the legs touching the floor, so small pertutbations won’t cause a loss of equilibrium.
Let’s design a thing to be like this other thing that came about through random mutations.
AI continues to prove it is Artificial but not Intelligent. I want to know what arrogant persons decided that trying to create a superior electronic intelligence would use ‘human intelligence’ as the model to teach it. Look around you, too many humans are dumb as a box of rocks.
Woman Marries ChatGPT-Generated Groom Through Breach in the Metaverse
This seems SO very pertinent here:
https://www.smbc-comics.com/comics/1763094803-20251114.png
Our built environment and our tools and our machinery etc is all designed for humans.
A general purpose humaniform robot can in principle do all that.
Also, the AI is one component of the robot, not the robot.
Silly to conflate chatGPT with a robot.
(Obviously, it’s aspirational at this time, but the technology will keep marching on)
I really thought the Russians were light-years ahead of anyone else in the development of this sort of technology.
I mean, look at Melania.
@10 John Morales made an important distinction. The software is not the hardware. Except for the melania robot which is automaton crapware.
In the news: https://www.vox.com/future-perfect/468492/artificial-intelligence-celebrity-voice-cloning-synthesize-michael-caine
↓
[pullquote]
As weird as it is to imagine an AI speaking with Caine’s inimitable Cockney accent — and frankly, no AI can do Caine better than these two guys in The Trip — at least he made the active choice while alive to sign it away. But ElevenLabs has also struck estate deals that let users hear narration in the voices of historical figures, like Judy Garland, James Dean, Maya Angelou, and AI pioneer Alan Turing. (That’s right, the genius who once said that if computers became smarter than humans, “we should, as a species, feel greatly humbled,” will now be lending his posthumous voice to the machines.)
The human body plan is designed to solve the problem of what if you start with a tube that eats on one side and shits out the other end, then bit by bit add complications. Eventually you get a quadruped, and you try to figure out how to climb trees. And then you decide fuck trees, let’s walk around the savanna. And then you start building stuff.
It seems somewhat unlikely that this is the optimal body plan.
As for the built environment: we modify it all the time to take machines into account. If you can provide an actual service to actual people, then we’ll replace the tree branches with an elevator, turn a rocky path into a nice cobblestone road, dig a canal, etc.
What do you mean by “stolen information”? Who was it stolen from? Is the original owner no longer able to use it?
I just saw this on Bluesky.
They have a point.
How long is it going to take before those AIs object to…forced labor.
“As for the built environment: we modify it all the time to take machines into account.”
Ahem, that’s kinda the point. General purpose robots that can operate in the exiting environment.
Easier to modify machines than the entire environment.
(Think stairs; replace them all with ramps? Elevators?)
‘How long is it going to take before those AIs object to…forced labor.’
Forever. Current AIs are not actually sentient nor sapient nor volitional nor emotive.
(Future ones, well, that’s to be seen)
—
Also, it’s kinda stupid to imagine doing what one is built to do constitutes ‘forced labor’.
The first company that can make a robot that can safely handle changing diapers will become the wealthiest company in the world.
cheerfulcharlie, or sexbots.
Ivan the Robot realsied he was in Russia, and collapsed in despair.
For those who might want an informed, if lengthy opinion on the hype and technical challenges around humanoid robots, here’s a relatively recent blog post by Rodney Brooks:
Why Today’s Humanoids Won’t Learn Dexterity
He talks of the challenges that need to be overcome to make a fully humanoid robot that could operate in human environments with roughly human-level competence, how these challenges are not easily amenable to the same approaches that work for LLMs and GenAI, and as hype greatly oupaces the progress (that is being made after all), the very definition of what it means for a robot to be humanoid will change and machines on wheels with specialized manipulators will be declared “humanoid” even if it is not at all what is being hyped now. Somewhat like a flying car used to mean a vehicle that can be used in all the ways that a car can be used, but also can fly, and now some projects claim the title of the flying car while being a form of urban helicopter entirely unfit for movement on the ground.
And also we are very far away from any humanoid robots that could be mass adopted.
I have a question: a lot of the robots in the video seem to fall down, but we see no footage of any of them getting up. Usually humans run over to put them back on their feet, the robots just seem to jerk spasmodically without any idea of what to do.
So…..can the robots get up? Is it too difficult? Wlll we have to get them those alarms for the neck or wrist that people use when they’ve fallen and can’t get up? A servant that we’ll always have to be helping to its feet? I can just see employees running up and down the factory line all day, their sole jobs being to get the robots back up.
I must point out that one-year-olds have this figured out: they put their palms on the ground in front of them and their legs under themselves, and lean on their arms to push themselves up. That doesn’t seem difficult, but then, they have inner ears that know how to balance.
My father, a human, figured it out once: he fell in the back yard and couldn’t get up himself, so he rolled over to the wall of the house and used it to lean on while he successfully stood up. The robots too stupid for that?
Really, I want to know: can they get up when they fall?
Another, but I won’t embed this one: https://www.youtube.com/watch?v=MO-bvndVIxA&t=45s
(The other one is 2 years old, this one 4 months old)
John, you found it! Thanks.
The one bot was awful slow and awkward, but the one on the field was quick, because it used the exact one-year-old’s method.
Well, at least now we know.
I can just see employees running up and down the factory line all day, their sole jobs being to get the robots back up.
You mean the robots who took those human employees’ jobs? I can see human employees first being told they had to train their robot replacements, and then being kept on after all, just to help their replacements walk to and from their workstations. Oh well, at least they wouldn’t have to help the robots go to the bathroom…
There is a joke is Stargate, in which O’Neil comments, while on a ship, “Why don’t you put seat belts in these things?”, to which the To’kra answers, “We prefer to just not crash.” I am reminded of this by the fact that Boston Dynamics, and others, already have freaking robots that are bipedal, which don’t claim to be “AI”, just to feed into the stupid AI insanity that we are having, which not only run, jump, walk, etc., but, “don’t tend to fall over, or if they do, can stand themselves back up.” So… our answer to this would be, I guess, “We prefer they don’t fall over in the first place.”? lol
the current ideas about “robots” seem to be making a humanoid is a major goal. It kind of reminds me of all the inventiveness surrounding flight. People were trying to make things that worked like a bird wing enables a bird to fly by flapping wings until they tried a different approach. I think that flapping wings flight has been achieved with small very light mechanisms, but flying is dominated by motors spinning fans mostly.
where the most advanced craft seldom could be mistaken for a bird.
I always thought of the “ornithopter” in Dune and all the “walkers” in Star Wars were exceptionally rediculous
re: ‘the current ideas about “robots” seem to be making a humanoid is a major goal.’
That was always the idea, from the very beginning.
It’s only when industrial machines began being called ‘robots’ that the idea they could just be anything and still be called a robot took hold.
(that’s what ‘bot’ stands for, as you doubtless know, as in ‘chatbot’)
—
So, now it’s more like ‘autonomous machine’, whether physical or virtual.
—
Also:
there was very little technology until recently that had the remote possibility of replicating a functioning humanoid robot beyond a marionette. same goes for flying machine replicating bird flight. making a device with that level of functionality is the problem not its appearance. That is what a couple bike mechanics realized it is flying not birds.
might be more practical to ignore the form at the start and focus on function is the simplest basics to match tasks and expected environment.
though big spiders or insects like robots might be a little unsettling for many. they could be pink and covered with “long soft fur” for comfort.
“there was very little technology until recently that had the remote possibility of replicating a functioning humanoid robot beyond a marionette.”
True. But the infosphere was saturated with what a ‘robot’ was, and it was always humaniform.
That it need no longer be so does not dispute that.
And the point is that when you wrote ‘the current ideas about “robots” seem to be making a humanoid is a major goal.’ you intimated it had not been so hitherto, and I corrected that. No more. Not personal.
But sure, autonomous uncrewed vehicles can be generically considered ‘robots’.
—
Also, for amusement, I offer this (because I myself use ‘uncrewed’ instead of ‘unmanned’):
https://unidir.org/files/2022-11/UNIDIR_Uncrewed_Ground_Systems_Primer.pdf
↓
Background
The U.S. Department of Defense (DOD) develops, acquires, and fields uncrewed aircraft systems (UAS, commonly known as drones) of varying sizes and capabilities. DOD has organized uncrewed—DOD uses the terms uncrewed and unmanned interchangeably—aircraft into different categories of aircraft to facilitate a common understanding of UAS within DOD and the integration of military UAS into the National Airspace System, and to communicate its UAS requirements to Congress. Congress may consider whether DOD’s UAS categories remain a useful basis for its oversight of DOD’s stated requirements and objectives for UAS programs.
—
(It will be interesting if that ongoing ‘woke’ transition is noticed by the current admin, I reckon)
I said that because I do not know all the history of thinking about “robots ” I do see the interest now represented by the posted video. I think that it might be more fruitful to focus on function instead of trying fit the technology into some familiar form but focus on the functionality we are trying to achieve
All true, but a humanoid robot that can competently and safely operate in spaces designed for humans has clear benefits too. With all existing mass-adopted robots to accomodate their technological limitations you either need to create a very specific environment, like an assembly line, or simply accept the limitations. Roomba is useful, but doesn’t navigate the stairs.
Setting aside for a moment, whether we can create such a robot in the near future, there are enough unpleasant and dangerous jobs that could be replaced with robots for everyone’s benefit. Of course, we could find other ways to either eliminate such jobs, or make them less dangerous. But there is an undeniable appeal in not needing to redesign the whole tool stack and just have a machine that fits right in with all the accumulated quirks of existing ways of dealing with the problem.
Of course, the capitall class doesn’t think of making dangerous and unpleasant jobs less so, even if they can approve of such notions in their PR talking points.What they wish is to replace all jobs with robots to avoid dealing with volitile and opinionated workforce
And after the relative successes of LLMs and GenAI, they think they are close to having their wish granted. Just through enough data at enough computaitonal resources and the robot will emerge. Hence the hype.
Now, going back to the feasibility of building a human replacement robot soon, the idea that GenAI were simply a result of throwing a ton of data into a big enough generic neural network with little to no preprocessing or domain-specific knowledge is profoundly uninformed. Just as one example, LLMs don’t learn the raw bits of text. The text first has to be broken down into pieces (tokens) which are then encoded aa vectors in a multidimensional space that preserve some relationship between the original pieces. This processing is not a part of the learning process and rather a result of decades of research into efficient encoding of text coming from the field of natural language processing. And we currently lack understanding of not only how to efficienly encode human movements for the machine to learn, but even precisely what kind of data we should encode. We are a couple of technological breakthroughs away from even trying to replicate the “success” of genAI in robotics.
And there’s a good argument to be made that genAI level of competence is not good enough for human-sized robots meant to operate alongside humans.