Elon Musk Has No Idea What He’s Doing with Tesla Bot(spectrum.ieee.org)
spectrum.ieee.org
Elon Musk Has No Idea What He’s Doing with Tesla Bot
https://spectrum.ieee.org/elon-musk-robot
87 comments
I laughed reading this because I agree. The voice control on the model 3 is so bad it’s insulting someone even thought to put it in.
Compared to what? Coming from anything named Toyota the Tesla voice control is light years ahead.
> But humanoid robots are much different, and much more complicated. With rockets, well, we already had rockets. And with electric cars, we already had cars, batteries, sensors, and the DARPA competitions to build on.
I don't think this line of argumentation holds up very well: we already have decades of research on humanoid robots too. They're just not good enough to be economically viable but they same could be said of electric cars and reusable rockets.
Why couldn't Tesla just hire enough people to replicate the published research and then improve it? The total amount of money invested in robotics worldwide is probably not even that high. The capital required to duplicate Boston Dynamics doing parkour is probably much less that what Tesla is already investing in custom AI hardware.
I don't think this line of argumentation holds up very well: we already have decades of research on humanoid robots too. They're just not good enough to be economically viable but they same could be said of electric cars and reusable rockets.
Why couldn't Tesla just hire enough people to replicate the published research and then improve it? The total amount of money invested in robotics worldwide is probably not even that high. The capital required to duplicate Boston Dynamics doing parkour is probably much less that what Tesla is already investing in custom AI hardware.
A self-driving car is just an intrinsically easier problem to solve. In fact, the first car that automatically centred a car within a lane using cameras to track lane markings hit the road in 1977! They had cars doing cross-country trips on highways with traffic in the 1990s!
Obviously there are complexities to it, which is why don't have any level 4 vehicles is just so much lower.
Humanoid robots are just several magnitudes harder because you first have to master the basic mechanics of human-like movement before you can move onto higher order things like avoiding running into a person. With a bipedal robot, a simple operation like walking up a hill is a major achievement. In the car domain, you essentially get that stuff for free.
Obviously there are complexities to it, which is why don't have any level 4 vehicles is just so much lower.
Humanoid robots are just several magnitudes harder because you first have to master the basic mechanics of human-like movement before you can move onto higher order things like avoiding running into a person. With a bipedal robot, a simple operation like walking up a hill is a major achievement. In the car domain, you essentially get that stuff for free.
Humanoid robots might be hard but they've been successfully demonstrated. It might also be argued that machine learning can make complex motions easier than if everything is hand-coded.
The humanoid form factor might be an entirely unnecessary distraction, a robot with more legs or non-human articulations would be just as useful.
The humanoid form factor might be an entirely unnecessary distraction, a robot with more legs or non-human articulations would be just as useful.
Boston dynamics has found machine learning to be minimally useful here, and has had better results hand-coding most things. That’s not to say that is necessarily correct for all scenarios, but BD has lots of extremely smart and motivated people and I would take their results seriously - so at very least it doesn’t seem to be as easy as throwing a big neural net at it, and any machine learning approach will require hard and careful work.
Re: successfully demonstrated, again Boston Dynamics might be the best example. Their bipeds are nowhere near the functionality levels Elon is describing - fine motor skills, vision, even walking (it’s my understanding that Boston Dynamics has to individually tune their bipeds for every course) are years and years away from hitting Elon’s goals at current rate of progress.
Furthermore, the BD bipeds have a battery life of about ten minutes. You need a revolution in either energy storage density or kinematic efficiency or more likely both to get a useful assistant bot out of it.
Re: successfully demonstrated, again Boston Dynamics might be the best example. Their bipeds are nowhere near the functionality levels Elon is describing - fine motor skills, vision, even walking (it’s my understanding that Boston Dynamics has to individually tune their bipeds for every course) are years and years away from hitting Elon’s goals at current rate of progress.
Furthermore, the BD bipeds have a battery life of about ten minutes. You need a revolution in either energy storage density or kinematic efficiency or more likely both to get a useful assistant bot out of it.
I hadn’t even thought of the physics constraints on the energy needed. This is a truly ridiculous goalpost he has set unless he has developed the arc reactor to put in its chest.
"Humanoid robots might be hard but they've been successfully demonstrated."
None of these robots can hang our laundry to dry, de-bone a chicken or cook a meal. They cannot perform any of the thousands of tasks we perform every day that require dexterity. They have just about mastered locomotion, and that's it.
None of these robots can hang our laundry to dry, de-bone a chicken or cook a meal. They cannot perform any of the thousands of tasks we perform every day that require dexterity. They have just about mastered locomotion, and that's it.
But the question is, why build a humanoid robot? It's much harder to balance (especially if it should be extendable with modules), probably won't be able to handle harsher terrain than four-legged models in the foreseeable future, and I don't see how they can do tasks significantly better than a simpler model with a few robot arms.
The stature isn't the main reason for the versatility of humans. I guess humanoid robots would be useful in situations where familiarity is important, but Tesla probably wants more than just a premium customer greeter with integrated voice assistant.
The stature isn't the main reason for the versatility of humans. I guess humanoid robots would be useful in situations where familiarity is important, but Tesla probably wants more than just a premium customer greeter with integrated voice assistant.
Our indoor environments are built for humans. For instance, robots with wheels would struggle with stairs. Short robots would have trouble reaching up and grabbing things from shelves. Wide robots would struggle to get down narrow hallways.
The four-legged Boston Dynamics robot can walk up/down stairs, is pretty slim and has a robot arm able to reach a bit higher. And there's still a lot of room for improvement.
Even humans can't always reach the highest shelves without a ladder. I'd like to see a humanoid robot successfully climb one up and down while holding an object; that alone would be an insane achievement.
A good humanoid robot costs a ton, to a point where I doubt it'll see much use in normal households. I'd love to be wrong in that aspect, but I haven't yet seen an affordable robot beyond hobby project standards. Imho a humanoid body would be a nice bonus once everything else is working reliably and there's enough demand, not what I'd start out with.
Even humans can't always reach the highest shelves without a ladder. I'd like to see a humanoid robot successfully climb one up and down while holding an object; that alone would be an insane achievement.
A good humanoid robot costs a ton, to a point where I doubt it'll see much use in normal households. I'd love to be wrong in that aspect, but I haven't yet seen an affordable robot beyond hobby project standards. Imho a humanoid body would be a nice bonus once everything else is working reliably and there's enough demand, not what I'd start out with.
Fair points. However, I think the biggest problem with a short robot would be people tripping over it. Even if it had a big 'ol arm sticking out like a robotic giraffe, it'd be hard to gauge the dimensions of the base, especially given it may not be clear what direction it is facing if the arm can reticulate about as it pleases.
So, a robot in a human environment must be about human height and width, but also its width should be roughly uniform from top to bottom like a human.
I suppose you could build a quadruped of those dimensions, but it's hard to imagine what it'd look like. I am not aware of any creatures like that occurring in nature, likely with good reason.
So, a robot in a human environment must be about human height and width, but also its width should be roughly uniform from top to bottom like a human.
I suppose you could build a quadruped of those dimensions, but it's hard to imagine what it'd look like. I am not aware of any creatures like that occurring in nature, likely with good reason.
> However, I think the biggest problem with a short robot would be people tripping over it.
Somehow that has not turned out to be a major problem with the closest thing we already have in widespread "use": Dogs.
I think we can agree that humans are pretty good at not stepping on a retriever-sized thing by accident. Not to mention that these robots will probably not be very quiet (servos can be pretty noisy). So if that's the biggest problem, we'd be in pretty good shape.
Somehow that has not turned out to be a major problem with the closest thing we already have in widespread "use": Dogs.
I think we can agree that humans are pretty good at not stepping on a retriever-sized thing by accident. Not to mention that these robots will probably not be very quiet (servos can be pretty noisy). So if that's the biggest problem, we'd be in pretty good shape.
> For instance, robots with wheels would struggle with stairs.
Just ask the Daleks.
Just ask the Daleks.
> why build a humanoid robot?
For one, humanoid robots are a bit less creepy than the alternatives. Have you seen the BD robots?
That’s not as important for industrial use, though.
For one, humanoid robots are a bit less creepy than the alternatives. Have you seen the BD robots?
That’s not as important for industrial use, though.
The reason you want humanoid robots is because environments that have humans in them are generally built around human ergonomics.
Well, you gotta start somewhere. I'm getting tired of people dismissing any effort of changing the world.
These are the same people that were laughing at him when he launched tesla. The same people whole said Apple was doomed, because they only had a smartphone. Etc. etc. etc.
These are the same people that were laughing at him when he launched tesla. The same people whole said Apple was doomed, because they only had a smartphone. Etc. etc. etc.
"I'm getting tired of people dismissing any effort of changing the world."
I am getting tired of people praising Elon joke projects as world-changing.
Its a voice assistant, we already have them, there is nothong world changing. Also Hyperloop is and Boring company are pretty useless, and are likely to stay that way - passenger capacity of every hyperloop design and of the tonnela are pethetic compared to a normal train
I am getting tired of people praising Elon joke projects as world-changing.
Its a voice assistant, we already have them, there is nothong world changing. Also Hyperloop is and Boring company are pretty useless, and are likely to stay that way - passenger capacity of every hyperloop design and of the tonnela are pethetic compared to a normal train
I note your omitting: falcon9, starlink, model3, and powerwall from your assessment that Elon’s ideas aren’t world changing.
And some were laughing as he announced full autonomous driving next year ... five years ago.
Or lets talk about his Mars settlement.
There a things that are realistic but people thought it's too expensive or too niche and there are things that sound like straight out of science fiction.
Musk is good in the former and not so good in the latter.
Musk is good in the former and not so good in the latter.
I think it's just lazy to dismiss everyone's criticism as one thing. Sure, there are idiots out on the internet who are going to crap all over what anyone does for literally any reason. However, the people are criticising are actual people and they do actually have arguments that you should think about. Some of those arguments are bad, but a lot of them aren't. When Elon Musk decided he was going to revolutionize travel with the Boring company a load of experts in town planning and infrastructure came out and said - "Hey, we're experts, we understand this problem, you aren't even solving the right problem" and so far, they seem basically correct. The boring company could completely succeed at its goal, and it wouldn't matter.
There are going to be lightweight criticisms of this concept (Musk invites those by making someone dress up in a morph suit and dance like a robot). But there are also going to be more realistic criticisms such as the fact that whilst Humanoid robots have remained science fiction, non-humanoid robots have proliferated massively across industries that were previously fully manual. It'd be great if we could build a humanoid robot to replace all those humans trudging around Amazon fulfillment centres- but we don't need to, because those robots have no reason to be humanoid in the first place.
I'm tired of people pretending that just the aspiration to change the world excuses you from real criticism of your ideas. There are plenty of people out there changing the world- some of them mentioned in the article, and those people actually thought out what they were doing before getting a guy in a morph suit to dance in front of the worlds press. Let's not forget, we're still waiting for the autonomous driving promises Musk has failed to deliver on for years now.
There are going to be lightweight criticisms of this concept (Musk invites those by making someone dress up in a morph suit and dance like a robot). But there are also going to be more realistic criticisms such as the fact that whilst Humanoid robots have remained science fiction, non-humanoid robots have proliferated massively across industries that were previously fully manual. It'd be great if we could build a humanoid robot to replace all those humans trudging around Amazon fulfillment centres- but we don't need to, because those robots have no reason to be humanoid in the first place.
I'm tired of people pretending that just the aspiration to change the world excuses you from real criticism of your ideas. There are plenty of people out there changing the world- some of them mentioned in the article, and those people actually thought out what they were doing before getting a guy in a morph suit to dance in front of the worlds press. Let's not forget, we're still waiting for the autonomous driving promises Musk has failed to deliver on for years now.
> "Hey, we're experts, we understand this problem, you aren't even solving the right problem" and so far, they seem basically correct. The boring company could completely succeed at its goal, and it wouldn't matter.
A lot of experts in cars and rockets claimed that Tesla and SpaceX are terrible ideas and they were proved wrong many times. Incumbents are too often dismissive of new ideas.
Many of the arguments against the boring company are "social arguments" from people who are ideologically committed to bikes and public transport. They have very little to do with whether it is possible to build cheaper tunnels.
A lot of experts in cars and rockets claimed that Tesla and SpaceX are terrible ideas and they were proved wrong many times. Incumbents are too often dismissive of new ideas.
Many of the arguments against the boring company are "social arguments" from people who are ideologically committed to bikes and public transport. They have very little to do with whether it is possible to build cheaper tunnels.
I think you're missing the point - which was to engage on the actual criticism and not just hand wave about "Oh lots of people said stuff". Sure, in any given case you can find someone making a bad argument on the internet, it doesn't mean there aren't good arguments, and in many cases if you listen to the good arguments, you realise what Musk is taking years to realise.
> A lot of experts in cars and rockets claimed that Tesla and SpaceX are terrible ideas and they were proved wrong many times.
Can we see some references to these experts?
Can we see some references to these experts?
Look up anything Toyoda says about electric cars.
Why do you act like we have not already started?
You are making the exact same dismissive mistake Elon is making: Just ignoring the existing work, or assuming it is somehow just shoddy.
You are making the exact same dismissive mistake Elon is making: Just ignoring the existing work, or assuming it is somehow just shoddy.
>With rockets, well, we already had rockets.
The author uses naivety, but this statement is very questionable to anyone interested in spaceflight. No, "we" didn't have any vertically landing reusable rocket capable of autonomous reentry, except one primitive atmospheric demo vehicle with manual primary controls. Let alone engines such as Raptors in the western world.
Anyhow, the presentation itself didn't seem to take itself very seriously, so I wouldn't think too much about the (very difficult) feasibleness of this, they must be aware. Hell, it may just be an elaborate advertisement for jobs.
The author uses naivety, but this statement is very questionable to anyone interested in spaceflight. No, "we" didn't have any vertically landing reusable rocket capable of autonomous reentry, except one primitive atmospheric demo vehicle with manual primary controls. Let alone engines such as Raptors in the western world.
Anyhow, the presentation itself didn't seem to take itself very seriously, so I wouldn't think too much about the (very difficult) feasibleness of this, they must be aware. Hell, it may just be an elaborate advertisement for jobs.
If I remember correctly Elon Musk actually said "We had no idea what we were doing" while talking about development of the first Dragon cargo capsule, so I guess he would be the first to admit that. That doesn't mean that they are not even going to try. You see that happening over at SpaceX with Starship development right now: start by trying a best guess ("Starhopper" was like a water tower with massive legs and an engine) and then build, test, improve, fail occassionally,...
Starhopper wasn't even the first design, before that they tried composites and had to scrap that. It really is trial and error.
Nobody knows for sure if Hyperloop, Self driving Teslas, Neuralink, Starship, Cybertruck, Starlink, etc. are going to work out.
But we know with 100% certainty we'll never find out unless we try.
But we know with 100% certainty we'll never find out unless we try.
Extrapolating my father's Model S purchase into this domain:
"look I spent $50k on this android - it makes fart noises and tries to jump off the pavement into a bush every few days"
"look I spent $50k on this android - it makes fart noises and tries to jump off the pavement into a bush every few days"
I wasn’t sold before, but I think you’ve just sold me on it.
Long term elderly care facility manned by Tesla Bot and Hyundai Bot team polishing Boston Dynamics's early prototypes springs immediately to mind.
Sounds like a nightmare for the residents!
"That will never work" is probably what has been said the most about Musk projects.
It's also correct a lot of the time.
Hyperloop didn't work out and the remaining related projects are looking more and more like scams. In the last few years no noteworthy progress has been made and even the most basic questions haven't ever been answered.
The Boring Company, which ripped off the original The Boring Company, has one finished project which didn't get close to the original vision and cost enough that further projects were put on hold.
Neuralink is cool but Musk made some really wild claims in their name without much to back them up.
SpaceX and Tesla do some amazing stuff and hopefully they keep being successful. But that doesn't say anything about the rest.
Hyperloop didn't work out and the remaining related projects are looking more and more like scams. In the last few years no noteworthy progress has been made and even the most basic questions haven't ever been answered.
The Boring Company, which ripped off the original The Boring Company, has one finished project which didn't get close to the original vision and cost enough that further projects were put on hold.
Neuralink is cool but Musk made some really wild claims in their name without much to back them up.
SpaceX and Tesla do some amazing stuff and hopefully they keep being successful. But that doesn't say anything about the rest.
Many of those side bets that "don't work yet" are much newer than Tesla and SpaceX and have had much less effort put into them. Claims that "they will never work" are predicting the future based mostly on personal biases.
For example Hyperloop is just a concept that other people are trying to implement with very limited resources.
For example Hyperloop is just a concept that other people are trying to implement with very limited resources.
> For example Hyperloop is just a concept that other people are trying to implement with very limited resources.
Virgin Hyperloop raised $160 million from investors and all they can show after 7 years is a successful trip through the massively downscaled 500m test track, at a fraction of the advertised speed.
So far nobody could show that it can compete with standard high-speed trains, how those by far largest vaccum chambers are supposed to be maintained and protected, or how the metal tube's massive temperature-dependent expansions are going to be handled. Those are huge engineering challenges and all they seemed to focus on so far are some public partnerships and interior designs.
Musk was wise to get off that train.
Virgin Hyperloop raised $160 million from investors and all they can show after 7 years is a successful trip through the massively downscaled 500m test track, at a fraction of the advertised speed.
So far nobody could show that it can compete with standard high-speed trains, how those by far largest vaccum chambers are supposed to be maintained and protected, or how the metal tube's massive temperature-dependent expansions are going to be handled. Those are huge engineering challenges and all they seemed to focus on so far are some public partnerships and interior designs.
Musk was wise to get off that train.
"Claims that "they will never work" are predicting the future based mostly on personal biases."
No, there are transport and contruction proffeshionals explaining that economics of the Hyperloop cannot work, it requires more effort to build and has 100x less capacity and worse safety than any bullet train.
The Boring Company doesn't even make sence, how are they making tonnels cheaper? Have they achieved anything at all? There is no 'core of innovation' behind it
No, there are transport and contruction proffeshionals explaining that economics of the Hyperloop cannot work, it requires more effort to build and has 100x less capacity and worse safety than any bullet train.
The Boring Company doesn't even make sence, how are they making tonnels cheaper? Have they achieved anything at all? There is no 'core of innovation' behind it
Boring corp, modified, will be immensely valuable on Mars, for digging habitats / etc.
We're still waiting for Automatic Self Drive, not long ago he acknowledged it's harder than he thought it would be. Musk is a good marketeer though, he knows how to make headlines. A Tesla Bot will also likely be harder than he expected.
This is a project they haven't done anything about yet, from the presentation it looks like they just came up with the idea that they would work on this within a week before.
Musk's fans give him too much credit for announcing future work.
Musk's fans give him too much credit for announcing future work.
I wouldn't be surprised if they get to the point of go get me an X from the fridge in a year but really the tipping point of usefulness will be go make me some properly seasoned scrambled eggs. Very few jobs are constrained to just go get x or help move Y so they will run into same problem as other current robot companies finding a customer. Spot is quite capable yet amazon hasn't replaced its pickers with it so what is the key feature musk needs to implement?
One foreseeable use:
Sending astronauts to Mars surface and bringing them home is energetically horrific. It's not so much the getting to Mars orbit bit that's hard work (at least, from Earth orbit), as descending into a gravity well and then getting back out of it again.
But current Mars robots have a huge control lag due to speed of light delays (up to 30 minutes between Earth/Mars, depending on where both planets are). It'd be easier to explore the surface using robots if they were controlled by astronauts in orbit -- say, in a small space station dug into the surface of Phobos or Deimos for radiation shielding.
And the easiest robots for a human to control directly using a conceivable circa-2030s AR rig are either a car-like machine or a human-body-shaped machine.
You could trial such a device on the Moon (as part of the Starship HLS plans NASA is currently backing for the Artemis Program), as an adjunct to actual astronauts on EVA (which is always a risky task). And the software for a human-controlled humanoid robot would be vastly easier to develop than the sort of fully autonomous control system Boston Dynamics is aiming for, hence (I suspect) Musk's over-optimistic projection.
TLDR: I think Musk is trolling Boston Dynamics et al, and his real goal is something a lot more practical. (Either that or he's been hitting the magic mushrooms again.)
Sending astronauts to Mars surface and bringing them home is energetically horrific. It's not so much the getting to Mars orbit bit that's hard work (at least, from Earth orbit), as descending into a gravity well and then getting back out of it again.
But current Mars robots have a huge control lag due to speed of light delays (up to 30 minutes between Earth/Mars, depending on where both planets are). It'd be easier to explore the surface using robots if they were controlled by astronauts in orbit -- say, in a small space station dug into the surface of Phobos or Deimos for radiation shielding.
And the easiest robots for a human to control directly using a conceivable circa-2030s AR rig are either a car-like machine or a human-body-shaped machine.
You could trial such a device on the Moon (as part of the Starship HLS plans NASA is currently backing for the Artemis Program), as an adjunct to actual astronauts on EVA (which is always a risky task). And the software for a human-controlled humanoid robot would be vastly easier to develop than the sort of fully autonomous control system Boston Dynamics is aiming for, hence (I suspect) Musk's over-optimistic projection.
TLDR: I think Musk is trolling Boston Dynamics et al, and his real goal is something a lot more practical. (Either that or he's been hitting the magic mushrooms again.)
The previous thread on this topic actually laid out a plausible general "Musk strategy"
(edit: here https://news.ycombinator.com/item?id=28243811 )
IIRC it went some thing like:
1) Declare a ludicrous "pie in the sky" / "City on Mars" goal.
2) Use that to inspire, hire clever but starry eyed people who will work 7 days a week in service of that.
3) Iterate rapidly on do-able things in the general direction of the goal.
4) Reach some some point that's far from the original goal, but is still a huge achievement (e.g. re-usable rockets to LEO).
5) Profit!
6) Use the profits to start again, with another field that has synergy to the last (e.g. if you have electric cars, then add solar panels and battery packs)
In terms of humanoid robots this makes sense, the "10% to the goal" point is useful, and ties into the space plan. Robotics and space have synergies.
Note: original post calls the big goal "relatable, unoriginal" - this is a feature not a bug: to inspire, pick something that people already understand. Something that you don't have to explain.
(edit: here https://news.ycombinator.com/item?id=28243811 )
IIRC it went some thing like:
1) Declare a ludicrous "pie in the sky" / "City on Mars" goal.
2) Use that to inspire, hire clever but starry eyed people who will work 7 days a week in service of that.
3) Iterate rapidly on do-able things in the general direction of the goal.
4) Reach some some point that's far from the original goal, but is still a huge achievement (e.g. re-usable rockets to LEO).
5) Profit!
6) Use the profits to start again, with another field that has synergy to the last (e.g. if you have electric cars, then add solar panels and battery packs)
In terms of humanoid robots this makes sense, the "10% to the goal" point is useful, and ties into the space plan. Robotics and space have synergies.
Note: original post calls the big goal "relatable, unoriginal" - this is a feature not a bug: to inspire, pick something that people already understand. Something that you don't have to explain.
While Musk lost some credibility over cryptocurrency related topics, it is important that he manages to repackage these already "thought of" stuff in an almost Steve Jobs sort of fashion and insert into the public consciousness. Apart from the hiring, many companies start up just to get acquired by Tesla, or to compete with them. Which creates the much needed innovation by itself.
I think he uses his media power for a good purpose even if he himself or his company cannot reach such goals, he creates the stepping stones in a future that I really like to see happen.
I think he uses his media power for a good purpose even if he himself or his company cannot reach such goals, he creates the stepping stones in a future that I really like to see happen.
Just for reference, you’re extrapolating all the way to things like digging holes in moons of other planets from a starting point of a guy dancing around a stage in a white spandex outfit.
Yes, yes I am. (Failure is always an option!)
However, I think digging holes for radiation shielding in a smallish asteroid-like body -- especially if it's a gravel pile similar to 101955 Bennu -- may prove easier than landing, provisioning, and returning human beings from the Martian surface.
(101955 Bennu, target of the OSIRIS-REx probe, turned out to be a rubble pile. It's an order of magnitude smaller than the Martian moons, which are nevertheless also believed to be rubble piles -- they're not large enough to have significant surface gravity, let alone enough to melt/fuse their rocks.)
However, I think digging holes for radiation shielding in a smallish asteroid-like body -- especially if it's a gravel pile similar to 101955 Bennu -- may prove easier than landing, provisioning, and returning human beings from the Martian surface.
(101955 Bennu, target of the OSIRIS-REx probe, turned out to be a rubble pile. It's an order of magnitude smaller than the Martian moons, which are nevertheless also believed to be rubble piles -- they're not large enough to have significant surface gravity, let alone enough to melt/fuse their rocks.)
But a humanoid is a pretty useless form factor for that kind of work. Humanoid robots are good in the kind of environments where humans evolved, or that they built for themselves.
Asteroid mining is pretty far from that.
Asteroid mining is pretty far from that.
You misunderstood.
Astronauts in Mars orbit need radiation shielding. So: dig their spacecraft/habitat into one of the Moons.
Humanoid robots: put them on Mars, instead of landing astronauts. That way, you don't need to bring them home afterwards. Meanwhile, with a good AR/VR rig, the astronauts (dug into the Phobos base) can control them remotely with no lag.
Astronauts in Mars orbit need radiation shielding. So: dig their spacecraft/habitat into one of the Moons.
Humanoid robots: put them on Mars, instead of landing astronauts. That way, you don't need to bring them home afterwards. Meanwhile, with a good AR/VR rig, the astronauts (dug into the Phobos base) can control them remotely with no lag.
But why humanoid robots, rather than robots with a shape that is actually optimised for the task?
I'm now picturing "humanoid" only in the loosest sense, e.g. a head and arms only, for remote operation by a human inside the Mars (or Phobos) base, mounted on the outside of a cybertruck, like some Terminator Centaur.
The operator would be able to toggle between "drive the truck" and "pick up rocks" modes, depending on which shape is appropriate for the task at hand.
The operator would be able to toggle between "drive the truck" and "pick up rocks" modes, depending on which shape is appropriate for the task at hand.
But that is not what was announced, and very little of the hard problems involved in building an actual humanoid robot like the one Musk announced are relevant to building what you suggest.
> But that is not what was announced
yes, so what? We expect that.
Previous comment here touches upon the connection and inevitable gap between the inspirational BS "what was announced" and final outcomes.
https://news.ycombinator.com/item?id=28255768
yes, so what? We expect that.
Previous comment here touches upon the connection and inevitable gap between the inspirational BS "what was announced" and final outcomes.
https://news.ycombinator.com/item?id=28255768
Robots digging holes on other planets is already a very real thing. It has been done so many times that it's boring by now.
It is easier than dancing and genuinely viable as a product, with that viability increasing with the degree of autonomy.
It is easier than dancing and genuinely viable as a product, with that viability increasing with the degree of autonomy.
Yes I thought something similar. Basically, whenever Musk announces anything, I ask myself “how does this relate to Mars?”
Mostly it’s pretty obvious. Tunnel boring machines to dig hyperloop tunnels to take passengers from the starship spaceport to the nearest city on Mars? Check. (Why isn’t he building hyperloops himself? No need to create a vacuum on Mars… but the vehicle research will be interesting)
Mostly it’s pretty obvious. Tunnel boring machines to dig hyperloop tunnels to take passengers from the starship spaceport to the nearest city on Mars? Check. (Why isn’t he building hyperloops himself? No need to create a vacuum on Mars… but the vehicle research will be interesting)
Tunnels are one of the most important things you could build on any planet or satellite like the moon.
Remember there is no atmosphere nor magnetic fields on other planets to protect you from cosmic rays, so a good solution is make people live underground.
Some time ago a teacher offered me to make my Master thesis on that(I did not take it) because there were people already working on this problem.
BTW, places like the moon are very interesting. Because there is no fluid erosion every single grain of sand is extremely abrasive.
Remember there is no atmosphere nor magnetic fields on other planets to protect you from cosmic rays, so a good solution is make people live underground.
Some time ago a teacher offered me to make my Master thesis on that(I did not take it) because there were people already working on this problem.
BTW, places like the moon are very interesting. Because there is no fluid erosion every single grain of sand is extremely abrasive.
Tesla is an infrastructure company. He’s building infrastructural components for Mars as much as Earth, for sure.
Deploying robots on the moon would be extremely useful even if they have minimal autonomy and have to be remote controlled from Earth.
All current efforts are rovers with scientific instruments, nobody has yet attempted to build something capable of clearing rocks for a launch pad let alone mounting solar panels.
This needs to move on from research projects into actual series production.
All current efforts are rovers with scientific instruments, nobody has yet attempted to build something capable of clearing rocks for a launch pad let alone mounting solar panels.
This needs to move on from research projects into actual series production.
The last thing you need for those things is humanoid robots.
Especially with the 2.6 second lightspeed delay in any remote control inputs.
Especially with the 2.6 second lightspeed delay in any remote control inputs.
What will the Bot be used for? I can see why it'll be useful in the future when we have AGI. But until then, I don't see what advantage it'll have over things built specifically for a particular job, which is already hard enough to achieve.
Not clear that hammering nails or digging holes in the ground requires AGI.
You'd be surprised how much "deal with unpredicted stuff" construction work entails.
Indeed, robots already can hammer nails into walls. Or dig holes. But we still pay construction workers to do these things (well, usually operating tools/machines - but we decidedly don't let those devices operate themselves). And it turns out that the results of them following orders too literally (like a robot would) tends to lead to funny (or tragic) videos clips.
See also this recent HN article, which discusses (the failures of) automating a much more restricted task in a seemingly straightforward domain: https://news.ycombinator.com/item?id=28054606
Indeed, robots already can hammer nails into walls. Or dig holes. But we still pay construction workers to do these things (well, usually operating tools/machines - but we decidedly don't let those devices operate themselves). And it turns out that the results of them following orders too literally (like a robot would) tends to lead to funny (or tragic) videos clips.
See also this recent HN article, which discusses (the failures of) automating a much more restricted task in a seemingly straightforward domain: https://news.ycombinator.com/item?id=28054606
Tesla could bypass a lot by training the robots by having humans possess them and control them to train them to do simple tasks[0]. Easy data.
https://japantoday.com/category/tech/japanese-convenience-st...
https://japantoday.com/category/tech/japanese-convenience-st...
I think the title is spot on, but I also think you need to add this to the title: "...but Elon will figure out faster than anyone else"
I think Elon Musk is exploring new stuff to do at Tesla to keep talented AI people in the company and not make them go away and search for new exciting problems to solve.
Even if the robot will never be real, he can keep these people excited and maybe gather new talent
Even if the robot will never be real, he can keep these people excited and maybe gather new talent
Also it's kind of an interesting project that can probably be worked on at fairly low cost given it uses the car hardware. I get the impression it's more a case of let's see if we can hack this stuff into a bot rather than a commercial plan.
One thing is certain: it can’t (yet) be used as a firefighter.
That is, if it shares the AI stack with Tesla AP.
That is, if it shares the AI stack with Tesla AP.
I like Elon Musk even though I have no expectation that this will work. I like that he is willing to try. Most billionaires in the world are incredibly boring. They just keep doing what they've been doing trying to earn more money. Elon Musk is a style of billionaire that other billionaires should aspire to - "I guess we'll try to make humanoid robots now."
Maybe if they could get it to a reasonable price point and get it to follow someone they could sell some of these as novelty type thing.
Maybe if they could get it to a reasonable price point and get it to follow someone they could sell some of these as novelty type thing.
cyberpsybin(5)
“It's almost like the dev who worked on the feature left, and nobody has taken over that duty yet.”
That was two years ago.