I feel like Marques is missing the point. A human shaped robot will be significantly harder to make than a specific special purpose robot. However, if you can solve that problem the utility from the bot will be orders of magnitude more capable and the utility of it is nearly endless. Not to mention the cost benefits of focusing on a single form factor compared to 10's, 100's, or 1000's of specially purposed robots to achieve the same capabilities the Tesla Bot MAY be able to perform.

Doesn’t Elon want to colonize Mars? And hasn’t he said a bunch of people would die going to Mars? Wouldn’t these robots be perfect for sending to Mars to build and prepare it for colonization without the loss of human life?

I think Elon is aiming to send those to Mars for tests and probably to build the infrastructure on Mars.

There's a huge problem with the "action specific robot" argument. If someone builts one of them, it is only useful for that purpose, and that is vastly more expensive, and human labor continues to be cheaper (nobody wants to do that kind of investment). In contrast, if you have a humanoid robot, multiple companies will have interest in them, making them a smaller investment, achieving 90% of the benefits, with 10% of the investment they would've needed to make a specific robot, for that specific task.

Seriously ? I Never thought MKBHD would be this simple-minded about innovation and new tech, I always watch your videos and tech reviews, first because of you are a Tech enthusiast. Your opinion on this new Robot is unimaginative

The point of the robot is to have something that can do almost every job so you don't have to design a new robot for every situation. Yes if you're talking about a single situation then it's cheaper to design a custom solution but by the time you're onto situation 1000 and every one of them required an expensive design and testing process with custom ai to make it happen then the Tesla bot seems like a good idea. Remember that this is how Elon thinks.

Was cool seeing your interact with the community like this and just have an a discussion format about this. Very healthy informative and really interesting to watch!

I would LOVE to see y'all interview Tom Scott!

I love how Andrew is so pessimistic about technology development lol 😂 "I'll be dead before X"

The thing about a bunch of purpose built robots is... you need space for all of them! Theoretically: a humanoid robot can literally wash clothes with its hands... And save the square footage area which would otherwise be taken by a massive washing Machine!

Is washer-dryer combo not a thing in the US? Cause Marques talks about it like it doesn't exist 😂😂😂

Been waiting for this one… Lol I actually work for that grocery company with all the bots, it’s called Ocado and mainly works in the U.K

Loved the Discord discussion. Very patient and deliberate with the Q&A :)

Specified Null's comments were amazing! Great episode everyone - thank you.

Good discussion. I think the single use form factor is missing a small point: the largest training set we have is how humans move in order to accomplish tasks. If you create a new shape/design, you have no examples to mimic and it becomes a reward based feedback loop instead, which would also be significantly limited in scope. Generalized is the key using massive sets of data, human form is the most documented of all forms.

I don’t want thirty robots in my house. I want just one or two to help me out with everything.

I really love the format you guys are putting out. First a highly polished video about something than a deep dive as a podcast. Perfect! I think the humanoid robot makes sense if it could do all the tasks a human can do like a personal butler. So it could do all the stuff that's not fun but doesn't occur on a daily basis so it wouldn't make sense to have a specific robot for the task. Like building furniture. like a janitor basically. But that's so complex that's so far away

I think a good parallel to this discussion would simply be: How do you feel about general purpose AI? That is the end goal of AI research, after all. The general idea has always been that the human shaped robot would go along with Gen AI.

I think this is what you are missing Marques: you posed the question “is it easier to improve the Hubble telescope, or make a humanoid robot?” Also, “is it easier to make an automated grocery store, or a humanoid robot” also “is it easier to make a dishwasher with a robot arm that can reach across your house and grab dishes, or make a humanoid robot” I think the question is answered when you combine it into one question: “is it easier to improve the Hubble telescope, make a crazy robot dishwasher, make an automated grocery store, make specialized robots for every task in a factory, etc, or to make a humanoid robot?” I think it’s easier to make the humanoid robot.

I personally like the idea of a spot robot with an arm/ arms that can be a better option. But great brainstorming! I think by talking about different ideas we can get to a better result!

Essentially the question I would like to ask is if the human form factor is flawed than what other form factor that would be more adaptable across our life span and all the activities we have accomplished so far? I would say that human form factor is crucial for the maximum range of mobility to make it adaptable to "all" tasks. And that has to start some where and let this be a step forward to that direction.

VERY innovative conversation about the humanoid portion. I think that Marques is spot on. The thing is, tech and applications are developing almost exponentially , we are already making machine and services that act autonomously. It would make more sense to make everything else autonomous and have it interact with each other. We can easily make a machine that sort dirty dishes, wash them, dry them and sort them. we have recycling depots that can weigh and sort different sized bottles- we can use that technology in some type of way to help. to make a robotic butler seems like it just kind of kills the inspiration of AI. we can easily make a 90% autonomous kitchen and laundry system. you still have to go sit down and eat, you still have to select the food you want. you have to do SOME TYPE OF WORK lol . so you will have to AT LEAST put your dirty dishes in the sink and do a couple other things obviously. and the grocery store bit, with the amazon warehouse-style grocery store, and and TESLA, your car drives itself to the store, pops the trunk and a robotic cart loads the trunk, and the rest is history. laundry machine they already have machines that can wash then dump the clothes in the dryer (stacked on top of each other). Essentially, our lives would just be restocking, reorganizing and online shopping.

This is the same thing they're doing with vision based self driving: an approach which is harder initially but if successful has a much larger use case and addressable market

You can make it do anything that you can have another person do, it can teach you anything, show you anything, get you anywhere, etc. All while doing everything you ask without question, as efficiently as possible...

If you want to send it to Mars, you are going to need general purpose robot where it may not be best at everything but it could do everything

Noah's (Discord) example of why we have to learn how to drive and get a license, vs walking up and down stairs: the determining factor is how much of a safety risk does each scenario pose to other people around you. Walking stairs does not really affect others around you if you fail at doing it properly, where as driving badly can have much more dangerous consequences. It's why the learning process of one over the other is much more controlled in society.

I think the idea of a humanoid robot is to promote the ability to interact into a human focused environment. There is a psychological component to make humans less startled or fearful to interact with a strange, unfamiliar shape...but the real benefit could be a robot that can drive a legacy car as well as /better than a human chauffer (sit in the drivers seat, operate the pedals, gears and wheel as well as the switchgear), move around a cityscape, into a home/office, negotiate stairs, prepare a meal, feed the disabled...sure a more specially designed robot could perhaps do most of these tasks slightly more optimally, but this is a human designed world and for a robot to do as many things as people can, the human form needs to be duplicated.

I'm halfway through this, sorry if this was said towards the end but my thinking is: Yes, having specific application "robots" is more efficient for the future. But a human-like robot makes more sense for us to get that future-like experience right now in the short-term since it will fit most scenarios. It won't be the most efficient for each tasks but it will fine for most tasks while having the huge advantage of being able to do MOST tasks, which is the real value of it right now. Not that we even need it right now (for most people) but this is how I see this being useful

You should interview Lex Fridman about the human-robot interaction side of things, he seems to have some interesting ideas, I don't know if I agree with everything but could give an interesting perspective.

If you look at home automation it is built around special purpose robots that are typically integrated with a voice assistant. Lights, cameras, locks, garage openers, blinds, thermostats. It's cheaper and more efficient for me to have these special purpose devices than one robot. If i want to lock my doors, close all my blinds, and turn off my lights I can literally give one command instead of having a robot run all over my house to do each individual task. They have robo mowers for lawns as well, so u don't have to do that either. And we aren't even talking cost. A highly mobile general purpose robot is going to be insanely expensive.

You cannot mention the Tesla Bot announcement without their insane new chip that they'll be scaling in the near future. Software is the real issue in robotics, not necessarily hardware anymore. Tesla will be in a unique spot with their ability to train AI at insane speeds.

Here is an idea: someone has to screen what people are going to talk about. So make this text based and I every 30 minutes or whenever someone reads out a text version of the question. We need quality control

For those wondering how can they possible make a robot so thin when Atlas is so bulky? Remember how bulky Spot was during early development? Things as they are today are not things as they can eventually become. Boston Dynamics, or any other company, will always start off trying to make it functional before working on slimming it down and making it practical and consumer friendly. Not only is the tech advancing rapidly, but with a mega-cap company like Tesla behind it, who already have some great advancements with their vision AI and D1 chip, imagine what they can do.

So for a human shaped robot I think the benefits in the short term would be that it would be cheaper to manufacture a single design that can do multiple things then having either increased manufacturing capacity or a higher cost per unit. And second you could use all the information gathered from learning so many individual tasks that the AI could likely be used to build better factories. The human form isn't needed but the capabilities are. So cheap manufacturing the quickly get the AI out and to learn. Refine the software then refine the form. Perhaps a quadraped would be better but perhaps the increased weight meant you would need a bigger battery and was therfore more expensive.

I understand that we aren’t “there” yet with generalized human shaped robots technology wise but that does not mean it’s not the best option. I’d rather buy one humanoid robot to take on some of the tasks in a world that we shaped to work for a human figure and navigate through it than buy a ton of robots to execute one simple task, there’s definitely a price difference between both options

One thing I think is being viewed the wrong way is people thinking it's easier to make a single-purpose robot to do any 1 thing vs a humanoid bot that can do it. But the more relevant comparison if you're making a multi-purpose robot that uses AI, is what would be better at solving the most problems at once rather than solving one at a time, and also future proofing that solution? So in terms of solving the most things at once, the more apt comparison is what's easier: Designing and building hundreds, or thousands of single-purpose robots to do one thing each, or designing 1 robot that can do them all? If you try to compare 1 to 1 then yes it will seem that it's easier to just make the single-purpose robot for this one thing, and then that one thing, and then that one thing, etc... But looking at the bigger picture and using first principles thinking, one can definitely make the argument that what it would take to build a multi-purpose robot that can do all the things might actually be less than what it would take to build the hundreds or thousands of single-purpose robots that would be able to do everything we do as humans in the long-run. In terms of future proofing: You would have to continue making single-purpose robots in perpetuity to solve every scenario that comes up, or if you make the use of AI and neural nets, a humanoid bot can learn how to do new things as they come up and continue to bolster it's capability over time without having to design something completely new every time, just like a human.

You still need a person to operate application specific robots, so the general purpose robot can operate them

the thing is about the general human shaped robot, yes of course robots for specific tasks are going to do better for these specific tasks, however once you have a good human shaped robot, it can do most the other tasks for when the user either doesn't have a specific robot or they maybe can't afford a robot for every tasks. In the end there will be both, the world is mostly setup for humans, so there are a lot of places where a human shaped robot will make sense and will be a thing eventually.

I feel application specific robots are improving with time, and the definition of robots has been visualized as per movies but any technical application that is automated to do stuff comes under the radar of robots. So Tesla trying a human shaped robot could be not a lot useful in its first iteration but it could lead the door to improve the process and eventually when released we could actually compare a humanoid robot vs application specific robots. Great Podcast though!

I'm no robotics expert but personally I think that the future of robots don't need a humanoid form factor. On one hand, I can see how they are superior in terms of intuitive-ness as in the far future, operating them may be as simple as giving them a voice command as if it's just another human being. However, I also agree with Marques' viewpoint about how the future may be in super specific robots performing super specific tasks. This route also has the advantages in repairability, as a humanoid robots' versatility may also become its downfall in the event of a malfunction; if the robot is out of action all the responsibilities and functions are unable to be performed whereas if one non-humanoid single function were to malfunction, the other robots in a persons' use will continue to operate as normal minimising disruption. Also, I don't think that the world isn't inherently designed for the human form, rather, everything that we as humans create try to make everything easier for us perform. I can't really explain it thoroughly, but for example we could completely omit the need to do groceries, such as the act of travelling, browsing stores and purchasing items, if we were to create a system where all these functions were automated, like if all products were delivered through a delivery tube simply by purchasing items through an interface, whatever form factor that may take in the future. This in my mind would be faster than programming robots to do the groceries the way we would. This may sound evil-villainy but I think that we can only surpass human limits and optimise functions if we abandon the human form.

You know what this got me jonesing for? MKBHD & The Studio giving their thoughts on pieces of tech from sci-fi movies like Minority Report, Elysium, Total Recall, etc.

Very excited to see OnePlus buds pro review 😁😁

Discord input was dope!

I would argue that a humanoid robot is more disconcerting and creepy than just a thing that looks like a machine. I’d be freaked out if I’m walking through target and there’s a bunch of iRobot robots walking around getting stuff. Grocery stores already use robots on wheels to do stocking and what not. Walmart and Kroger. There’s also robots in use that roll down the sidewalks of my city to do food deliveries.

Started getting groceries delivered during lockdown.... Hmmm... So I'm not sold on the groceries usage. So glad that they mentioned the Tommy Scott Auto grocery Warehouse video. That was scary cool.

I’m sure they’ll manage to make something that walks and demos some specific applications well, but this is taking a form factor which demands a large number of degrees of freedom which makes the control much harder and it sets up expectations in people who interact with it so beyond just being creepy anyone who runs into one will think it’ll have human level intelligence and capabilities and when it doesn’t meet those, it’ll feel worse than if it had a non-anthropomorphic form

I would think the best option could be an adapted human form. Something with legs that could fold into a car like a human but split maybe into 4 vs 2 for better balance, or have treads on the feet to “roll”/“cruise” around vs walk but again split and walk up stairs… but with a mostly humanoid top.

Finally the background is symmetrical !

I saw myself casually chilling in the event 😂

That was the best episode ever!

What a great podcast! The idea of transplanting our head tho, when we reach 90yrs old could be simpler. If we have a humanoid robot that we can control, and AR is also developing, then the Tesla bot could be the vehicle itself in experiencing everyday life and doing different tasks through haptic feedback etc.

Overall I appreciate the thoughts and perspective you guys present. But wow, starting at 11:35, a lot of what follows about AI, over the next few minutes, is misleading. The problem of self-driving is so difficult *precisely because* it really does require general-purpose AI. And the claim that FSD "has a very specific set of objects that it is going to react off of" I think illustrates the disconnect. The set of objects and scenarios that a self-driving vehicle should be able to identify and react to is nearly limitless. Doing this literally requires the car to understand the context of its surroundings. An AI designed to drive a car on the street has to be no less general than an AI designed to operate a humanoid which walks down the street, let alone one which carries out boring, repetitive tasks that don't weigh thousands of pounds and move at deadly speeds around other humans. While it may be true that a purpose-built robot will excel at the task it is built for, we clearly can't have a separate robot for every task a person wants done without ending up with dozens of robots per person. It's simply not practical to produce or purchase yet another robot for each new task, not to mention how inefficient and uneconomical it would be to have personally-owned robots that sit idle for 90% of the time.

1:00:00 That laugh lol!! GUYS I LOVE THE PODCAST!!!!

love ur podcasts

I don't think Tesla Bot will be as general as a lot of people are thinking. I don't think we're anywhere near the level generality with current AI to do general tasks like that. Even full self-driving is a massively narrow AI problem compared to telling a Tesla Bot to go and do your grocery shopping for you, and then tell the same robot to clean your house for you. It has to master multiple domains of knowledge, environments, be able to understand a wide variety of commands and respond appropriately. I just don't that level of generality being achieved in the next couple of years at least. Assuming they can standup a prototype of this in the next couple of years, it will still probably be really brilliant and a massive engineering achievement but let's keep our expectations realistic.

that discord session ..thats a great community...

I think that you guys are kinda missing the point. This is a job replacement robot so instead of trying to create a job for the robot to do look for job a job or jobs that are repetitive and boring. I personally think most minimum wage jobs fall into this category. Fast food you make the same thing over and over all day. Store clerks spend a fare bit of time stocking the store with new inventory. Hotels room attendance clean and organize each room almost the same way over and over again. I can't imagine janitorial work being terribly complex and even if I am wrong I don't think the tasks would very much day to day. So imagine if the head of some major ferm can buy three or four of these bots and replace there buildings entire janitorial staff, cafeterias whate staff, and the bulk of their overnight security staff. 3 to 4 bots to replace possibly 30 to 40 minimum wage jobs in one building.

When the slack notification goes off and i start to freak out that it was me haha

If the world was run by pessimists, then we’ll still be in the stone ages. The best tech UKpostsr with one of the most open mind for technology and innovation, with one of the most uneventful pessimistic individuals on any tech platform.

Yeah for outer space stuff i think it makes way more sense to build an specific specialized robot. Right think about it, you're in space, on a satellite or the moon. Both are environment's that are not the earth and are not built around us as humans. It also has a much more narrow scope. "fix the telescope".

Wanted to post this in the discord but it's too long, hoping maybe someone sees it. I loved the latest podcast. TLDR: building robots is hard, I don't think they'll pull it off in a short space of time. Assuming they can make them, I believe they will start in niche applications and industry, move to public services, health care etc, then roll out more human mannerisms allowing more support / human relationship roles. Hopefully there is legislation and world agreement for them not to be weapons.... Legal and ethical questions about robots with emotions? Nuralink, summary of my opinions, and finally a question about if this is a good idea environmentally. So I had some thoughts on the Tesla robot I'd like to share as someone who's in the electrical and electronic engineering sector as an undergraduate having worked for / with a couple of companies. Robots aren't my specific field but I have a good amount of experience working on electric vehicles. So I'm also pretty pessimistic about how far Tesla will get in the time frame they've laid out, as there are very good reasons Boston Dynamics robots look they way they do. Humans are pretty unstable, which allows us to be nimble, and if you think of humans as robots, we have the most powerful computers in the known universe in our heads, our skin acts as a sensor suite across our entire body with a crazy amount of detail, and we have tons of joints ligaments and muscles that can move in a ton of directions, which allow us to maintain balance and execute the perfect movements to stay balanced. This is very hard to replicate in software, if you can make hardware good enough to feed that software with the data and movements it requires. It is however achievable given enough time and let's assume for the sake of my further thoughts that tesla pull it off in the relatively close future (like 5 - 10 years) So I would argue that tesla is playing a very long game (like most of their projects, ie renting a car service model). People initially will be uncomfortable with human-like robots, they are threatening and make us wonder our place in the universe and will we be obsolete etc... The initial roll out would be niche markets imo, with custom software running per task (rather than a software that can do every task like alluded to for a lot of the podcast). It would be similar to spot, doing dangerous tasks or understaffed jobs with little human interaction. Obviously industry would be the first market as they'll sell packages to support the clients properly and can monitor each contract individually and get a good PR image for the products. Once people are more comfortable with the concept in general, they will start to occupy jobs in the public sector. I recon things that require operating in the human world, collecting garbage, cleaning public spaces, maintaining plants and such, fixing roads. Once they are in society, and have been proven to be safe, effective and trustworthy it would only be a matter of time before they took on more roles. I think one of the major things being worked on behind the scenes would be emotional support, interaction and human like body language etc. Nailing this would allow robots to take on roles of nurses, therapists and other carer roles. These roles, humanising robots, I recon would lead to society giving them names, imagining (or experiencing) personalities and at that point people would want them in their homes, helping with tasks and doing all the things that humans need to live in a 1st world utopia. This is quite optimistic thinking, assuming no military contracts or oppressive police regemes had been developed out of these programs and hopefully we can put the fail safes in to protect against this. If we did end up in this utopia, the next stage would be forming proper relations with these robots leading to a strange world with legal and ethical issues, do they have rights? Are they still property if they can "fake" a personality? Are they approaching human with enough development? That is a whole world of avenues to dive down which I imagine Issac Arther would do a far better job of exploring. As for the idea of getting Nuralink to a point of putting brains in these robots, that's either a dystopic future where the rich live forever, or a post scarcity world where we have space mining and a foothold in the universe allowing us to do all the things that we dream of in science fiction. In conclusion, because I've made this post way too long, I think these robots will take a role in industry, assuming they are fully developed. And there's a chance that they will be integrated into our societal structures as assistants. Beyond that, I have no idea, and the rate at industry adopts, changes and moves (even with Elon's crazy push for tech) society will be very different and maybe more or less accepting of this tech. I will leave on a question though. What would be the environmental impact of building these? Assuming it's at scale? Will it fit the future where we need to be green? Will it help with that?

The best way understand humanoid form. Give a serious thought to answer below question. Imagine you need to establish a Colony on another planet. And you need maximum flexibility regarding the tasks you need to perform to study, adapt and build the infrastructure. What would be the best approach? Let me remind you, mars rover was completely useless when a dust storm covered the solar panels. Let me remind you that Perseverance rover is very complex as a locomotive and very inflexible with the way it can work with the on board science experiments. Argument is, it would be much better long term is we can send two humanoids and bunch science experiments in a box. It could even repair a faulty instrument and could recover the hardware sent on previous missions. Think about it.

"Don't forget to drink water" 😂😂😂

My Google home did not make me a cup of coffee. But Noah is right, all about that software. AI day showed they are already creating models to emulate human behavior for pedestrians based on camera data from the cars.

We shouldn't forget that Tesla is also a company so essentially the humanoid robot might be a marketing item, there are already a lot of companies that are making non humanoid robots. So the engineering complexity will be worth it at the end if they want to stand out

Thank you all god bless

13:23 I mentioned this on your main channel video. Wouldn't a self-driving shopping cart with an arm to pick up & place groceries be better at grocery shopping than a human pushing the cart? 15:14 I guess you addressed that 1:09:49 Unless you have some sort of remote control

I think you guys are missing the point here. Yes, task specific robots work better for a single task. How many robots would someone without full use of their limbs need to have an improved quality of life? How would a task specific delivery robot deal with hundreds of years of building codes and building processes around the world to deliver a package? Stairs, ramps, elevators, various doors, finding the correct location (cabinet, doorstep, delivery boxes) are all special cases. If you want to get a robot to do all possible things for deliveries use the current form factor for delivery people. Most of your recommendations are based on a first world country ultra gentrified areas. How would a third world country deal with a fully robotic grocery store? Please look past your life, your experience, and how this would work in your life. See beyond your limited views of the world. I wrote this partway through the video, as I’m writing there’s some good points that are similar to what’s being talked about.

Unrelated: I just don't understand why they don't break down and add front small lidar and sonar, just for an extra layer and check, for safety. Their cars probably wouldn't run into cop cars as much, if they diversified their input sensor array. Not saying cameras are bad, just saying why not get some extra data to add to the mix. Lidar sensors are a lot smaller now. It doesn't have to be a or b, it can be both.

The humanoid form factor robot gathers a rather specific constellation of data that has utility value of itself. It has a vast potential for broadening and deepening the machine-human interface. Theoretically easier to relate to and accept. Sets up a vector for crossing the 'uncanny valley', for good or for bad.

1:00:47 i lost it hahahaha

It's weird that people forget that they can and probably will make more than one robot, this one proves a concept, they will evaluate and reassess if something else is needed, but for the particular area they are targeting initially seems to make sense.

Human form is by far the most versatile robot. If one is going emulate something , might as well be human.

How about the fact you have to maintain and repair all of those little robots? Versus a single robot that would require general maintenance and updates. I think about simple stuff like how I have an iPhone and a MacBook yet they use different charging ports -.- now imagine the inconsistencies once we have more tech.

The general purpose humanoid robot will eventually get here, but I think it will be awhile (10+ years). In the mean time, application specific robots will become more and more pervasive. The longer the general purpose robot takes to get here, the more value (and lower cost) it will have to provide to be viable in the market

I don't know if the Tesla bot is being designed for being used on Earth. To me it seems like Elon might be using the R&D budget of Tesla to build robotics for SpaceX's exploration and colonization of Mars. Using Tesla Bots to build the human infrastructure on Mars before humans get there and in return they'll be able to sell it to customers on Earth.

How cool is it that we're even having conversations like this? Ultimately I think it's just hard to come to terms with the idea that a useful humanoid robot could be possible soon. I think fundamentally most people (including Marques) would agree that if you could just tell a robot to do something instead of doing it yourself because you don't want to... and they do it... even if you could do it better... that would be pretty fucking cool. And obviously humanoid is the form TESLA would go with because it can do the most things... because humanoid humans already do the most things... and because Tesla doesn't want to make grocery stores or dishwashers.

56:46 this is basically Upload but a physical version

What’s the intro song? Absolute banger

Can see this happening down the track: Nuralink will stream your conscious into the chip then you can change “vessels”

Call the last guy from the Discord stage on the podcast! He seemed really knowledgeable :D

The last discord member was on the right track except did not take the idea of Nueralink far enough. There will be no need to transplant the head or the brain because you will be able to "backup" your brain digitally. That is the future. If I am not mistaken Elon at a Nueralink event even mentioned backing up the brain I think. Obviously we are messing with Westworld at that point and time and "fidelity" becomes a concern. I think the thing Elon is trying to fix with one robot to rule them all, is that just like Tesla AI in the cars, they have to be able to be prepared for all instances. Similarly a robot would need to do the same. When you start building robots for one task and any sort of "oddity" is thrown in to the mix they will shut down and need human intervention, but if you design a build a robot that can adapt and work around unknown obstacales....is that not better?

23:06 Did the editor forget to put the link? 1:07:29 This: ± exists

We need to come together more than ever

The robot model reminds me of the uniform for fencing

The problem with comparing Tesla bots to existing attempts at building useful robots, and even some not so useful ones from Boston robotics is that Tesla's approach is drastically different. Just like Tesla's approach to self-driving was drastically different. Other self-driving and other robotics approaches see their problems as software to capture complex rules. Tesla went straight to neural networks. That is a very very big difference. They bet big on this by equipping every car with expensive hardware to collect data necessary for training. They are now confident enough with this approach that they think they can also solve robotics in the same manner. They probably think that their vector space approach is generalizable enough for training robots that do simple jobs. My guess is that their significant improvements in their ability to create realistic simulations can help them train the first version of robots on simulated data. From then on their deployed robots will be able to collect additonal data necessary to address other real world issues that simulations miss. The main purpose of humanoid shape is mass production for reducing cost. Building task specific robots will be expensive. Humanoid robots can also be trained by softare to possibly specialize on tasks.

Now my Google wants to make coffee

The last call is basically, talking about cybernetics. You might as well just download and upload at that point, just copy and paste ourselves. 😄

13:30 Best use case that comes to my mind is a caretaker robot for elderly people to make their life easier 24/7.

Do american grocery stores not deliver? Most of the big ones here in the UK deliver at a smallish fee or free if you order enough. I guarantee it would be cheaper to get every grocery shop I'm going to do for the next 20 years delivered for a fee over purchasing a robot concept which will cost tens of thousands best case scenario.

Example, the Humanoid Form, could easily transform to a quadriped and become a spot shaped, or it could contort, beyond human form to adapt to SEVERAL different form factors, to adapt to many if not all situations. Even wheels could emanate to provide for speed and mobility. It's easier to work from the Humanoid Form down than from a specific form factor like a rumba bot up. Don't disparage the Humanoid Form Factor too much friend.

Use of these robots- In military finding someone in harsh weather and terrain rescuing poeple from biulding fire etc In construction In shipment and fullfillment centres

One of the use case I can imagine is to combine neuralink with it. extract and upload memories through neuralink into the bot. Also the bot should have 4-6 hands, 2 is very common.

We could probably make better phone calls, text faster and take better photos if we had separate dedicated devices for each, but few want to deal with the hassle

What is the cost to rebuild every retail store in the us to be like Tom Scotts example vs developing a multi purpose robot?

That final guy had me until he said I would need to remove my brain. No thanks. I thought he would go much more of a VR route. Maybe it is just my pandemic life, but thinking all you need to go out is a VR headset and the robot, it would be pretty cool. I can be the one going to the grocery still from home. My Tesla (hypothetical) and robot are delivered to the store, and I start controlling the robot with VR and shop what I want. It would also be cool to be able to connect to someone else's robot by VR, maybe like floo powder from Harry Potter. You could be in the same room with someone in another country with the VR. Would be cool and not reaaaaally impossible, all you need to do is possess a robot via some future connection, VR, and a controller.

In future we might need multiple number of robots, having dedicated robots for each task can be inefficient. Better idea would be to have a design more robust to perform multiple task. Number of robots manufactured would be less as a single robot can do multiple things which would help to efficiently use resources on earth. I dont support the idea of human shape robot but robot which changes shape based on task kinda Transformers

Did the dude just describe Krang from TMNT? At around 01:00:00

That last guy went totally cyberpunk 😅

the humanoid form factor is extremely useful in things like elder care. A high functioning hff robot would be an invaluable aid for body-unfriendly care tasks like moving and lifting elderly clients.