The obstacles to scaling up humanoids

I've owned several Roomba type robots, both actual Roombas and competing brands. None of them have really saved any time or labor. They always get stuck under furniture or tangled on charger cables. They don't work on stairs. And clearing dog hair from the roller is a huge hassle. I fully expect to still be paying a human cleaning service decades from now.

I have vacuum robots. I'm considering a lawn robot. My suburban city has two large ones mowing the parks, had never heard of them previously. Mostly worried about pets and critters.

> But neither of these are exactly flying off the shelves.

Roombas and lawn robots are all extremely popular.

Of course there always be "There is no reason anyone would want a computer in their home" people, that's life.

Roomba is pretty mediocre at a single job it's kind of able to do.

Humanoid robots _potentially_, _hypothetically_ can do anything that human can do because they are designed for environment, tools and equipment that we designed for our bodies.

> Humanoid robots are at least as hard as useful self-driving cars

A self-driving car only has to do one thing - drive. It's also got a stable wheeled base and only a couple of degrees of freedom - got to steer and regulate it's speed.

Even if the only thing you wanted the humanoid robot to do is drive your car, it'd be massively harder for it since it's got all those degrees of freedom, will be bouncing around in the drivers seat, and presumably doesn't even know how to drive.

If the humanoid is more than a gimick - meant to be general purpose, then it needs an AGI brain and ability to learn for itself. It's not going to be learning in a simulator like your FSD car - it's be learning on the road like your teenage kid.

I think some of the car role will port to robot role.

Vision for cars already includes object detection, and the better that is, the better robot object detection gets. The same for "human ran out on road" would work for "walking in house, small human is now in front of me, stop!".

I wonder how much of the one will port to the other. A house has paths aka "roads", inside and out. Places the robot may walk, and not. So path navigation is a thing too. Maybe 'getting around' is mostly solved, while of course other challenges are still there.

Sort of replying to others in this part, the reason people are all hung up on humanform, is that our entire world is made for humans. Whether stairs, doors, sidewalks, doorknobs, cupboards, or even space to walk in a small kitchen... it's all made to work with human shape and size.

(Yes, while there is wheelchair access mandated, that doesn't extend to the inside of every home, and all the spaces in homes, and even then everything we have is designed to be operated by fingers/arms/hands.)

So if you solve humanform, the robot can go anywhere and manipulate/do anything a human can. That means no change to the environment when you get one. Right or wrong, that's why everyone is after humanform.

ASIMO is 25 years old. PETMAN is 16 years old and Atlas is 12.

It’s actually going faster because it does not require public approval the same way that driving cars do because you’re in public space.

We’re seeing a lot of robotic trials happening in private warehouses and on private test ranges at pretty rapid scale

Beyond that the methods for transfer learning behavior cloning behavior authoring are very robust so that I can get joint angles directly from a human via instrumentation through vision or even commodity sensors which captured trajectories that can be immediately applied to robotic joint positions.

The real challenge is actually capturing demonstration recordings from humans because it’s the hardest thing to instrument. The core task is instrumenting data capture of existing human tasks that are not done through machines, such that they can transfer to machines.

This is easiest done with existing human operated robots because the instrumentation is free, so data can go directly into real2sim2real pipelines.

There might seem counterintuitive but most of the actual technical bits and bites are already there it’s re-orienting the economic and logistical process of labor execution that is the major challenge.

I will say though, I’m seeing less and less barriers there as time goes on. Employers really want to not have to hand human employees

(a few specific towns, which have done various things to smooth operating conditions for Waymo.)

It's useful, don't get me wrong, but when Waymo can handle Cairo and Rome, I'll consider it a solved problem.

I own at least three! Dishwasher, laundy washer, laundry dryer...

I would get a roomba but it can't do enough fine detail to be worth it.

and so is the safety margin for a humanoid. The consumer market is huge only if the robots are highly reliable and work very well both of which are not true at the moment. Things will change but it will take quite a bit of time and much more research.

They're all dog bots adapted to stand on back legs. None are true humanoids in the first place, none of them even have articulating pelvis.

"The vision" is that instead of building 9999 specialist robots for 9999 different tasks, you mass produce one robot model that can do all of them.

Less efficiently, sure, but for the manufacturing, logistics, maintenance? The economies of scale are immense.

The reason why we weren't doing exactly that back in the 80s isn't that universal humanoid robots somehow weren't desirable. It's that for a universal humanoid hardware to be useful, you need a fairly universal AI to back it.

That "universal AI" was nowhere to be seen back in the 80s, or the 90s, or the 00s. Now, we finally have a good idea of how to build the kind of AI required for it.

I think the reasoning is that the world is built for humans, so if you need a robot to do arbitrary human tasks in an unmodified environment then perhaps that gives an advantage.

The premise itself seems bogus though - there's plenty of tasks such as traditional assembly line and conveyor belt automation where a stable robot bolted to the floor, with a wired power source and custom manipulators is going to be a much better option.

For a mobile robot stability and reliability are key, and it's hard to see how a humanoid robot would be anything other than a massive downgrade for applications like Amazon's warehouse robots, hospital drug delivery robots, mall security robots, robot vacuum cleaners, etc. Wheels for the win.

OTOH there's the dream/hype of a domestic robot doing all your household chores, where humanoid form might actually be a plus, but at this point that's a pipe dream, and I seriously doubt many people really want C-3PO in the kitchen washing the dishes even if he is managing to do it without breaking anything or short-circuiting himself. It's like a 60's vision of the future, with people in flying cars or living on mars. No product-market fit.

If you can have one robot do multiple tasks at 80% of the capability of a special-purpose robot, then that’s a win. Buy a fleet of them and just repurpose them as necessary.

If they’re humanoid then they can already use tools, equipment, and access methods we already use for ourselves.

What part of the vision doesn’t make sense?

They're struggling to find buyers because they can't do anything useful _today_. For the price of a car it's obviously hard to find early adopters.

What's the worst that could happen? https://www.rottentomatoes.com/tv/humans

Humanoid robots are notable worse than humans in many aspects that impact productivity. If a humanoid robot is ultimately 33% as productive as a worker in a developing country who gets a wage of $10k USD annually and works 8 hours per day every day, then then robot has to cost less than $10k annually all in to be a good replacement. Assuming a 5 year useful lifespan and $2k in maintenance per year, results in the robot needing to cost ~$40k before it can replace a human's productivity. And that is inclusive of training and setup, and I doubt we'll have robots that are capable of learning as quickly as average humans without dedicated specialists training them... which raises the cost.

In general, you can get a dedicated machine for most human tasks that is easily 10-1000x productivity if you have a few million in capital. There are tasks on the margin where human flexibility and dexterity that having a human operate a $10k sewing machine is going to be very very hard to replace.

Can't machines work a 7-day, 24-hour schedule? That said, humanoid robots strike me as a jack-of-all-trades tool, our environment is full of things that are optimized for human-sized and -shaped users, but if you can purpose-build your robot for a factory, it's going to be more efficient at a narrow set of tasks there.

The biggest bottlenecks are hardware design and software design. Materials science to an extent, particularly battery materials, but we could build robots with currently-available materials and power density if only we knew how to make them work usefully enough.

I'm not against the concept and I agree the manufacturing can be scaled. There just isn't a product yet.

Isn't the biggest bottleneck just that they need to adequately and reliably be able to do useful work at a better price-performance ratio than a human ?

And that's just not the case yet?

Article calls out demand — so far there has been no demand for large numbers of humanoid robots.

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.

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.

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.

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.

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

Forklifts do pretty great

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

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.

Moore's Law was primarily a tool for arguing with finance guys to let you borrow the enormous amount of money to build the new fab- see, this law says that if you don't let us raise the money, then the other company will- see this straight line? It's inevitable, so give us the money.

We could have chosen that scale of investment in battery technology, certainly, but the finance guys said that it wasn't profitable to invest that much in it. China was willing to invest into this field for national security reasons (the threat was of the USN cutting off oil traffic in the Indian Ocean, well beyond the range of the Chinese military). Semiconductors themselves were the result of (US) national security investment- the Minuteman ICBM guidance system was the first large scale use of IC semiconductors.

Given how quickly battery technology has advanced- and how profitable it has been to build these factories- I think the evidence is that the finance guys were wrong about the profitability and importance of battery research/construction, and the national security investment jump-started the whole thing, just like with semiconductors.