> A Weibo post expands on this, one section translating roughly to the following: "the 100% self-developed 'Xiaomi Pengpai Intelligent Manufacturing Platform' is the brain of the factory, injecting soul into the factory, allowing the entire factory to have self-perception, self-decision-making, and self-execution capabilities, and can independently diagnose equipment problems, improve process flows, and realize full-scenario digital management from raw material procurement to delivery, becoming a true smart factory that can evolve by itself."
I'm going to need a lot more detail before I can believe this. It's too much fluff to be taken at face value.
It looks to me like China will move from using cheap labour to cheap electricity from renewables for its manufacturing a lot faster than the rest of the world can keep up.
That clean energy is even cheaper than their already cheap coal powered power stations - they are going to go even further eating our manufacturing breakfast.
Whatever you say about value of software. You still need to build machines and all the parts it runs on. Not to forget all parts of the building it runs on and all parts of the networks it connects to and everything to power it...
Capability to manufacture full hardware stack is critical. China might not be top at the silicon chips, but they have extreme competence for below that.
This. I wonder why people say India will be the next China, as if they are still living in the past. The future is about cheap energy powering robots, not cheap labor. The era of growth based solely on cheap labor is slowly ending with robotization.
China's installed solar power capacity rises 55.2% in 2023.[1]
It added 216 GW of solar PV capacity alone in 2023 that was equal to 14% of the world's total installed solar PV capacity, more than what many countries have ever installed in total.
India added 7.5 GW of solar in CY 2023, a 44.1% drop from 13.4 GW installed in CY 2022.
So India 7.5GW vs China 216 GW.
"China is already the largest industrial robot market in the world. It accounted for 52 percent of robot installations worldwide in 2022" [2][3]
I don't see how they can compete in renewables or robots. Catching up to China will be extremely difficult.
The difference is that China has the complete chain, from the raw materials, processing of said materials, manufacturing, integration, maintenance. They have the IP and know-how. India may have the land and the sun, but it lacks the hard parts.
>India can just buy the panels from China and be in nearly the same position.
Can it though? This will be impossible because of Indian domestic politics. You will also never get the same benefits by buying as in producing them yourself.
Everyone had always been moving towards such direction at least since days of Charlie Chaplin... It's just that marketing automatically produced goods had always been strangely hard.
Everytime I see something like this I am more convinced that deep learning/AI focus should go away from language into other domains where it can make a bigger impact on less compute.
It already has. Language is just the most accessible by most people. But deep learning is not new and is not newly applied to problems. But it’s unclear how much of this article is hype or new.
Gigafactory was also supposed to be 100% automated in a much more difficult assembly than putting a few boards into a metal case and gluing a screen on. They failed in practice and balanced into something more practical. (I’d not even in this article they eventually nod that it’s not fully automated just the “important parts”)
Playstation factory was also almost fully automated: "PlayStation 4 units come off the assembly line at a pace of about one every 30 seconds." https://news.ycombinator.com/item?id=23748664
Of course it will. AI is driving a flood of development in robotics. A home robot is definitely coming in the next 10 years and an assembly line worker is going to be an extinct job.
We already have home robots in the form of vacuum cleaners. If you're referring to humanoid robots for domestic chores then forget it, 10 years is way too optimistic. The safety problems alone of having a powerful heavy robot moving around with people and pets will take decades to solve. AI in the form of LLMs is effectively impossible to test for safety critical applications with so few system constraints. The risk of product liability lawsuits will prevent this technology from being sold anytime soon. Some limited industrial use may be feasible because there the robots can be carefully monitored and have limited interactions with humans.
Plus there are remaining hardware problems in terms of power and actuators that will take longer than 10 years to solve in a cost effective way.
Seems pessimistic. I already have an oven that can burn the house down, a dishwasher that might flood out my kitchen, a refrigerator that can fail and ruin a fortune in perishable foods. Hell, my lawn mower might remove a foot!
Why is a robot that might step on the cat, somehow inconceivably more hazardous?
Those examples are unlikely to kill you. A humanoid robot can. Let's say it quickly moves its arm on a collision course with your head. The arm can weigh around 10 kg. The impact can easily kill an adult.
I'm a bit less pessimistic than OP. It's possible to run safety checks (e.g. collision avoidance) on a lower, deterministic level, and use generative, black box, 'AI' for higher-level planning.
The definition of "quickly" for roboticists are different from how it's usually interpreted. I'm not a real one, but to me it seems this tree of comments contain couple more such definition disagreements.
Like animals the robot arm should also be compliment and not always be moving with full force because it is going to hit things even with active collision avoidance.
Static home appliances don't move around and the space of possible dangerous failure modes is limited. This stuff is easy to test for safety. And a refrigerator failure isn't a potential safety issue.
LLMs tend to fail in novel ways that are impossible to reason about. What is the expected failure rate for mistaking a human infant sleeping on the floor for a large doll that needs to be picked up and put away? You can't prove that the rate will be zero. We don't have good ways to model and estimate those failures rates in such complex systems.
My point is, robots are not essentially new in their home-environment risks. We accept some level of risk now. The pessimistic view that home robots are impossible to deploy until their risks are reduced to zero, is nonsense.
My theory: we'll have home robots as soon as they're affordable and have a market demographic. Like any other technology.
We have higher standards for designed devices than sentient ones, especially ones that have a long history like dogs.
I love my dog, but appliance with the QA tolerances of dogs would not sell well. Some dogs were bred for helping you work, others were bred as guard dogs. If I went around selling something like a 75% toaster, 25% gun device, and we don’t know which parts come from which, I think I would go to jail.
These claims are just on the edge of believable to me, but tantalizing -- they say pretty clearly you can dump a product spec (which I imagine is still a highly specific CAD file of some sort, not a "make me a cool phone" type instruction), and the software side of the factory can instrument raw material production into a final product.
I don't think there are any steps in this process that couldn't be automated with today's tech; however I bet there are a lot of steps in this process that have a human checking in and approving or clicking 'redo' right now.
Taking the step to add self perception, hooking up sensors to everything is smart and very interesting, adding logistics connectivity is very interesting, basically the whole thing is intriguing.
I have a lot of questions, though -- it seems like it's 10 years too early to have this working -- but if you're planning long range, the experience here will almost certainly be baseline work you need done for building a better more successful autonomous factory over the next few iterations.
It would be an incredible production advantage for the company that's first to really getting to 'automated automated' -- once you had that platform, you could point it a lot of places.
This is the kind of thing that's actually necessary for sending humans to Mars. But a couple of levels higher - we need an autonomous factory to build other autonomous factories, that can be packed into a launch vehicle (or assembled in space). Until we have auto-factories on Earth, Mars is but a pipe-dream.
> we need an autonomous factory to build other autonomous factories
Good plan but let's start with something more simple than a phone factory. How about paperclips? Once that problem has been solved [1] we can be confident we can create the optimal Phone Maximiser.
I’m sorry but I remember quite recently when 3d printing was promised to do this and it failed miserably. Mostly made a lot of figurines. And a few specific use cases where it shines. Just as one would expect from a Gartner hype cycle.
100% agree with you; don't care if it's about 3d printing, AI, or rockets. My point is that until we can build a fully autonomous factory on Earth, we're not going to Mars.
Well a lot of jobs should be automated and more so those jobs that the pay check is low, but also another categories of jobs will be created, it might generate some unemployment but also new opportunities, how many jobs the computer took away and how many it generated?
Well they can make a billion for all I care, I'm done with Xiaomi. After =~ 10 years of using their increasingly buggy, crappy and at the same time expensive phones, I'm out. I'll still stay on Android since that's what I know to use (and to some extent, develop for), but for my next phone I'll reach deeper into my pocket for a Samsung. I heard good things of Google Pixel as well but meh ... too technical. Samsung's got a bit of glamour as well where I live.
Here's a non-exhaustive list of the bugs on my Xiaomi phones:
- Battery draining from 100% to 10% overnight, while in sleep mode, energy usage showing "screen activity". Goes on and off, I suppose depends when the Chinese government decides to do a data examination.
- Video mode on the camera takes 10-30 seconds to start. If you plan to film Bigfoot, make sure you turn the camera on some minute before he makes his brief appearance. No update ever fixed this, if anything it broke other things as well and made it worse (sometimes it just hangs and never starts video mode).
- Speaking of updates, the rule with Xiaomi is that any update will break more than it will fix. With guaranteed some new annoying crap, usually worse than what you're trying to fix. After my last Xiaomi slowed to a crawl, whenever I charged it it basically bricked itself, nothing but a hardware restart turned it back on, I broke my rule and updated. Only to end up with a notification "Looks like you performed a factory reset (spoilers: I didn't!). Connect to the network and restore MIUI" with two buttons "Not Now" and "Restore". Pressing "Not Now" will keep popping the annoying notification and "Restore", after taking an inordinate amount of time, fails. Great job, Xiaomi developers!
- Taking pictures is fine but to examine them I need the patience of a saint. For 20-30 seconds or more it just hangs and shows me a white preview. I guess it needs to pass them through the Chinese cloud for approval first or something. One workaround I found is to take a picture, open gallery, press "share" on the white loading crap and send it to WhatsApp. I can then examine it in WhatsApp while Galery still hangs on Chinese cloud AI processing.
Idk man I've had a Xiaomi Redmi phone for over 4 years and it's honestly a rock solid stable device that hasn't given me any need to switch yet.
Camera is decent, screen is good, battery life is good, performance is good. Like honestly, I'd probably never buy a flagship from them, but for a beater phone I wouldn't get anything else.
Same here, you can do better with other brands on the flagships, but the Redmi devices are usually the best in their class (once you flash a custom ROM on it because all the shit about MIUI is true).
It sounds like all your problems are with Xiaomi's software. You could check https://wiki.lineageos.org/devices/#xiaomi whether your device is supported, and if so, keep your hardware while enjoying much less annoying software.
Edit: right now might not be a good time, because the most recent Quarterly Platform Release introduced breaking changes and many devices have not been tested and/or fixed yet, so they're marked as unsupported for now: https://fosstodon.org/@LineageOS/112670224607386323
Have they got their own version of Android? All those issues sound like software issues.
I prefer stock Android, I just don't trust any manufacturer to change it, and I don't see why on earth it's needed. Just gimme the most vanilla version of the OS.
Interesting; I had a Xiaomi Mi 9 SE for over 3 years, and it's one of the best phones I've ever owned. It was rock solid, and battery life was amazing, it would go 2 days of medium use before needing charged.
The article participated in none of these tropes, whatsoever. The posts here, to the extent they are negative, list specific Xiaomi product complaints. Chill.
I'm going to need a lot more detail before I can believe this. It's too much fluff to be taken at face value.