Hacker News Reader

I don't dismiss Tesla's successes at all, but they are primarily repackaging and improving technologies that already exist. They didn't need to invent the lithium-ion battery, or the electric motor, or the automobile; all of those things already existed and were available to be adapted to Tesla's desired use.

Humanoid robots, however, are a much harder problem than building a mass-market-ready electric car. In order for someone to bring to market a robot that can perform general household tasks, for instance picking up dirty laundry around the house, putting it the washing machine and dryer, then folding or hanging it up, then they first need to make multiple fundamental several-order-of-magnitude breakthroughs in the fields of machine vision, robotic locomotion, energy storage, artificial general intelligence, etc.

I'm not dismissive because I don't think Tesla is full of brilliant people who can do amazing things. I'm dismissive because building a useful, mass-market humanoid robot would be an achievement of incredible magnitude that's as far outside of the current envelope of technological achievement as the moon landing would have been to the Wright brothers.

You've left out manufacturing. Building a car factory + supply chain is quite the challenge.

Tesla had the best and the brightest - I've worked with them before they were acquired by Tesla on a daily basis: https://teslagrohmannautomation.de/en/

and still had trouble scaling. Manufacturing is considered to be one of the hardest things about building cars - electric or not.

Edit: What's with the downvotes? Disagree? Comment below then.

According to musk manufacturing is the hardest part of rocketry as well.

that makes sense. rockets are big dumb objects with incredibly tight engineering tolerances. humanoid robots are anything but big and dumb, and there's no incremental path to success between a bunch of servos wired to a mannequin and having it do the dishes.

>Building a car factory + supply chain is quite the challenge

And yet Elon announced to the world he was going to have the most automated factory in the world, the machine that builds machines, and that the factory was the actual product. People lapped this up. 2 years later they were hand-assembling cars in a tent.

Now do you see the problem?

>Manufacturing is considered to be one of the hardest things about building cars

Again, you say this as if it's some kind of revelation. It's obvious. The only shock is that Musk and the people he was surrounded with didn't realize it.

They tried to do something quite crazy with manufacturing and it didn't work. Musk has often ignored industry standard and often it works. In this case it didn't and people will now use this forever to show how stupid he is.

However what happened then is that they effectively corrected their mistake, fixed their production, reached their production targets with excellent unit margins.

They learned from the episode and realized they need to reduce part count and made a whole series of innovation and if you look at how their new Model Y are going to be developed in the new factories you see a huge improvements. They developed a new alloy for casting, worked to put largest casting process into production, totally redesigned the battery for manufacturing and massively simplified the battery pack.

To the point where Tesla are quite competitive in terms of weight and ease of manufacturing. They already have industry leading unit margins and some of these innovations will only hit next year.

That part of the story somehow is never mentioned when people say 'hahahahha build cars in a tent'.

They are not doing so bad it seems:

https://pbs.twimg.com/media/E9FARebWUAYd5NI?format=jpg&name=...

>They tried to do something quite crazy with manufacturing and it didn't work.

What did they try to do that's "crazy with manufacturing" that hasn't been done by every other car company on the planet?

>fixed their production, reached their production targets

Have their plants ever reached production capacity?

>To the point where Tesla are quite competitive in terms of weight and ease of manufacturing.

And here we have the point. Tesla are competitive, fantastic. That's a long way from all of the early claims about upending the entire manufacturing industry through factory automation the likes of which the world has never seen. Funny how that part of the narrative disappeared.

>They are not doing so bad it seems

Looks to me like legacy car makers are catching up, no?

I was responding to the comment above mine that didn’t mention manufacturing but pretty much everything else. It might be obvious to you but general populace have no idea about difficulty of manufacturing. Some engineers haven’t got a clue on what goes into building things by millions.

Having Rosie the robot maid from the Jetsons at home is the vision of the future being sold, but as you point out, that level of technology is very much still the realm of science fiction. If you're watching the other parts of the AI conference this was presented at, self-driving cars are practically here. The reality, of course, is that that actually aren't yet. But you can buy a Tesla today to get an electric car taste of that future, even though it's not actually quite here.

A bipedal furniture dolly, however, is as unsexy as it is possible, using today's technology. It requires none of the impossible breakthroughs, though energy storage seems like it would be a problem. (Quad-copter drones seem to have follow as a control scheme figured out already.)

Doing this under Tesla means the earliest versions don't need to sell at all, they just need to be useful internally. Which I'm sure is why this is coming up in the first place. Think of all the places a forklift can go in a factory vs a human. The forklift is extremely useful, but still limited to flat, paved terrain. A bipedal robot base solves that, enabling Tesla to make more use of available space.

If this bot follows Tesla's strategy for self-driving (which still isn't here), it'll ship with "all the necessary hardware for Rosie the robot maid-mode", and then consumers will have to wait another couple decades for the software to actually catch up.

> They didn't need to invent the lithium-ion battery, or the electric motor, or the automobile

They do have their own batteries, their own motors, and a pretty unique automobile. I'm not so sure the point stands. If the "they are only using existing technology" line was true, you'd have seen a ton of companies from China, or the traditional automakers, take the market before Tesla had any chance to compete. Somehow Tesla had time to build their own factories and parts from scratch and still beat them...

> "but they are primarily repackaging and improving technologies that already exist."

Did you mean: they are innovating? IMO there's not enough appreciation for getting to commodification and scaling the existing tech. It's a very difficult problem, from what I understand.

> I'm dismissive because building a useful, mass-market humanoid robot would be an achievement of incredible magnitude

Agreed. But they don't have to start with "mass-market humanoid robot", there are 1000s of niches where then can execute "land and expand" strategy. Same playbook as with Tesla or SpaceX: start with something niche (sport car for rich people/gov contracts), iterate to expand ($30k EV/satellite internet everywhere).

This isn’t “no wireless, less space than a Nomad,” though. This is like RCA announcing an iPod in 1961, reasoning that because they already make magnetic tape and portable electronics, it’s only a matter of time before they’ll be able to fit 1000 songs in your pocket.

Fair point, however, I look forward to coming back to this comment in 5 years (1966).

Maybe I’ll be eating humble pie, however, my guess is that by then it will be out of prototype phase and in use.

In 40 years, 2001

This is a good point that most of us have difficulty to grasp. The innovation speed is actually increasing.

While Tesla is spending a lot of time and effort on their AI (from hardware to software), they might be able to reuse it later for other things, automating more tasks, and at scale. There is a compounding effect.

I don't think Tesla is kidding itself with Optimus, the v1 would be close to useless, and they know it. But it would allow them to learn, build a team of experts and get to the state of the art, to eventually improve it and go beyond. Maybe the main added value at the time would on the software only i.e. mechanically the robot would not be much better than the Boston Robotics of the day. And it could sufficient to actually make a big difference between a Boston Robotics and Tesla Bot ?

I am curious to see what they will end up with. I am not worry about the timeline. I was not expecting a humanoid robot in the next 10years. Tonight announcement is not really changing it, except that now, Musk put a stick on the ground, and we might witness some actual progress.

I am blown away by how many people repeatedly under-estimate Musk's companies despite him eventually proving them wrong basically every time.

True. That's why the cross-country autonomous drive that didn't happen in 2017 is imminent.

Exactly. People were saying it was impossible. He announced it to happen in 2018. It will happen 3-4 years later than planned. And people use that to bash him, forgetting the elephant in the room which is that he made the supposedly impossible possible.

OP was sarcastic.

I know. I wasn’t clear enough I guess. I should have said « You said exactly the kind of ridiculous argument that haters repeat ». My answer was also sarcastic.

Can you describe the manual labor field where 1% failure rate is acceptable? I am struggling to think of one.

I suppose it depends on the magnitude of the failure.

I worked in a factory that was the distribution centre for all the cold and frozen food in every Safeway in the sate. There were a few hundred of us who had to pickup a heavy box from a conveyor belt, turn around, take a step or two and put the heavy box down on a pallet. When the pallet was full someone would come and take it away. Repeat. For 8-10 hours a day, 6 days a week.

Upstairs people were manually putting those boxes on the belts. It destroyed people's bodies, I'm happy I only worked there for a few months each year to put myself through University.

The factory was very high tech and cutting edge, though the humans at each end of the conveyor belts were unavoidable.

We messed up all the time - dropped boxes, did bad stacking and made pallets fall over, crashed pallet jacks and knocked over pallets etc. The failure rate was probably around 5%, I'd guess.

Starting pay was around $30/hr, double that on Sundays, and 2.5 on Public Holidays

A humanoid robot that actually works most of the time will change the world.

This doesn't seem like the kind of problem that would require a humanoid robot to solve though? It's just that even with current state of the art robotics, things like grabbing cardboard boxes (easily crumpled or dropped, mass distribution within may be uneven or unstable, may already be broken or leaking etc) and stacking them into a stable 3D pallet is actually quite hard, so much so that Boston Dynamics managing to do some parts of this in optimal conditions with uniform, lightweight boxes was hailed as a breakthrough:

https://youtu.be/uuO6oeO0-ts

Every manual job I’ve ever done I’ve probably failed more than 1% of every movement. But then I correct. In a car moving at high speeds (focus matters more), I do this less often, but I have been in crashes.

Well, anything that if you do incorrectly someone (or some robot) will likely spot it before the failure gets to a customer. For instance, if you have a robot that picks fruit but 1% of the time the fruit is not ripe or is worm-eaten or something, then you can have a process where the picked fruit are inspected a second time and the bad ones thrown out.

Another case where 1% failure would be acceptable is weeding. Suppose you have a bunch of robots removing invasive English ivy from a forest. Maybe once in a while they'll pull up salal or a fern or something by mistake, but that's fine because those aren't endangered species and in the absence of ivy the ecosystem will quickly recover.

Picking oranges off a tree.

Cleaning up trash in a parkland.

Folding clothes.

Searching a forest for survivors.

Unclogging a sewerage pipe.

For many menial tasks, a 1% failure rate (or indeed any failure rate higher than a human) is generally not a problem so long as the economics stack up and safety isn't a concern.

Im 100% sure Tesla can mass-produce the hardware for such robots. But the biggest challenge is in software, And then the FSD debacle comes to mind.

There's major advances in AI needed to get to full self driving and even more advances for a useful autonomous humanoid robot.

But then again, those will not come to fruition just by waiting, someone has to work on this and if Tesla advances the state of the art, I can only welcome them. But I don't expect a Tesla bot in the way Elon poses here in the foreseeable future.

The hardware for such things doesn't exist yet, we don't know how to make it.

Have you seen Boston Dynmamics robots? And their clunky size and shape? And compared those to the life-like body of the Tesla demo?

We don't know how to make such robots yet.

Oh, I'm 100% sure the robot hardware Tesla will make will not look like what is shown in the presentation, because that design makes no sense. There's no room for joints to move etc. The BD hardware looks cluncky, but it is robust.

> The flaw with full self driving right now is not that it doesn’t work sometimes.

.. doesn't work many times.

If my robot folds my laundry correctly for 99 days and on the 100th day shreds my wardrobe, I don't think that will print money (at least not for very long).

From what I’ve seen FSD makes minor mistakes that at high speeds lead to danger or death. I can’t see a minor mistake in folding laundry leading to the laundry being shredded. Musk said the robots would not be strong (only can lift 5lbs). So unless the robot picked up scissors it couldn’t shred the laundry.

Unless I’m missing something?

FSD has not been rolled out yet, so you’re attributing to FSD some errors that are actually made by humans. Yeah maybe these humans were confused by marketing to think it was FSD, just as you think it was FSD, but that doesn’t make it FSD.

The downvotes are fascinating. It’s like doubling down on misinformed thinking but… to what end?

Tesla is quite literally selling a package called "Full Self-Driving". I don't see how it has not been "rolled out yet" unless you're confusing Tesla's bullshit marketing lingo for level 5 autonomous driving.

It’s described as a future feature that you can prepay for though.

What, in the fine print after the giant "FULL-SELF DRIVING CAPABILITY" banners?

Sure, that would be bad. So it only needs to be good enough to fold correctly for 99 days, and then on the 100th day just do nothing (i.e., detect a problem and hard STOP).

Then a hard reboot will kick it back into line and we're good to go again.

Obviously that's not good enough for driving a car, but it certainly is in all the manual labor jobs I've ever had.

A robot that folds laundry would obviously have torque limits while handling clothing, so the prospect of destruction would be minimal.

The error domain of a laundry robot would be misidentifying an unusual item and incorrectly folding it, or incorrectly sorting it. If someone sold a humanoid laundry robot which folded and sorted my clothes correctly 99% of the time, I'd consider that miraculous.

Tesla is not without problems. Poor drive quality. After spending 40K+ for the car, I was cited for not having front number plate. I had to figure out how to tape the frame, spend more time appearing in court. I prefer not doing such things than having automatic s/w update for fine tuning fart noise.

You're blaming Tesla because you didn't use the front license plate holder that they give you, ready to go?

why did they put it in the trunk when all other car dealers fix it and give ? And why do you see 80% of Tesla'a on the road without the front number plates ? Also if you put the number plate it may not reach their advertised 0-100 mile in 3.x seconds ?

Because a lot of people like to leave it off, especially on expensive cars, not exclusive to Tesla. It looks better and the chances of getting a ticket are actually pretty low. There are plenty of people who've driven around without a front plate for many years without getting cited, and it usually just a fix-it ticket without a big penalty. It's not generally because people are ignorant of the law. And it's not required everywhere either.

I’m happy they left it up to the customer to put the front plate on. It makes the car look worse and I’ll accept the risk of a citation.