Humanoid Robots for Moon and Mars

[crandles57]

This is exactly what I have been talking about.

That they need to use Optimus here on Earth first, to validate a whole multitude of things about Optimus and humanoid robots in general. That they haven't been, to any great extent, tells me that Optimus (and the many other humanoid robots around the world) are not yet ready for doing "work" on a daily basis - something that would be required if Optimus is sent to Mars.

I could comment that that sounds like you haven't been paying attention to how much I have been agreeing with you. Maybe that hints at an attitude of it is an argument to be won rather than constructive dialog?

If I was taking the argument approach I might comment that encyclopedia require neutral point of view and Musk mentions Optimus will be sent and lots of people get hyped up about it so you have to throw in some caveats for balance but ... oh oops I just did.       

[crandles57]

The Starships are suitable as habitats for quite a while, and solar storms are not really an issue on Mars - the atmosphere does provide significant shielding against that.

So you are saying this from grokipedia is overblown?
https://grokipedia.com/page/SpaceX_ambition_of_colonizing_Mars#human-health-risks-and-mitigation

Human health risks during SpaceX's planned Mars missions encompass acute and chronic threats from interplanetary transit, surface operations, and long-term habitation, including radiation exposure, gravitational alterations, toxic regolith, and psychological strain. These challenges arise from the absence of Earth's protective magnetosphere and atmosphere, prolonged microgravity en route, and Mars' 0.38g environment

The atmosphere is mainly CO2 and nitrogen, not much hydrogen? I thought hydrogen was good at providing protection hence water and methane providing good shielding but maybe the thin CO2 and N2 is enough?

[Paul451]

Hydrogen rich materials are good radiation shields for ships/stations because they reduce the secondary radiation that comes from breaking up heavier atoms. (Which turns a single incoming high-energy particle into a shotgun blast of medium energy, but still dangerous, particles.) Gives you the best per-kg per sq_m protection. There's no other "magic" to hydrogen. If you don't care about mass, just shove more of anything in the way.

Hence, it doesn't apply to atmospheres, where you either have excess mass to also absorb the secondaries or you don't. Mars' atmosphere wouldn't be any better if it was pure hydrogen.

[sanman]

Hydrogen rich materials are good radiation shields for ships/stations because they reduce the secondary radiation that comes from breaking up heavier atoms. (Which turns a single incoming high-energy particle into a shotgun blast of medium energy, but still dangerous, particles.) Gives you the best per-kg per sq_m protection. There's no other "magic" to hydrogen. If you don't care about mass, just shove more of anything in the way.

Hence, it doesn't apply to atmospheres, where you either have excess mass to also absorb the secondaries or you don't. Mars' atmosphere wouldn't be any better if it was pure hydrogen.

I thought Hydrogen is supposed to be good as a moderator (hence the reason why H2O is used as a moderator in reactors)
Hydrogen = Proton = similar in mass to neutron, therefore kinetic collisions by neutrons with hydrogen re-distributes their kinetic energy more equally between them, which helps to slow down those neutrons a lot, especially across successive collisions.
Likewise, it works the same for proton radiation too (proton-proton collision)

[Vultur]

The Starships are suitable as habitats for quite a while, and solar storms are not really an issue on Mars - the atmosphere does provide significant shielding against that.

So you are saying this from grokipedia is overblown?
https://grokipedia.com/page/SpaceX_ambition_of_colonizing_Mars#human-health-risks-and-mitigation

Human health risks during SpaceX's planned Mars missions encompass acute and chronic threats from interplanetary transit, surface operations, and long-term habitation, including radiation exposure, gravitational alterations, toxic regolith, and psychological strain. These challenges arise from the absence of Earth's protective magnetosphere and atmosphere, prolonged microgravity en route, and Mars' 0.38g environment

The atmosphere is mainly CO2 and nitrogen, not much hydrogen? I thought hydrogen was good at providing protection hence water and methane providing good shielding but maybe the thin CO2 and N2 is enough?

Well, not so much overblown as generic enough to be meaningless. But that text also doesn't mention solar storms. Solar storms =/= GCR.

The importance of a magnetosphere is however massively overblown if you're inside even a very thin atmosphere.

[Twark_Main]

That they need to use Optimus here on Earth first, to validate a whole multitude of things about Optimus and humanoid robots in general.

They have been. There are videos.

That they haven't been, to any great extent

That's a very different (and very vaguely worded) goalpost.

You could interpret that Optimus doing  "great extent" of the work means that over 50% of all human labor is replaced by Optimus. I don't think that's what you mean, but the fact that your words could even be interpreted that way makes it clear that you've left yourself a loophole big enough to fly a jumbo jet through. 

tells me that Optimus (and the many other humanoid robots around the world) are not yet ready for doing "work" on a daily basis

Since you left yourself that "great extent" escape hatch, this observation could equally be explained by the fact that Optimus is currently in low-scale production. This clearly limits the "extent" of total Optimus work, and it also means the unit price of each Optimus is still rather high.

You also have (and correct me if I'm wrong) zero real-world knowledge of exactly how much Optimus testing Tesla is doing behind closed doors.

If Tesla wasn't doing any real-works testing of Optimus then I would agree that the project is doomed. However as written, this is actually just saying "Ron wants more videos."   

[Vultur]

That they need to use Optimus here on Earth first, to validate a whole multitude of things about Optimus and humanoid robots in general.

They have been. There are videos.

They are *testing* it. They don't appear to be *using* it.

[crandles57]

That they need to use Optimus here on Earth first, to validate a whole multitude of things about Optimus and humanoid robots in general.

They have been. There are videos.

They are *testing* it. They don't appear to be *using* it.

True but what is the best use for Tesla at present?

the estimated start date for the Tesla Optimus Gen 3 (V3) production line is widely projected for early 2026. While Tesla aims for pilot production and internal testing throughout 2025, mass production of the refined V3 design is slated for the following year.

I have no idea whether they have ~50 v2 and 10 v3 from pilot production or ~250 v2 and 20 v3. Whatever, clearly the AI is not yet fully developed, trained and tested so the best use for the ones they have got is clearly in training and developing the AI rather than putting the few they have to use in Tesla production lines.

It is quite possible that they need 3000 Optimus v3 in order to do 3000 years of training with them so about mid 2027 before they put them to use in Tesla factories and sales are going to be a while after that.

Or make up your own numbers as to what you think will be needed. 

[Twark_Main]

That they need to use Optimus here on Earth first, to validate a whole multitude of things about Optimus and humanoid robots in general.

They have been. There are videos.

They are *testing* it. They don't appear to be *using* it.

What's the (meaningful) difference?


"They're not using them to do work in a factory?" They are.

"They're not using them for-profit commercially?" That comes back to the current cost and production scale.

"They're not renting them to outside companies even if it isn't profitable yet?"  Why would they, and what difference would that make?


Functionally there's no difference between "using" and "testing" that helps Ron's argument.  If anything testing is the ideal version of "using," because it's been optimized and instrumented to best accelerate the Optimus R&D timeline.

[crandles57]

"They're not using them to do work in a factory?" They are.

What is your evidence for this? How many are used to displace how many production worker employees?

Yes we have seen videos of them at tesla factory doing things a human would likely be doing and that humans could probably do faster. The speed may not be a killer economically though it might be if it means you need more stations to do the task particularly if the space is in short supply.

The video provides evidence that they have tried training Optimus to do the task, it doesn't prove Optimus continues doing the task as opposed to moving on to many more training tasks as part of ascertaining the best/easiest/cheapest approach to training Optimus.

Similarly Musk saying things like they are doing tasks that human production workers do. I don't think he has added that they will continue to do these tasks.

So what is the evidence that Optimus continues to do work of production workers as opposed to doing some training to see how well different training methods work?

[Coastal Ron]

Interesting article on this topic:

Why Humanoid Robots and Embodied AI Still Struggle in the Real World | Scientific American

Article tagline:

General-purpose robots remain rare not for a lack of hardware but because we still can’t give machines the physical intuition humans learn through experience

This echos some of the same issues talked about by iRobot's founder:

Why iRobot’s founder won’t go within 10 feet of today’s walking robots - Ars Technica

For the Scientific American article I think it pretty much boils down to a similar issue that autonomous cars have to deal with, in that these autonomous systems don't have a 3d map of the physical world, and don't have the intuition the animal world does about how the physical world works. In the article they explain it this way:

To explain the gap, Meta’s chief AI scientist Yann LeCun has noted that, by age four, a child has taken in vastly more visual information through their eyes alone than the amount of data that the largest large language models (LLMs) are trained on. “In 4 years, a child has seen 50 times more data than the biggest LLMs,” he wrote on LinkedIn and X last year. Children are learning from an ocean of embodied experience, and the massive datasets used to train AI systems are puddles by comparison. They’re also the wrong puddle: training an AI on millions of poems and blogs won’t make it any more capable of making your bed.

As to how to train:

The second approach is simulation. In virtual environments, AI systems can practice tasks thousands of times faster than humans can in the physical world. But simulation runs into the reality gap. An easy task in a simulator can fail in reality because the real world contains countless tiny details—friction, squishy materials, lighting quirks.

That reality gap explains why a robot parkour star can’t wash your dishes.

As to when these issues can be dealt with:

Among those working on robotics, there is broad disagreement about how quickly that gap will close. In March 2025 Nvidia CEO Jensen Huang told journalists, “This is not five-years-away problem, this is a few-years-away problem.” In September 2025 roboticist Rodney Brooks wrote, “We are more than ten years away from the first profitable deployment of humanoid robots even with minimal dexterity.” He also warned of the dangers that robots pose because of a lack of coordination and a risk of falling. “My advice to people is to not come closer than 3 meters to a full size walking robot,” Brooks wrote.

Nvidia CEO Jensen Huang has a vested interest in forecasting solutions that are more near term, so I tend to ignore those types of prognostications from CEO's that have vested interests in what they are saying. And from what I've seen, I'd be surprised if the 10+ year estimate wasn't correct. I wish it would be sooner, but I just don't see progress in solving the issues holding back humanoid robots.

With all of that, if cargo ships are sent to Mars within the next 10 years, I have no doubt that humanoid robots will be among the cargo. But I also think there will be other non-humanoid robotic systems that will probably do most of the planned work. I think the humanoid robots will have a limited range of responsibilities, plus they will be gathering data to help the next generation of humanoid robots to go to Mars.

[StraumliBlight]

Impressive demonstration of robots understanding context.

[Cheapchips]

This echos some of the same issues talked about by iRobot's founder:

These are all interesting challenges.  There are fair retorts from other experts online to most of those, not just CEOs.  Rodney Brooks / Roomba are the Tory Bruno / ULA of the robot vacuum world after all. 

[Coastal Ron]

This echos some of the same issues talked about by iRobot's founder:

These are all interesting challenges.  There are fair retorts from other experts online to most of those, not just CEOs...

I think the only valid retort would be to show an actual humanoid robot, out in the wild, doing 24/7 work of some mild complexity.

Otherwise it's just more "all hat, no cattle". 

[Vultur]

That they need to use Optimus here on Earth first, to validate a whole multitude of things about Optimus and humanoid robots in general.

They have been. There are videos.

They are *testing* it. They don't appear to be *using* it.

What's the (meaningful) difference?

That there's currently no evidence that they are good enough to be useful in a real world situation (as opposed to a curated demo) even on Earth, much less Mars.

[Twark_Main]

I think the only valid retort would be to show an actual humanoid robot, out in the wild, doing 24/7 work of some mild complexity.

Otherwise it's just more "all hat, no cattle". 

That there's currently no evidence that they are good enough to be useful in a real world situation (as opposed to a curated demo) even on Earth, much less Mars.

Tomorrow not here yet. More at 11.

Thanks. I'll need these links in a few years when people are asking for proof whether anyone actually thought humanoid robots were unworkable.

The future called, and the conventional wisdom is that it's very unfair (and probably merits investigation) how Tesla and SpaceX could "skate to where the puck will be" while others were fearful and hesitant and debating. How cruel of you, to mislead Elon Musk's poor competitors into bankruptcy like this!    


Seriously though, "put up or shut up" is exactly what Tesla is doing. They don't waste time bothering trying to convince anyone, and neither will I. You do, however, forfeit your right to come crying back to me saying how unfair it is that nobody warned you about the huge impact of humanoid robots, or Tesla's impending dominance in the field. 

[clongton]

That there's currently no evidence that they ARE good enough to be useful in a real world situation (as opposed to a curated demo) even on Earth, much less Mars.

There was a time when there was no evidence that electricity would be good enough for a real world situation.
There was a time when there was no evidence that air flight would be good enough for a real world situation.
There was a time when there was no evidence that gasoline automobiles would be good enough for a real world situation.
etc. etc., etc.
In all of these cases, and for tens of thousands more, after considerable time and effort was expended, all of them became more than good enough for a real world situation.
Why should humanoid robots be any different? We're just not there yet.
My generation made orange juice by growing a tree, picking the orange, cutting it and squeezing the juice out. It took a little time.
People today are skeptical if they can't get everything quickly, like mixing orange powder into a glass of water.
So yea, there's currently no evidence that [humanoid robots] ARE good enough to be useful in a real world situation.
This is hard, expensive and time-consuming. So just give it some time. Patience, my young Padawan.

[Vultur]

But for the purpose of preparing for human arrival on Mars, the question isn't "will humanoid robots be useful someday".

If the humanoid robots are supposed to prepare things before humans land on Mars, they have to be ready *before humans are ready to land anyway*. Otherwise the humanoid robots become a *delay* to the schedule relative to the previous, no humanoid robots plan.

I am not questioning whether they'll "ever" be any good. I'm questioning whether they'll be good enough for totally autonomous (tens of minutes light lag) operations when SpaceX starts sending cargo missions - say the 2028/2029 synod. That launch window opens in 36 months. (The 2026 window, less than a year away, seems much less likely.)

[Twark_Main]

But for the purpose of preparing for human arrival on Mars, the question isn't "will humanoid robots be useful someday".

If the humanoid robots are supposed to prepare things before humans land on Mars, they have to be ready *before humans are ready to land anyway*. Otherwise the humanoid robots become a *delay* to the schedule relative to the previous, no humanoid robots plan.

If you have a plan with no robots that lands at the same time and has the same risk to human life, why would you ever consider robots in the first place?

At far as I know, such a plan doesn't exist. The plans I'm aware of use humanoid robots to lower risk to humans. So it's not "just" a delay even if Optimus is late, you're still buying the advantage of reduced risk to human life by using robots instead of no robots.

I am not questioning whether they'll "ever" be any good. I'm questioning whether they'll be good enough for totally autonomous (tens of minutes light lag) operations when SpaceX starts sending cargo missions - say the 2028/2029 synod. That launch window opens in 36 months. (The 2026 window, less than a year away, seems much less likely.)

An eternity in AI time.

Again nobody will try to convince you, least of all me. "It neither picks my pocket nor breaks my leg." I am satisfied to just have a better internal prediction of SpaceX plans.

[edzieba]

At far as I know, such a plan doesn't exist. The plans I'm aware of use humanoid robots to lower risk to humans.

There are no shortage of 'humans to Mars' plans that do not involve humanoid robots, including architectures both with and without Starship (from Von Braun's Mars Project to DRM 3.0, so spanning over half a century).
On the other hand, those involving humanoid robots is currently one suggestion that there may at some point be the concept of a plan.