Remix.run Logo
Lapra 2 days ago

Humanoid robots are probably never coming. The fact is - flesh and blood humans pay for their own upkeep. Wear-and-tear, particularly on a heavy lifting robot, would probably be their biggest cost and might always outweigh the cost savings.

LeifCarrotson 2 days ago | parent | next [-]

I've commissioned dozens of robot cells (6-axis industrial arms for manufacturing are old, proven tech) and the wear and tear costs have been inconsequential. Even a large arm like a Fanuc R2000iC only uses about $0.50 in electricity per hour, some cells use significant power for pneumatics (in particular, compressed air venturi vacuum generators).

A couple grand for gearbox rebuilds every few years, replacement vacuum cups or worn hard tooling as needed, troubleshoot electrical issues as they arise... and your quarter million robot cell ($60k of that is the robot, most of the rest is NRE labor) will only need one human instead of eight to spit out parts every 60 seconds for the next decade.

Unless you think the humanoid robots are going to wear out significantly faster than existing robots, wear and tear costs are negligible.

With tight process controls, turning a work cell that has multiple humans doing manual labor for material handling, fastening, inspection, labeling, etc. into one intelligent human keeping the automation well adjusted is a solved problem. Eliminating that last human - the one that makes decisions instead of moves materials - with a humanoid robot is going to take decades.

bbarnett 2 days ago | parent | prev | next [-]

We can make things that last for decades, we just choose not to. Planned obsolescence is a business strategy, as is rapid breakdown of things we buy.

A generic example, fridges could easily last 40 to 50 years without maintenance. They wouldn't be all that more expensive either. Volvo, and the B-52 bomber program showed this, with Volvo having some models unchanged for 20 years. The B-52 has been in service longer than most people have been alive.

Each time an early wear or failure point is found in the B-52, it is documented, fixed, and rolled out to all B-52s. Their ancient, but more reliable than newer bombers and require less maintenance.

We could do this for everything. Design a fridge, and after 10 years collect the failures and see how they broke. Keep selling the same fridge, the same parts, and eventually it's a rock.

We don't do this, companies don't do this, because it's not best for profit.

So my point is robot maintenance could be minor, and if it was purely a lease model, would remain minor... because the company would profit from lower overall maintenance costs.

Lastly, compare a robot to a car driving 100s of thousands of KM. I've driven new cars to 150000km with almost no failure of any kind (except brakes. tires). So maybe not as bad as thought.

rimunroe 2 days ago | parent | next [-]

> A generic example, fridges could easily last 40 to 50 years without maintenance. They wouldn't be all that more expensive either. Volvo, and the B-52 bomber program showed this, with Volvo having some models unchanged for 20 years. The B-52 has been in service longer than most people have been alive.

B-52s require regular inspections and maintenance just like any other aircraft. A fridge is less complicated, but it's still a machine. Even my grandfather's clock needed some work done every couple decades, and it didn't contain refrigerant, a compressor, fans, or have to deal with condensation.

bbarnett a day ago | parent [-]

Yes but how old are the B-52s? And how much maintenance?

All the parts that have been shown to wear rapidly, have been reengineered and updated across the fleet. Compared to newer platforms, it's a rock.

rimunroe a day ago | parent | next [-]

I was responding to you saying we could make fridges which could easily last 40-50 years without maintenance (and somehow not cost much more).

> Yes but how old are the B-52s? And how much maintenance?

They get heavy, months-long maintenance where they replace major components (often including major structural ones) every four years. Presumably there is more frequent minor maintenance before then. I don’t imagine their age of 70ish years is really relevant since I assume they’re approaching Ships of Theseus at this point.

bbarnett a day ago | parent [-]

I was responding to you saying we could make fridges

Well OK, but you starting out by discussing the B-52, so I responded in kind.

They get heavy, months-long maintenance where they replace major components

Unless the documentary I watched was wrong, and of course it could be, that isn't "just replace things because". Instead, it's "if those components need replacement".

Now, I do recall that sometimes they'll discover an early-wear component, and do fleet-wide upgrades to fix that flaw. But that's different than maintenance to replace worn parts, for of course all B-52s fly different missions, have different wear as a result.

My point is, some of those B-52s are being inspected, but not having much done to them, where as others a lot more.

I don't think you can really, fairly compared a long running platform like the B-52 with a newer aircraft. Not in terms of stability of the platform, because the concept here is "fixing engineering defects that exhibit early wear".

Whenever I buy a car, I attempt to never buy the first year of a model revision. I wait until near the end of that run, often 4 or 5 years in, as car manufacturers constantly update assembly and build to deal with parts they've seen as early-wear. This isn't really debatable in a meaningful way, it's simply what's done.

And that's my point. If you look at my original post, I specify that the way to get 'stable', is to keep the exact same platform, and improve early wear points.

The post I responded to said:

Wear-and-tear, particularly on a heavy lifting robot, would probably be their biggest cost and might always outweigh the cost savings.

So you can see why I was specifying how this can be mitigated.

But all of that said, we clearly know this really isn't true. While people will still use shovels, we now have backhoes for a reason. People could use horses, but we have cars for a reason. Mechanical replacements exist for almost everything, and the wear and tear is cost effective and worth it, even with the price of fuel, maintenance, and regular upkeep.

And, even along with the fact that companies engineer for planned obsolescence and forced replacement.

rimunroe 16 hours ago | parent [-]

> Well OK, but you starting out by discussing the B-52, so I responded in kind.

...because you used B-52s as an example in your argument that fridges could be made to run maintenance free for decades.

> Unless the documentary I watched was wrong, and of course it could be, that isn't "just replace things because". Instead, it's "if those components need replacement".

Yes, that's how maintenance works. Why would you have thought I was saying it was just for fun? The every-four-years maintenance (PDM) involves a deep inspection of the whole aircraft and replacing parts that are fatigued or otherwise showing signs of wear. I would assume they're doing this because they've found that four years is a good rate to find and replace parts that are going to fail well before they're likely to. Airplane parts wear out after use. Frames experience stress from flying and become fatigued.

Similarly, refrigerant can leak when seals fail. I imagine that on the span of decades outgassing might be a problem too. I haven't dealt with compressor failures personally, but they're mechanical components and I assume eventually lubrication gets to be a problem. Rubber gaskets degrade over time. Fungus/algae can grow in the lines used to drain condensation to the evaporation pan and eventually block them, causing the fridge to build up ice and eventually leak water. Fans will fail over time because bearings wear or insulation on windings degrade.

> My point is, some of those B-52s are being inspected, but not having much done to them, where as others a lot more.

That inspection involves nearly complete disassembly. Which documentary did you watch? Aerodynamic surfaces and structural elements get replaced from fatigue and wear. Parts get removed because of corrosion (airplanes get exposed to a lot of moisture). New components get introduced because systems get upgraded. This is normal and expected, but it's all maintenance. From an article[0]:

---

> Foreman said it typically takes a B-52 between 220 and 260 days to go through depot maintenance, depending on parts availability and whether a bomber has any age-related stress fractures or corrosion that need to be repaired. The Air Force is still trying to figure out how much more time the upgrades might add to that schedule, he said.

> Cracking and other structural issues are common on the six-decade-old B-52, Foreman said, and sometimes require components to be replaced. But the Air Force is used to catching and fixing those problems, he said, and the aircraft should be able last well into the 2050s — perhaps to 2060 — without more in-depth structural upgrades.

---

> So you can see why I was specifying how this can be mitigated.

No, I can't, not to the degree you claimed. I don't think your argument about being able to make a maintenance-free-for-decades fridge--pardon the pun--holds water[1], much less without significantly increasing the cost. If you're expecting them to do all the work of discovering issues upfront, why would the cost of doing that not inflate the price significantly? Also, ongoing inspections of the fridge would cost money.

If your argument is that eventually we would be able to make humanoid robots where the maintenance costs are negligible, then sure, maybe. That seems within the realm of possibility.

But again, you said we could make fridges which didn't need maintenance for several decades. I don't think that's within the realm of possibility. You used something (the B-52) which needs regular maintenance as an example of why we should be able to make something else (a fridge) which needs almost none. This is what I was responding to.

[0] https://defensenews.com/air/2024/02/20/tinker-air-force-base...

bbarnett 11 hours ago | parent [-]

Well, I still feel the age of the platform, its use case, means that the B-52 is a good thing to point to. But, OK, fair enough... the rigors of military maintenance are indeed there, and of use. And as you've demonstrated, more rigorous than I thought.

I'll take another tact here.

https://forcedistancetimes.com/asymmetric-why-china-still-ca...

You may be aware, but it wasn't until 2017? that China could make a ball point pen from domestic parts. The tolerances of the ball, the machinery to make it, the quality of steel... just wasn't there.

Yet we've had that in the West for most of a century? Regardless, it's not about specific timelines, but instead about the fact that most of our modern world could not be crafted by hand. Modern molding, quality of steel and components, machine tolerances, all of this means we're able to make products to tolerances and specifications that were simply impossible before.

The reason I stressed the Volvo, the B-52, is because they are platforms known for fixing stress points, early wear, resolving manufacturing defects, and becoming far more reliable as a result. Unlike modern manufacturers, who mostly derive profit from full replacement of dead products, the profit motive for the Volvo and the B-52 aren't there in the same capacity.

I have nothing against profit, but in this case I do believe it has resulted in far less research into reliability. And the constant churn for 'new new', results in change with little benefit but a constant change to manufacturing methods, and so on.

It's so demonstrably bad, that Quebec (where I live) has passed laws stating that appliances such as fridges, are good not for a year, or two years, but instead a 'reasonable timeframe'. It's purposely vague, for example a single person buying a $1500 fridge, vs someone buying a $800 with a family of 4 might expect more wear and tear, and the warranty not quite as long.

But the reason behind this law, is fridges used to last a least a decade, meanwhile LG fridges often fail in just a few years these days.

Back to the main point, we have the ability to product parts with incredible consistence and to very finite tolerances. Modern failures are no more down to type of material (you cite rubber, and yes that can be an issue), or to unexpected wear.

What I've been trying to convey is, we have part manufacturing down quite well now. Longer term platforms without change, and constant improvement on unexpected wear points, would result in far better outcomes on that front.

This is a far cry from the original post I replied to, which felt robots would have little use from the overwhelming maintenance costs.

rimunroe 3 hours ago | parent [-]

I was never arguing that we couldn’t make things more reliable, just that a fridge which lasts 40 years without any failures isn’t realistic

dragonwriter 16 hours ago | parent | prev [-]

> And how much maintenance?

From what I can find, 50+ man-hours per flight hour, which is pretty far from maintenance free.

dragonwriter 16 hours ago | parent | prev [-]

> A generic example, fridges could easily last 40 to 50 years without maintenance. They wouldn't be all that more expensive either. Volvo, and the B-52 bomber program showed this, with Volvo having some models unchanged for 20 years. The B-52 has been in service longer than most people have been alive.

Neither Volvos (presumably, a reference to their cars) nor B-52s are maintenance-free, even if they have long service lives with proper maintenance, so I don’t see how either supports your argument that fridges could easily be made to last decades without maintenance.

the_sleaze_ 2 days ago | parent | prev | next [-]

Cost of human is much higher. Taxes, healthcare, breaks, brain-damage related to their emotional maintenance, safety requirements, etc.

No reason a robot can't work in a dark cave flooded with radon, and that is going to be cheap real estate.

zdragnar 2 days ago | parent | prev | next [-]

Human flesh and blood is pretty bad at upgrading itself, too. A sapient robot, or one with specific programming, might adapt itself as parts wear out when individual components, limbs, and other odds and ends are separately serviceable.

qgin 2 days ago | parent | prev | next [-]

I think it depends on the application. Employees tend to not be free in the 21st century at least.

ralusek 2 days ago | parent | prev | next [-]

If a robot costs $50k, lasts 5 years, and does the dishes and laundry every day, I'd consider it.

mdavid626 a day ago | parent [-]

Isn’t that the dishwasher and the washing machine?

BryanLegend 2 days ago | parent | prev | next [-]

Forklifts do pretty great

Teever 2 days ago | parent | prev [-]

Have you factored in the ability for humanoid robots to be able to do preventative maintenance and repairs on each other?

In many instances with repairing electronics and home appliances labour is the greatest cost, not the material. Sometimes it's as simple as replacing a 50 cent washer to repair something, or perhaps squirt some lube here or there regularly to prevent something from breaking down.

If it's the same for robot maintenance then robots being able to fix themselves and each other will change the equation on ownership tremendously.

Imagine if everyone had a domestic robot and if it broke down their neighbour's robot could repair it. That would be an extremely user friendly and cheap way to deal with the problem.

sjsdaiuasgdia 2 days ago | parent [-]

This makes a lot of assumptions about the field service potential of humanoid robots. A humanoid robot is so much more complex than something like a washing machine. There are far more things to break. Assuming humanoid robot maintenance will look like general appliance maintenance may not be a robust assumption.

"Replace tiny parts" option - Which parts is the manufacturer making available for purchase and what does the supply chain look like for that? What tools are needed to do the disassembly, part installation, and re-assembly? Can a humanoid robot out in the real world replicate the clean room conditions in which delicate components were assembled then sealed inside some compartment so dust can never get to them? Are we going to put heat guns and soldering irons in the fingertips of every humanoid robot to support self repair? There's going to be problems that can't be resolved with the kinds of tools available in the average household.

"Replace modules / components" option - Having to buy a whole new hand when you really wanted to replace a single finger joint impacts the value proposition of self repair, it's not a 50 cent washer it's a $1000 pre-assembled component. The repair is now definitely doable in the field, at least.

You might also be assuming humanoid robot manufacturers would not work specifically against self-repair. They make more money if you buy a new robot, or you pay them to fix your broken robot. Maybe "fix this other robot" ends up on a list of forbidden tasks the robot will always refuse to do...

Teever a day ago | parent [-]

That's the right way of thinking about it.

I think that you'd design it to use human tools as a bare minimum, so a soldering station, and a 3d printer, or even milling machines and lathes if needed.

But you're right, it'll be restricted from doing that. So the idea is you buy one, jailbreak it, and then get it to build a copy of itself.

It's like asking a genie for more wishes.

sjsdaiuasgdia 19 hours ago | parent [-]

> get it to build a copy of itself.

Where does it get the billion dollar semiconductor fab to make the chips for the copy?

Teever 14 hours ago | parent [-]

> get it to build a copy of itself.

Get it to assemble a copy of itself from a combination of available parts and anything else that it needs to manufacture from scratch.

sjsdaiuasgdia 14 hours ago | parent [-]

What motivation does the manufacturer have to make those parts available to you?

Teever 9 hours ago | parent [-]

It'll probably come down to a combination of regulation from an entity like the EU and the economic reality that it's easier to make a robot that has OEM parts that a consumer can access out of your supply chain.

sjsdaiuasgdia 6 hours ago | parent [-]

You've retreated from "buy one humanoid robot and it can make as many more as you want from scratch, it's like asking a genie for free wishes" to "there could be a regulatory framework that would require the manufacturers of humanoid robots to make a retail parts pipeline that allows your humanoid robot to build another humanoid robot at a cost lower than what the manufacturer would sell one for."

You didn't talk about the cost, but what's the point of having your robot assemble parts that the manufacturer will sell to you, assembled, at a lower price? It only makes sense if it's cheaper.