First, if you’d really like to kill three hours, you can watch the whole AI day right here:
…but if you have other things going on in your life, we’ll just jump to the good roboty bits here. First off, people on Tesla’s robotics team gave an overview of the robot’s hardware and software, which all seems to be quite by-the-book for modern humanoid robot design. The first robot demonstrated was, as Tesla bot mechanical engineer told the audience before the robot came out, going to be operating for the first time without any backup support or umbilicals or support cranes, so I imagine the whole team was quite nervous. I’m not sure why you’d choose to have the very first time to try something like this be in an extremely public test, but, well, I admire their fuck it, let’s do it attitude here. And, to their credit, nothing terrible happened as the robot demonstrated that it could walk just like you walked as a teenager when you were sneaking back home, long past your curfew, slowly and carefully:
This is their development robot, and as such lacks a lot of the body cladding and refinement seen on the other robot they showed, which is supposed to be closer to what the production robot would look like. Even though the development robot didn’t do all that much, the more sleek-looking other robot did significantly less:
This sleeker robot just waved, wore a Don’t Mess With Texas belt buckle, and made everyone around it nervous, sort of like your uncle from Plano at a wedding where he’s had way too much to drink.
Looking back at what was shown, there absolutely are some impressive things. With only around a year of development time to go from skinny guy in robot Halloween costume to (even slowly) walking humanoid robot is a big deal! The mechanical engineering team appears to be capable, and is showing they have the ability build a humanoid robot of some kind. But was anything genuinely new or innovative seen at this demonstration? No, not really. You can’t look at these and not consider what else is going on in the humanoid robot world. Hyundai-owned Boston Dynamics, for example, has been showing wildly more impressive videos of their robots for years. This video of humanoid robots doing parkour is from last year, for example:
That’s a hell of a lot more impressive than tiptoeing around and waving. It’s also worth remembering that carmakers have been making humanoid robots for a long time, and have been walking and running and doing much more impressive things years ago. Honda’s Asimo is one of the best known of these, as you can see in this video from 2014:
…and Toyota had one, too:
…and let’s not forget that GM sent a legless humanoid robot into space :
I bring all of these up to give some context to what we saw. If you’re going to make a big deal about a humanoid robot, you can’t ignore what’s come walking out before. But, we also have to be fair here: the Boston Dynamics robots are expensive, hydraulically-powered beasts that use a lot of energy and as a result aren’t really practical for mass-production or long-duration use. And, those are the sorts of things Elon Musk is claiming that the Tesla robots will be: affordable (he said under $20,000, which, in humanoid robot terms is extremely affordable) and he said that these robots will be mass-produced and capable of doing a full day’s worth of work. Robotics experts that weighed in on the demonstration seemed to have generally the same take-aways: what was seen was, you know, fine. But that’s it. Here’s a good summation from Christian Hubicki, Director of the Optimal Robotics Lab at Florida State University:
First, the team did a good job. They came a long way in about a year(?), going from zero-to-robot from the ground up. Also, doing a live demo without a tether (safety catch rope) is braver than people know.https://t.co/XxC2BXQOoM — Christian Hubicki (@chubicki) October 1, 2022 Not blown away, not laughing, either. Seems reasonable. But the problem is that all of the crucial parts – the fact that the robots can be mass-produced, sold for under $20,000, can operate in a variety of environments and accomplish all sorts of tasks via Tesla’s powerful AI – is just talk, and nothing even close to any of these goals was demonstrated.
Sure, we saw some clips of the development robot doing some sorts of “work” (carrying a small box, watering plants) but even in these clips it was always with an umbilical (pointed out by the red arrows) and there’s no information regarding what exactly is happening in these clips, as far as the robot’s “brain” is concerned.
We’re just going on Elon’s word here, and, I have to be honest, his track record isn’t exactly great regarding these sorts of promises. There’s no $35,000 new Tesla, there’s no Roadster yet, there’s no Tesla Semi truck yet, there’s no real Hyperloop that can whisk you from city to city at incredible speeds, the Cybertruck isn’t available yet, and the promise of self-driving Robotaxis has been coming “next year” for the past nine years. I think his engineers and software developers are bright and hard-working, but that does not mean there’s any reason to believe any of the claims made during AI day, at least not without proof, of which there is currently none. If Elon got up on stage and said “by the end of this year, you can spend six figures on a robot that can slowly and gingerly walk through your house and wave” I would absolutely believe him, because that capability has been well demonstrated. But everything else? Based on Elon’s track record, I’d counsel no one to undertake any breath holding. Of course, that doesn’t deter Elon’s die-hard fans, who are if nothing else impressively robotic in their dedication to whatever the hell Elon says, responding to entirely valid criticisms with tweets like
— DCI Gene Hunt (@DeutzAre) October 1, 2022 I’m not sure what the fact that the robot “uses ML” (machine learning) really means when there’s zero actual evidence that it either does or that it somehow matters, at least not until it can be shown “using ML” to actually, you know, do something. If some random guy “thinks” it’ll actually cost around $25,000, based on a complex series of ex recto calculations, is that good enough for you?
— OptimusBot (@BogdanLazarec) October 1, 2022 It’s still incredible to me to see the absolute, unquestioning support Musk is able to get, and the trains of thought this support leads to. Like this, which starts with general excitement and enthusiasm and then leads to some really incredible extrapolation:
— Tesla Owners Silicon Valley (@teslaownersSV) October 1, 2022 So, Tesla robot is step one, step two is “…a future of abundance. There is no poverty.” I’m really going to need to see the work there, just something in that middle step that explains how buying $20,000 a pop robots that would replace a lot of human jobs is going to eliminate poverty. I mean, I’m all for it, I just wouldn’t mind having someone walk me through the details, you know, so I don’t seem stupid at parties when I’m being served canapés by one of these silver fellas. There’s so much more counting of robotic chickens long, long before robotic eggs will be hatching:
$500b/$20k=would be 25 MILLION bots. At a 30% margin, @elonmusk could create $150 BILLION in gross profit for $TSLA (assuming 1yr of labor replaced)#AIDay2022 pic.twitter.com/EdASA4yrO0 — Meet Kevin (@realMeetKevin) October 1, 2022 Replacing all factory jobs in America? That’s a lot to get out of a demonstration of watering plants, but, sure, maybe!
— TESLARATI (@Teslarati) October 1, 2022 Oh, even better, it’s only three to five years away! Someone remind me why anyone trusts any timeline given by Elon Musk, again? Imagine having the benefit of the doubt that Elon gets from his True Believers. It’s incredible! Look at this long-ass tweet that’s just a lot of wildly optimistic daydreaming about the Tesla bot without any evidence at all:
A thread.
— Ezer (@EzerRatchaga) October 1, 2022
It’s not that I think conceptually these technological achievements are impossible – like all engineering problems, I’m sure there’s solutions to the demands of a general-use humanoid robot – but at this point, threads like these are just embarrassing fan fiction, and it’s a phenomenon seemingly unique to Tesla.
When Boston Dynamics shows their robots dancing or robbing vending machines or whatever, you don’t see these kinds of starry-eyed dreams of The Amazing World To Come like you see with Elon Musk or Tesla. It’s deeply strange.
Here’s the takeaway from all of this, as I see it: Tesla has built a humanoid robot, generally on par with what many other companies have been making, and it’s real, sure. I’m certain they’ll develop it even further!
But all of these claims that it’ll be cheap and work all day in any number of contexts and have powerful AI or ML or BS that will be able to figure how to do every job, anywhere, and free mankind from drudgery and poverty forever?
Come on. Get a grip, dummies.
Give the world cheap-ish generalized robots, and people will figure out what to do with them.
Right? This is the part that gets me. Best-case fantasy sci-fi miracle engineering scenario, this thing’s only reason to exist is to put millions of humans out of work. Elon’s been at war with his own labor force for years, this is his ultimate solution. In the long run it’s just about cutting costs, by eliminating those pesky, needy, willful, injury-prone, expensive human beings. It’s not just Tesla; every time I see a company proudly demonstrate a robot that can, like, flip burgers or whatever, I think the same thing.
To be clear, Tesla’s robot is not the key to unlocking a future of mandatory gay space communism. It’s not the key to anything. But the ideal of a post-scarcity society is, in my opinion, what we should all be working toward:
(1) capitalists (2) proletariats (3) means of production (4) bourgiosie
Maybe there is a possible utopia down this path, but the transition period would be absolutely grueling. Companies would use the lower production costs to increase their profit margins, while people lose jobs left and right. And much like communism, I am afraid the utopia part of the equation would keep slipping further and further away…
It’s always being pulled away by those in power because greed remains one of humanity’s greatest weaknesses.
Something about bootstraps no doubt.
I’m sure every business with a factory would love a fully dark warehouse where robots whir and products come out the other end, perfect and good, but a walk through a trim shop and final assembly line of an automaker is a good example of why humans just WORK; Easily trainable, can easily swap jobs, can potentially improve the product, and when the mid-cycle refresh comes and you need to route the harness under this widget instead of over, it’s a single-day training. I don’t intend to be negative, but there’s a reason we have robotic arms and wheeled drones in our factories instead of Marvin and Bender.
Machine learning can obviously simplify this, but if ML was a perfect method to teach a robot something, why can you fool a Tesla into stopping sooner by installing a bigger stop sign? This is assuming of course that Tesla’s AI people are talking to each other…
So, to look at the “replace all factory workers” guy: We have replaced factory workers where robots make sense, and those robots are designed for that purpose. This isn’t going to do anything better than the robots that are already real and in use even in a best-case scenario. And it’s probably not going to do anything better than the humans currently doing whatever job they’re doing.
The robot is a thing that exists because Elon Musk got high and watched The Jetsons.
So, to look at the “replace all factory workers” guy: We have replaced factory workers where robots make sense, and those robots are designed for that purpose.”
Maybe some factory workers have been replaced but there is still one HUGE area of labor ripe for replacement. One that would be best served with the most human appearing automatons:
Sex workers!
“This isn’t going to do anything better than the robots that are already real and in use even in a best-case scenario. And it’s probably not going to do anything better than the humans currently doing whatever job they’re doing.”
Au contraire! Done right they will be able to satisfy any kink, be unable to pass along any disease, never get pregnant, never judge the customer, never cheat the boss (if rented), be available 24/7/365, be ruggedized for special requests and be trackable if they end up in a trunk. No messy morality issues, no “cheating”. Hell let the jealous wife or crazy GF shoot or stab the damn thing, it won’t care.
Rosey the robot indeed!
Not necessarily. You can’t blackmail someone for having sex with a robot as there’s nothing (currently) illegal or immoral about it.
If the threat is to immobilize the robot I think that threat is even less than that with cars, especially those with OnStar. Folks have cited concerns of hackers taking over cars for decades yet AFAIK nobody has been killed when a remote hacker cut the brakes or steered the car over a cliff.
Come on, man, don’t take this guy out of context. He didn’t say “no poverty” was step 2. He just glossed over a few of the more minor steps:
- Tesla robot 2) Computer scientists produce perfect AI systems 3) fusion energy production becomes widespread 4) matter synthesizers are created (as in Star Trek and The Orville) 5) The world government is formed and distributes fusion and matter synthesizing technology equitably to everyone worldwide. 6) No need for anyone to work or produce anything. Your fusion-powered matter synthesizer can produce anything you need, including the latest model of the Tesla robot. 7) Poverty is eliminated and world peace is achieved. But the IMPORTANT thing is it all started with the Tesla robot! And/or with John Lennon singing “Imagine”. Anti-Tesla Historians will debate the details, I suppose. Teslabot needs to be small, efficient, do what needs done in the least amount of power ????. It’s only 73kg” In designing a humanoid robot capable of doing all sorts of work, Tesla is going to have a lighter, more efficient robot, because it doesn’t need to be able to do all the things Boston Dynamics is doing. For some reason. Musk is promising a MORE generalized robot that is somehow lighter, more efficient, and doesn’t need capabilities. I guess. So, an average American then? 😛 If history has shown us anything it’s that miniaturization will be part of the natural evolution of the technology. The hard part is getting it to work in the first place, which Atlas seems to be way ahead of the game on. Fuck off back to the lighting site, vatnik. Regarding getting electric cars with long range available for sale at a semi-affordable cost, Tesla did in the 2010s what the mainstream automakers could have started doing in the 1990s. Without Tesla, we might not have EVs available on the mass market today, simply because until Tesla offered them for sale, the mainstream automakers had no incentive to do so(EVs threatened existing recurring sources of revenue), and no competition willing to do so. The technology was ready enough for niche market acceptance and even early-adopter interest no later than about 1998, which would have been enough potential sales volume to justify a mass production run to get the cost down to something that your well-off suburbanites could afford. The Solectria Sunrise was getting more than 200 miles range in real-world driving and set a range record of 373 miles at the Tour De Sol, and the designer, James Worden, claimed it would have been $20,000 car in mass production, as but one example. “Think of Boston Dynamics as a hardware company while Tesla is more software.” Oh wait. That’s Tesla’s software. And infrastructure. I wish he would retire to privately-owned SpaceX where he can forget about creating investor narratives and concentrate his energy to get Starship off the launch pad. Twice the kids in half the time: https://m.youtube.com/watch?v=z0NgUhEs1R4
