I have explained this before. They have to design a factory right now to be safe for humans, so any factory that a robot works in, has to be designed for human safety. Also, like I showed you with Agilities humanoid robot. If there are stairs the robot has to be able to navigate those stairs. You could design other forms but the humanoid form is best since the factory is already designed around the humanoid form. You do not want to design a factory for a dog form and then put Humans into it to run it.
Let me ask you this. Can you think of any reason that Agility would be producing humanoid robots? They are getting ready to produce them so why would they do that?
Cosmetics and aesthetics come immediately to mind. The exact same reason why they put two âeyesâ and a âheadâ articulated separately from the âtorso,â even though thereâs no engineering reason for those things at all. A wheeled platform with an upright post with two arm-like appendages attached to it would have the same functionality - but it doesnât say âfuture.â
People have been designing and building robots for many years - and while thatâs a good job and makes valuable products, it doesnât have much buzz any more. But if youâre making humanoid robots? Ah, thatâs something that really sizzles. If youâre a start-up and you want equity investors and media coverage and big companies like BMW to let you partner with them, a plain old boring robot isnât going to get you in the door. Or attract retail investors when youâre ready for your IPO. But a humanoid robot that does the same thing, just with a lot more complications? Thereâs the ticketâŚ
So Musk and Agility are making humanoid robots for basically appearance. I could see Agility doing that because they do not have a brand but Musk? There isnât any reason for him to do it. He is in the news every day already. Also there are to many people trying to make humanoid robots. Letâs see what they say.
Agility: Agility Robotics says that its manufacturing facility will be the first to mass-produce [humanoid robots] which could be nimbler and more versatile than their existing industrial counterparts.
China: The Chinese guidelines say industries like healthcare, home services, agriculture, and logistics will likely see a rise in the use of robots in the coming years. The country also wants to put effort into developing humanoids to work in harsh and dangerous conditions and within the manufacturing sector .
Oh, sure. Musk as much as anyone. The Optimus looks the way it does for cosmetic/aesthetic reasons as well - which is why it hews so closely to the human form in areas where thereâs zero engineering reason to do so, like the shape of the head.
The first is for attracting talent. Tesla needs to (finally) solve autonomy. They need AI boffins to do that. But working for a car company isnât a traditional AI job path, and for the last several years the space has been filled with other firms (OpenAI, DeepMind, Anthropic, Inflection, Mistral, and the like) competing for AI talent. Hence, AI Day - Teslaâs effort to convince AI folks to have their career at Tesla rather than an actual software or AI company. To justify that, you need projects that will excite the engineers - get them eager to play in your sandbox. Doing the final-stage work on an (allegedly) nearly done FSD project that will end the companyâs foray into AI work isnât going to do it. Hence, Dojo and Optimus.
The second is for appeasing investors and building your brand. Teslaâs had to spend billions and billions on AI programs in an effort to make FSD work, including billions on building their own supercomputer (Dojo). Itâs not a great look to invest in all that and then toss it once your autonomy problem is solved. Nor do you want investors to start valuing Tesla as âjust a car companyâ - there always has to be a ânext big thingâ around the corner for Tesla. Not just for the investors, but also to keep Teslaâs brand in the news. If youâre not going to advertise, you have to be doing something more newsworthy than introducing a compact SUV.
And finally, the third audience is Musk himself. He wants the future to look like the future. More specifically, his personal 90âs-era cover of Omni magazine vision of the future. And he gets to pick the aesthetics. Thatâs why the Cybertruck looks like it does - that was Muskâs aesthetic choice, and he runs the company. And thatâs partially why Optimus looks like it does. If Teslaâs going to build AI robots, theyâre going to look the way Muskâs vision of the future looks like - not necessarily the way that makes the most sense for them to do their jobs.
Hereâs why it matters. From what I can tell, the main argument for using a humanoid form is that it works well with other humans. But if the goal is a factory without humans, then that argument for a human form robot starts falling apart.
You seem wedded to the status quo, specialized robots that do one thing.
Not at all - not in a general sense. But I thought this all started by talking about making assembly lines fully automated. When youâre talking about an assembly line, youâre talking about a series of tasks. One job done many times over. Of course, thereâs almost infinite complications in the ordering of jobs and finding places were multiple jobs can be done at the same time. (Trivially - and being done for decades already - assemble car bodies and engines separately, then bring those together. Or install both front and rear glass at the same time, or hood and trunk, or left and right doors.) These all eventually come down to one thing. So you make a robot that does that one thing (more realistically, a limited number of related things in a short sequence). Then you make another robot that does something else. Itâs almost a certainty that the best robots for those individual jobs would not be humanoid in any sense. Probably not needing any significant AI, either.
Where I could see a more generalized - approaching humanoid - robot with a rudimentary AI, is automating a low volume, semi-custom shop. Particularly where you need to quickly switch from producing one product to another. Kind of a general assembly line where the item being assembled changes frequently. Today, weâre assembling toasters. Tomorrow itâs blenders. Next week weâve got a run of toys and a couple different pneumatic tools coming up. Ahead of time, you train a single robot to do all the individual tasks necessary, then they share that training with the other robots.
To attract talent all He needs to do is call someone and say come work with me. Musk can get all the talent he wants. No need for a robot he has Tesla, SpaceX and now the neuralink. Who wouldnât want to work for him.
Your right FSD also how did I forget that. There is another project that he has that is just begging for people to work with him.
Musk is just going to build something that is better than everyone else because that is how he walks. He could build a robot on wheels if that is what he needed but he decided to make a humanoid, because that is what he needed. How do I know that. Because we have Optimus, itâs been built.
Someone who wants to be in the cutting edge of AI, if Tesla isnât doing an AI project. Which is why Tesla needed to have more than FSD on the to-do list.
Sure - but that doesnât mean heâs right, any more than they were right about the Solar Roof. Or that theyâd be able to have robotaxis in the Tesla Network in 2019. Or that thereâs a way to massively increase tunneling efficiency. Sure, itâs a cool idea to build a general-purpose android worker - but that doesnât mean the tech is ready for that to be a real thing. He can tell his engineers to go out and make a humanoid robot to do a job in his factory that isnât currently already being done by a robot, just by watching the human currently doing that jobâŚbut that doesnât mean that itâs going to happen.
If the argument is that these robots have to be shaped like humans in order to be able to work in human environments, it should be possible to explain away the literally millions of robots that are already working in human environments without being shaped like humans. I donât think âbecause Elon decided that is what he neededâ is a persuasive argument. It should be possible to actually offer the reason why you canât have non-human shaped robots working in human environments when thatâs already happening in literally millions of places.
First, we have had specialty robots for a long time now. Yet most of our factories are NOT completely automated. We still need lots of human workers. That suggests that while specialty robots are important, they arenât nearly sufficient. Musk himself found this out when trying to build the Model 3. You still need humans. To replace those humans, you need robots with human capabilities, ergo humanoid robots.
Second, humanoid bots can learn tasks from watching human behavior. While robots of any shape can potentially learn in this way, the more similar a robot is to humans, the easier the learning. It is difficult for a robot with pincers for hands to learn how to make a snowball watching humans.
How many car assembly lines do you know of that are fully automated? If the answer is none, then how do you replace the needed human workers? One obvious answer is through humanoid robots.
Why is that an obvious answer? After all, every single other human worker thatâs ever been replaced by a robot was replaced by a non-humanoid shaped robot.
This argument only works if the reason automation hasnât happened yet is because of the shape of the machineâŚbut I suspect thereâs not a whole lot of assembly line jobs where thatâs the case. Consider, for example, the meat processing industry - a notoriously labor-intensive manufacturing process that has resisted automation. But the reason that itâs resisted automation isnât because we lack robots with the form factor of a human - itâs because:
The manufacturing sector, including automotive plants, has relied on robotics for several years, but putting a bolt in a car is very different than working with meat, since no two animals are the same.
One of the challenges is replicating the human eye and touch. So far, robotic butchers arenât able to make precise cuts and can also struggle to accurately tell the difference between skin, fat, bone and meat in chicken and turkey facilities.
If you solve the âbrainâ and âeyeâ and âcuttingâ problems, then you can use robots in meat processing - regardless of whether they have legs or torsos or hips or what have you. Thereâs no need for robotic meat processors to look anything at all like humans.
Apropos of this topic, a Delaware court has voided Muskâs 2018 compensation package and ordered rescission of the contract:
âŚwhich will certainly make it far more difficult for Teslaâs board to give Musk what he says he needs in order to keep the AI and robotics programs at the company.
I already told you that, yet you refuse to listen. In fact we have had this discussion before and I have told you many times before. Just because you canât remember does not mean it isnât valid.
LOL really, tell me how many robots we have right now working in a meat processing plant that have a human form factor. You are just making stuff up now to suit your narrative.
None. We also donât have them in non-human form factor, because the job is too hard and robots arenât good with spongy, stringy, squishy things. Making the robots even more complex isnât going to change that.
Labor is much more expensive in a gig factory. Telsa in the early numbers would save $25 b over five or ten years. The numbers are a bit vague to me. The heart of it is expensive labor that can be replaced first. Later some less expensive labor can be replaced if productivity can rise. MCD can not feed more people in your suburban shop just because it could make 10k burgers per hour.
In harder conditions in factories productivity soars because conditions can be met at higher production rates. The human body can not take as much punishment.
The robot instead of just an arm is like the internet instead of just a PC.