Humanoid Robots for Moon and Mars

[JohnFornaro]

I eagerly await the first demo where Optimus is handed a shovel and told to dig a ditch...

Bingo. 

Included in this exercise would be removing big rocks and the final steps of covering up the trench. 

1.)  Do this on Earth first.

2.)  "I eagerly await the first demo where Optimus [does a job way better suited to an excavator, both mechanically and economically]."

3.) Pack a lunch, because you'll be waiting a while. I doubt this "demo" is a high priority within the R&D program. It certainly won't be a priority to use Optimus that way in production. There's no ditch-digging company out there champing to replace their excavators with Optimus.

4). Light shoveling in the garden? Sure. "Dig this hundred meter ditch?" You're crazy if you don't hire an excavator for a few hundred bucks, rather than risk your $25,000 robot. Even just the wear-and-tear on joints makes it uneconomical for production use.

5). If I didn't know any better, I'd say certain folks prefer setting their goalpost to a demo they know isn't a real priority for the company...

1.) All these Optimii need to demonstrate their skill and utility on Earth first.  The general body design probably should be the humanoid design. Changing gravity from Earth to Moon to Mars doesn't come into play, because the geometry of the bot doesn't need to change.  IOW, granting all the functionality to the bot on Earth is the hardest goal, not putting the bot in slightly different gravitational environments.

2.)  Here, the term "excavator" could apply to several machine designs.  A human digging a 100 yard ditch is every bit an "excavator" as is the one mining coal. People here are either lazily or deliberately including differently scaled versions of "excavator" in their speculations.

3.) From what I've observed, the various "demos" are underwhelming. While folding towels might be an algorithmic and robotic challenge, the task itself is relatively useless on the Moon or Mars.  And indeed, there are no Optimii in the terrestrial excavator marketplace, nor likely to be any at all for the foreseeable future.

4.) "Risking one's robot" is a false risk.  Nobody would mismatch the robot and the task. 

5.) People here are playing "fast and loose" with  the selection of goalposts.

[Cheapchips]

Interesting look at humanoid form factors from one of Agility Robotics' AI staff, 'How Human Should Your Humanoid be?'

https://itcanthink.substack.com/p/how-human-should-your-humanoid-be?r=3uj5js&utm_campaign=post&utm_medium=web&triedRedirect=true

[Twark_Main]

Ironically, if there's a straw man here, it's your "eagerly awaiting" a ditch-digging Optimus demo (with the heavy implication that it's doomed otherwise), when you know full well that it's a silly use case for humanoid robots and isn't a real priority in the R&D process.

...I'm getting really sick of people on this forum accusing me of saying things that I didn't say.

Reading Coastal Ron's brand of goalpost-always-just-out-of-reach pessimism and your agreement, it seemed clearly implied that you endorsed the pessimism, or at least you don't object enough to say anything.

I should've considered that you just enjoy watching anything with a shovel. I shouldn't have assumed.

Enough about past misgivings. I was wrong. So what is your take on the generally pessimistic view of Optimus as expressed by Coastal Ron?

[InterestedEngineer]

As a reminder, Elon Musk has stated that the Tesla Optimus robot will have these characteristics:

Weight: ~125 lbs (57 kg).
Height: ~5'8" (1.73m).
Payload: ~45 lbs (20 kg)

Oh, and Tesla did not demo the latest version of Optimus of CES. Plus Musk earlier in the year mentioned a goal of producing “5,000 to 10,000 Optimus robots in 2025.” No evidence that happened.

I mention all of this so that expectations can be calibrated for what could end up being sent on the first Starship planned to land on Mars. I have no doubt that Musk will send whatever the latest version of Optimus is to Mars, but I doubt that we can expect it to be very useful.

I don't know if Boston Dynamics Atlas (i.e. Hyundai) wants to get into space exploration, but they were at CES demoing Atlas, and Atlas looks pretty interesting. Not saying Atlas could do much on Mars either, since I don't think humanoid robots are getting "smarter" that quickly, but at least Atlas looks like it is going into serial production.

I eagerly await the first demo where Optimus is handled a shovel and told to dig a ditch

Quoting myself to remind everyone to blame ME for the idea, not some innocent bystander.

To be fair, I also eagerly await someone investing in low volume manufacturing plant and a design for Mars-capable excavators to dig those ditches  Not to mention the large vacuum chamber with dirt for testing it.  Either with its own completely custom automation or operated by an Optimus.

Which one do you think we'll be waiting for the longest?  One is some software/neural net stuff and shovels whose company is controlled by The Guy trying to get us to Mars, and a trip to the Mojave desert (or maybe the gully behind JPL).

The other is a billion dollar manufacturing plant for a plausible but distant future market whose volume is relatively low and has no earthly application as a backstop or driver.   And where training the AI is implausible because if we've learned ONE thing about AI is that it requires incredibly high volumes of varied data sets to train properly (you have learned that... right?   It's AI Econ 101 at this point.  I hope)

[Twark_Main]

I'm sure the bystanders appreciate it. 

large vacuum chamber with dirt for testing it

Such thermal-vacuum chambers already exist, but as InSight taught us you can play in a sandbox but that doesn't mean it's representative of the digging conditions on Mars.

billion dollar manufacturing plant for a plausible but distant future market whose volume is relatively low and has no earthly application as a backstop or driver. And where training the AI is implausible because if we've learned ONE thing about AI is that it requires incredibly high volumes of varied data sets to train properly (you have learned that... right?   It's AI Econ 101 at this point.  I hope)

I don't think the Mars-optimized Optimus (MOPtimus) will be a completely different product line with its own factory. I expect there will be significant commonality with Terrestrial Optimus (TOPtimus), in both hardware and software.

While I eventually do expect millions of Optimus bots on Mars, you don't need to build that factory today. It will take at least a decade, and I expect most of the mass to eventually be manufactured on Mars.

Probably they'll still import the AI chips for a while, but ultimately I expect Tesla wants an AI chip fab on Mars too.

[Paul451]

To be fair, I also eagerly await someone investing in low volume manufacturing plant and a design for Mars-capable excavators to dig those ditches
[...]
Which one do you think we'll be waiting for the longest?  One is some software/neural net stuff and shovels whose company is controlled by The Guy trying to get us to Mars, and a trip to the Mojave desert (or maybe the gully behind JPL).
The other is a billion dollar manufacturing plant for a plausible but distant future market whose volume is relatively low and has no earthly application as a backstop or driver.

Coincidentally, at that same CES, Hitachi Construction Machinery showed off their driverless earth-mover concept-vehicle. The mining industry is investing heavily in remote operated and autonomous vehicles, due to safety, labour costs and labour retention.

[Coastal Ron]

...
Enough about past misgivings. I was wrong. So what is your take on the generally pessimistic REALISTIC view of Optimus as expressed by Coastal Ron?

There, fixed it for you... 

[InterestedEngineer]

I'm sure the bystanders appreciate it. 

large vacuum chamber with dirt for testing it

Such thermal-vacuum chambers already exist, but as InSight taught us you can play in a sandbox but that doesn't mean it's representative of the digging conditions on Mars.

billion dollar manufacturing plant for a plausible but distant future market whose volume is relatively low and has no earthly application as a backstop or driver. And where training the AI is implausible because if we've learned ONE thing about AI is that it requires incredibly high volumes of varied data sets to train properly (you have learned that... right?   It's AI Econ 101 at this point.  I hope)

I don't think the Mars-optimized Optimus (MOPtimus) will be a completely different product line with its own factory. I expect there will be significant commonality with Terrestrial Optimus (TOPtimus), in both hardware and software.

While I eventually do expect millions of Optimus bots on Mars, you don't need to build that factory today. It will take at least a decade, and I expect most of the mass to eventually be manufactured on Mars.

Probably they'll still import the AI chips for a while, but ultimately I expect Tesla wants an AI chip fab on Mars too.

I think you read that upside down.

I'm talking about the factory to make AI automated vacuum-capable excavators in low volumes - a financially and technically infeasible idea that keeps coming uo.

I just assume as a matter of course that Optimus will be vacuum capable and produced in a high volume factories.  Given Elon talking (and more importantly investing) it's pretty much a given that's the plan.

Which one will dig the ditches is the real question.  Heck maybe someone should counter with tunneling robots where you collapse the tunnel (or just don't collapse tunnels), given Elon is actually working that.

[redneck]

[quote author=T

I'm talking about the factory to make AI automated vacuum-capable excavators in low volumes - a financially and technically infeasible idea that keeps coming uo.

I just assume as a matter of course that Optimus will be vacuum capable and produced in a high volume factories.  Given Elon talking (and more importantly investing) it's pretty much a given that's the plan.

Which one will dig the ditches is the real question.  Heck maybe someone should counter with tunneling robots where you collapse the tunnel (or just don't collapse tunnels), given Elon is actually working that.

I assume as a matter of course that you are unaware of the current availability of remote controlled machines that run on batteries. Excavators to dump trucks to front end loaders for starters. And the availability of custom configurations of them. Unlikely that Optimus will be able to muscle through a process available right now from custom machines.   Equipment will have to be vacuum adapted in either case, just much easier with existing equipment.

[clongton]

Interesting look at humanoid form factors from one of Agility Robotics' AI staff, 'How Human Should Your Humanoid be?'

https://itcanthink.substack.com/p/how-human-should-your-humanoid-be?r=3uj5js&utm_campaign=post&utm_medium=web&triedRedirect=true

He asked the question:

"Both of these robots are incredible pieces of hardware, but we must ask, why should Figure's robot look so human while Boston Dynamics opts for such a strange form factor?   Is it just a gimmick that Atlas can turn all the wat around? If we're moving away from the human form, why not just go all the way and make a robot that's fully optimized for its task ...?"

The answer is that we are not moving away from the human form. Atlas is still a human "form", but is optimized for greater dexterity and mobility. These 2 "forms" will serve different purposes in different working environments, each one optimized for its designated tasks.

Please, just don't put any skin, hair and clothing on any of them.
I don't want to see any of the BSG models.

[InterestedEngineer]

[quote author=T

I'm talking about the factory to make AI automated vacuum-capable excavators in low volumes - a financially and technically infeasible idea that keeps coming uo.

I just assume as a matter of course that Optimus will be vacuum capable and produced in a high volume factories.  Given Elon talking (and more importantly investing) it's pretty much a given that's the plan.

Which one will dig the ditches is the real question.  Heck maybe someone should counter with tunneling robots where you collapse the tunnel (or just don't collapse tunnels), given Elon is actually working that.

I assume as a matter of course that you are unaware of the current availability of remote controlled machines that run on batteries. Excavators to dump trucks to front end loaders for starters. And the availability of custom configurations of them. Unlikely that Optimus will be able to muscle through a process available right now from custom machines.   Equipment will have to be vacuum adapted in either case, just much easier with existing equipment.

I am aware of them.

How do you vacuum adapt something that requires a ton of rubber to work? (rubber outgasses quickly in a vacuum).

Who is going to remote control them from where?

[redneck]

[quote author=T

I'm talking about the factory to make AI automated vacuum-capable excavators in low volumes - a financially and technically infeasible idea that keeps coming uo.

I just assume as a matter of course that Optimus will be vacuum capable and produced in a high volume factories.  Given Elon talking (and more importantly investing) it's pretty much a given that's the plan.

Which one will dig the ditches is the real question.  Heck maybe someone should counter with tunneling robots where you collapse the tunnel (or just don't collapse tunnels), given Elon is actually working that.

I assume as a matter of course that you are unaware of the current availability of remote controlled machines that run on batteries. Excavators to dump trucks to front end loaders for starters. And the availability of custom configurations of them. Unlikely that Optimus will be able to muscle through a process available right now from custom machines.   Equipment will have to be vacuum adapted in either case, just much easier with existing equipment.

I am aware of them.

How do you vacuum adapt something that requires a ton of rubber to work? (rubber outgasses quickly in a vacuum).

Who is going to remote control them from where?

Machines don't necessarily need rubber. There are steel tracked and steel wheeled machines all over the place. Other than that, there is very little rubber in equipment that can't easily be replaced by metal. As it often is already in certain applications and environments.

Autonomous operation of a machine is the next step from remote. Massively easier to program a device to do what it normally does  than to expect a robot to operate it.

[Twark_Main]

...
Enough about past misgivings. I was wrong. So what is your take on the generally pessimistic REALISTIC view of Optimus as expressed by Coastal Ron?

There, fixed it for you... 

Gott'em!!!       


The fact that that's the part (and the only part  ) you could muster a reply for is very telling about the strength of your argument.

[Twark_Main]

I'm sure the bystanders appreciate it. 

large vacuum chamber with dirt for testing it

Such thermal-vacuum chambers already exist, but as InSight taught us you can play in a sandbox but that doesn't mean it's representative of the digging conditions on Mars.

billion dollar manufacturing plant for a plausible but distant future market whose volume is relatively low and has no earthly application as a backstop or driver. And where training the AI is implausible because if we've learned ONE thing about AI is that it requires incredibly high volumes of varied data sets to train properly (you have learned that... right?   It's AI Econ 101 at this point.  I hope)

I don't think the Mars-optimized Optimus (MOPtimus) will be a completely different product line with its own factory. I expect there will be significant commonality with Terrestrial Optimus (TOPtimus), in both hardware and software.

While I eventually do expect millions of Optimus bots on Mars, you don't need to build that factory today. It will take at least a decade, and I expect most of the mass to eventually be manufactured on Mars.

Probably they'll still import the AI chips for a while, but ultimately I expect Tesla wants an AI chip fab on Mars too.

I think you read that upside down.

I'm talking about the factory to make AI automated vacuum-capable excavators in low volumes - a financially and technically infeasible idea that keeps coming up.

The "factory" is just a machine shop.

"Vacuum-capable" is the only custom requirement here, but the digging software (which is the hard part) is common between Mars and terrestrial excavators.

I just assume as a matter of course that Optimus will be vacuum capable

I do not (that requirement would be silly), but again making one or even 100 custom units in a machine shop isn't difficult.

[Twark_Main]

Please, just don't put any skin, hair and clothing on any of them.
I don't want to see any of the BSG models.

The "skin" will just be a disposable / washable cover to keep dust out, so you should be fine.

[clongton]

Please, just don't put any skin, hair and clothing on any of them.
I don't want to see any of the BSG models.

The "skin" will just be a disposable / washable cover to keep dust out, so you should be fine.

That's fine, as long as it's not engineered to look like a human being.

[Coastal Ron]

"Both of these robots are incredible pieces of hardware, but we must ask, why should Figure's robot look so human while Boston Dynamics opts for such a strange form factor?   Is it just a gimmick that Atlas can turn all the wat around? If we're moving away from the human form, why not just go all the way and make a robot that's fully optimized for its task ...?"

The answer is that we are not moving away from the human form.

Regarding human form robots, it is important to remember that throughout the history of the Industrial Revolution humans have relied on non-human form mechanical inventions to dramatically increase human efficiency.

So we have a LOT of history with non-human form machines, but practically none for human form machines used to produce economic value.

Part of that could be that their manufacturers are still trying to figure out what jobs human form robots can actually do in the real world, and maybe also that they have been reluctant to put them into those jobs for reasons we (the public) don't know about.

For Mars, certainly a lot of non-humanoid heavy, medium, and light equipment will be required, and will likely be the most practical and efficient equipment for their jobs. But I would imagine that there will be lots of small jobs that here on Earth would normally be done by humans. Maintenance is one, and that will be especially important.

So could the machinery being sent to Mars be designed for humanoid robot maintenance?

Or, what other "important" tasks should humanoid robots be tasked to do?

[Twark_Main]

Interesting look at humanoid form factors from one of Agility Robotics' AI staff, 'How Human Should Your Humanoid be?'

https://itcanthink.substack.com/p/how-human-should-your-humanoid-be

He asked the question:

"Both of these robots are incredible pieces of hardware, but we must ask, why should Figure's robot look so human while Boston Dynamics opts for such a strange form factor? Is it just a gimmick that Atlas can turn all the way around? If we're moving away from the human form, why not just go all the way and make a robot that's fully optimized for its task ...?"

The answer is that we are not moving away from the human form. Atlas is still a human "form", but is optimized for greater dexterity and mobility.

Right. Restated, they're really asking "if you're making a humanoid robot more general-purpose, why not make it less general-purpose?"

[redneck]

So could the machinery being sent to Mars be designed for humanoid robot maintenance?

Or, what other "important" tasks should humanoid robots be tasked to do?

One of the crucial jobs that Optimus would perform on Mars and other destinations of interest is provision of photo ops. Product placement. Salesmanship targeted at buyers on Earth. Anything useful that gets done will be gravy.

[InterestedEngineer]

Autonomous operation of a machine is the next step from remote. Massively easier to program a device to do what it normally does  than to expect a robot to operate it.

That's probably the biggest arm wave we'll see all week.

AI requires two things:

1. Massive data centers to crunch the data
2. Massive data.  From large numbers of sources that have a huge variety of data.  For example all the Tesla cars out there with cameras


You can do (1), but i'm not sure where you'll get the billions from

(2) is impossible with low volume equipment in any reasonable time frame.