Proof of Concept to Test Humanoid Robots | Hacker News Reader

Proof of Concept to Test Humanoid Robots

thehumanoid.ai

18 points0xedb25 comments

The human body is sub-optimally designed for most hard work humans do (which is why that work is "hard"). I laugh every time I see AI videos of a human-shaped robot harvesting crops: we have very, very effective crop-harvesting robots right now, and they are shaped like big boxes on wheels because that's a much better shape for doing that.

Universality matters though. It's less interesting that a hyper specific machine exists for a task than that the same machine might be able to do a wide range of tasks, provided the price point is right.

“Less interesting” is an interesting value to compare things that are typically measured by utility. Human form factor robots are definitely more interesting to us as humans, but really only economically viable for high mix low volume tasks (of which there are many).

But past a certain scale special purpose machines will always be more cost effective.

And more annoyingly they will no doubt be given modular behavioral capabilities that require separate subscriptions to use (even the big cube-shaped farming robots do this)

The human form is terrible for most productive things. We are slow, weak, short, and inaccurate. Robotic arms are the true multitalents of manipulating the physical world.

That doesn't say anything about the human form: it says something about the human body.

But it's also not very accurate on that count: we are actually very strong compared to mechanical systems of a similar size, weight and energy structure.

You're mistaken. It says something about the shape of humans. Bicycles, which are powered by the human body, show the inferiority of bipedalism to wheels, when it comes to fast and efficient locomotion.

Those crop harvesting robots can’t do anything else though. They’re also not very good at weeding, or picking berries or tea. Things that require finesse. Also imagine not having to use the god awful amounts of pesticides we currently use. You’ve got to think of these humanoids as universal. You should be able to tell the robot picking weeds to stop and go do the grocery shopping ideally.

Why? One tool for one job. I let my gardener robot keep gardening while my grocery shopping robot goes to the store.

A gardener robot should be able to:

- plant new plants (hold plant in pot, remove plant from pot, shake excess soil, dig hole using trowel, place plant in hole, pat down earth, water plant using watering can)

- dig up weeds (using e.g. a hoe, fingers)

- set up a trellis (attach trellis to wall using drilled-in screws; wrap vines around trellis)

- water plants with hose (unwrap hose, turn on tap, spray plants)

etc.

What form factor will beat humans at all these tasks?

Four different drones, or one drone with four tool attachments. The robot shouldn't need to "use a trowel"; it should be a trowel.

Because these things are going to be priced like cars. Most households can only afford one or two.

The use case claimed here is (a) they can move around (b) they are "universal".

But

(a) Those things look like they need a wide berth to move around and flat terrain

(b) Those end effectors are far from universal. The payload weight seems so low that it even dropped an empty box at https://www.youtube.com/watch?v=3FIXjy2GWTg&t=150s

The weirdest part about the box carrying humanoids is that this problem has already been solved by fork lift style robots [0] which are being sold by many companies. When people talk about universality, having a central warehouse with movable pickup & place locations is all you need. The only thing that would be interesting is to build a universal loader/unloader that can take parts out of the boxes and insert them into machines factorio style, but that loader doesn't need to be humanoid and could instead just be put on a movable cart.

[0] https://www.hairobotics.com/products/haipick-a3

The bot may be notionally "universal" but will only operate on the DLC you buy from the robot rental company. Want it to wash dishes? That's the $20/mo dishwashing pack, or for one low price you can get the entire housework pack for only $80/mo.

It’s got to get good enough for open source versions at some point though.

Eventually, but consider the power envelope for compute in an android is 10-100x lower than for a car*, while at the same time the range of tasks a general purpose bot has to be good at is a strict superset of driving a car, and that open source attempts at self driving cars are far behind the proprietary attempts, and that compute efficiency is only doubling every 2.5-3 years (Koomey’s Law), so expect something like 20 years before these are properly general purpose and also not remote controlled.

Still plenty of value from special purpose and from remote control, but that's the timeline for solving both at the same time.

* even with the compute being external, the global electricity supply is presently a few hundred watts/capita and already being used for all the other things we want electricity for, hence all current anger about data centers making electricity too expensive, but renewables could tilt this to 8 billion robots with 1kW compute each in as little as 10 years if we brush aside all the decarbonisation efforts and keep burning petrol in cars and gas in stoves etc., otherwise more like 15 years for that.

That would shock me. These things are going to have a TPM

What’s interesting here isn’t the humanoid form factor, it’s the systems integration. Plugging robots into Siemens’ industrial stack means they’re being treated like first-class nodes in existing logistics workflows, not special demos. If humanoids can reuse current automation software, safety models, and ops tooling, that lowers adoption friction a lot. The real question is whether reliability and MTBF get good enough to compete with simpler, non-humanoid automation at scale.

In this use case, the robot autonomously picked totes from a storage stack, transported them to a conveyor, and placed them at the designated pickup point for human operators.

Well, yes, you can use a humanoid robot for that, but there are far simpler robotic solutions. There are lots of systems for handling standardized totes.

Clearly that’s not the point. The end goal is essentially to build a robot that can function as a human slave would in the past.

I noticed they used the wheeled version for the test, so calling it a humanoid feels like a bit of a reach to me. The speed of sixty boxes an hour seems pretty slow if they want to replace actual people on the line.

What constitutes a "humanoid" robot is a matter of some debate: https://james.darpinian.com/blog/the-humanoid-alignment-char...

They look so hilariously slow and bad in the video, and it’s a really simple back and forth task, with empty crates.

humanoid robots require an order of magnitude better battery technology that does not exist yet.That technology will change a lot more than just having viable robots

Reply on news.ycombinator.com