News on China's scientific and technological development.

mossen

Junior Member
Registered Member
So thoughts on Baidu's ERNIE?
China's AI effort is dictated by the same two forces that shape Western AI: chips and software.

China can longer use CUDA or other software stacks. Building its own is possible, but will take time. As for chips: it can't buy top-end GPUs anymore and while it can design its own, it cannot fab them because they are forbidden from using bleeding edge fabs from TSMC.

As a result, China's AI efforts will be also-rans until and unless it manages to fix these two vulnerabilities. Probably not what people here want to hear, but I also attacked Zero Covid when lots of folks here defended it long past its usefulness. Harsh truths need to be told.
 

BlackWindMnt

Captain
Registered Member
China's AI effort is dictated by the same two forces that shape Western AI: chips and software.

China can longer use CUDA or other software stacks. Building its own is possible, but will take time. As for chips: it can't buy top-end GPUs anymore and while it can design its own, it cannot fab them because they are forbidden from using bleeding edge fabs from TSMC.

As a result, China's AI efforts will be also-rans until and unless it manages to fix these two vulnerabilities. Probably not what people here want to hear, but I also attacked Zero Covid when lots of folks here defended it long past its usefulness. Harsh truths need to be told.
I personally don't agree with this view it just wrong in context of computing power. China can definitely create the computing needs to train and run those AI model. It's will just take up more server space and energy.

Well China has access to cheaper energy then western or at least European locations. Also can also make the needed infrastructure like server locations cheaper.

Also China created one of the if not the fastest super computer on the world using domestic designed 14nm chips if I'm not mistaken.
 
Last edited:

ACuriousPLAFan

Brigadier
Registered Member
I read couple of years ago about a Chinese plan for a huge particle collider bigger than CERN's hadron Collider. What happened to that plan? Its called the Circular Electron–Positron Collider (CEPC)
I also remember this and i also remember the west freaking out how this should be a global operation and not a one country project o_O..
Its kind of like that big ass radio telescope China made first it wasn't needed and now that the US neglected maintaining their own im sure the scientific community is glad China made one.
From the Chinese Academy of Sciences (CAS) just around a week ago:

Please, Log in or Register to view URLs content!

Roughly translated:
Wang Yifang Representative: The Ring Positive & Negative Electron Collider <<Technical Design Report>> is Basically Completed

Published:2023-03-11
Source: Science and Technology Daily Lu Chengkuan
  There is another new development in the Ring Positron Collider (CEPC), which is of high interest to the international high-energy physics community. "We have basically completed the Technical Design Report of the CEPC, which will undergo international review this year." Wang Yifang, a member of the National People's Congress and director of the Institute of High Energy Physics at the Chinese Academy of Sciences, revealed in a recent interview with Science and Technology Daily.
  CEPC is a scenario proposed by Chinese scientists in 2012 about the future high-energy collider. Scientists hope to use it to study some key unsolved scientific questions and new physical laws such as the Higgs particle, the early evolution of the universe, the loss of antimatter, and to search for dark matter and other new particles.
  In 2018, the Conceptual Design Report of CEPC was officially released. According to the conceptual design, the CEPC will be a "big ring" of 100 km in circumference, built 50-100 meters underground.
  "With the support of the CEPC pre-research project, we have overcome a number of key technical difficulties such as superconducting high-frequency cavity, enhancer extremely weak magnet, vacuum coating, digital beam current measurement and control equipment, and developed related prototypes. And the superconducting high-frequency cavity has reached the best international level." Wang Yifang said, superconducting high-frequency cavity can be accelerated by a very high energy efficiency to charged particles, equivalent to the modern particle gas pedal "engine".
  Wang Yifang introduced: "We developed the prototype of superconducting high-frequency cavity, technical indicators are absolutely international leading, for China's construction of international leading high-frequency free electron laser device and the future high-energy positron collider to provide technology and equipment assurance."
 

ACuriousPLAFan

Brigadier
Registered Member
I could see it as a minor supplement to the global supply. Hopefully lithium becomes less and less needed as sodium ion batteries start getting more popular. One way it could be done economically is to bundle all seawater extraction projects in one plant.

When you desalination seawater, you already have to pump millions of tons of seawater inland and filter it, producing a concentrated brine. Instead of dumping this brine straight back into the oceans, you use this concentrated brine to extract uranium, bromine, gold, lithium, sodium, potassium, magnesium and whatever valuable minerals or elements that you need. This would save cost, since you're already pumping so much water inland, might as well do something with the brine before dumping it into the ocean, not to mention that you're already filtering the water for bio-fouling and as concentrated brine, it should be easier to extract whatever minerals or elements you need since the concentration should be much higher than seawater.

Of course in practice I have no idea if all this extraction processes will play nice with each other, seeing as some of them will need chemical treatment that may interfere with each other. I only know that lithium extraction needs an active electrochemical process and that uranium extraction involves fibermats that can absorb uranium passively. But it could save a lot of cost, if you're getting fresh water and a few other important industrial elements/compound out of your brine rather than just lithium.

If they do it this way, they could piggyback off China's growing desalination capacity and plants.
Speaking of which, once such technology becomes commonplace, China would need longer coastlines than what she is having right now. I guess Taiwan rises even further in importance, without being overtly obvious.

Though, when the time comes, I think we can expect Washington DC to scream "Chyna is a national security threat! Look how they are conquering and exploiting seawater! Sanction Chyna NOOWWW!!"

America also doesn't realize that you don't need a 100 kilowatt hour tesla with 900km range for daily life, most chinese are happy with a small mini-car with a 20 kilowatt hour battery that can go 200km on a single charge, which suits LFP and sodim ion perfectly. Their shock at the wuling mini EV popularity was amazing to see.
This is coming from Americans' obsession with huge objects. Big cars, big trucks, big houses, big lawns, big food, big supermarkets, big expressways, big expressways - You name it.

In fact, I found this recently-made video from Not Just Bikes on SUVs:

His focal point is this - Americans are so obsessed with "large vehicles = safer" and the "big is better and cooler" mantras - Such that "oversized" SUVs and pickup trucks become the most popular vehicles in the US car market compared to normal-sized compacts and sedans; how bad they are for road and urban environments; and how improper & unsafe they are for other road-users and pedestrians, etc.

I believe this translates pretty well to explaning why Americans are shocked at how popular and how common smaller, compact cars and minicars (i.e. Kei-cars in Japanese) are found on the streets and roads in not just China, but also Japan, Europe, India and elsewhere, for instance.
 
Last edited:

Serb

Junior Member
Registered Member
China's AI effort is dictated by the same two forces that shape Western AI: chips and software.

China can longer use CUDA or other software stacks. Building its own is possible, but will take time. As for chips: it can't buy top-end GPUs anymore and while it can design its own, it cannot fab them because they are forbidden from using bleeding edge fabs from TSMC.

As a result, China's AI efforts will be also-rans until and unless it manages to fix these two vulnerabilities. Probably not what people here want to hear, but I also attacked Zero Covid when lots of folks here defended it long past its usefulness. Harsh truths need to be told.

Could those photonic chips change that?
 

Andy1974

Senior Member
Registered Member
China's AI effort is dictated by the same two forces that shape Western AI: chips and software.

China can longer use CUDA or other software stacks. Building its own is possible, but will take time. As for chips: it can't buy top-end GPUs anymore and while it can design its own, it cannot fab them because they are forbidden from using bleeding edge fabs from TSMC.

As a result, China's AI efforts will be also-rans until and unless it manages to fix these two vulnerabilities. Probably not what people here want to hear, but I also attacked Zero Covid when lots of folks here defended it long past its usefulness. Harsh truths need to be told.
Well, you are missing the entire data picture here. All the software and chips are useless without high quality annotated data, of which China has plenty. They identified early that datasets are the key, and are building dataset exchanges which will give it an unbeatable advantage.
 

Eventine

Junior Member
Registered Member
Well, you are missing the entire data picture here. All the software and chips are useless without high quality annotated data, of which China has plenty. They identified early that datasets are the key, and are building dataset exchanges which will give it an unbeatable advantage.
China is rich in certain types of data, not others.

The English language internet is larger than the Chinese language internet, so for training something like ChatGPT or StableDiffusion, the West actually has access to more "language friendly" data than China; and in the worst case scenario, they can even limit Chinese access to the English language internet via IP bans, which won't be air tight but could be an additional obstacle for Chinese companies to overcome since the global language remains English.
 

Andy1974

Senior Member
Registered Member
China is rich in certain types of data, not others.

The English internet is larger than the Chinese internet, so for training something like ChatGPT or StableDiffusion, the West actually has access to more "language friendly" data.
You are discussing raw data, I am talking about datasets.
 

tphuang

Lieutenant General
Staff member
Super Moderator
VIP Professional
Registered Member
China's AI effort is dictated by the same two forces that shape Western AI: chips and software.

China can longer use CUDA or other software stacks. Building its own is possible, but will take time. As for chips: it can't buy top-end GPUs anymore and while it can design its own, it cannot fab them because they are forbidden from using bleeding edge fabs from TSMC.

As a result, China's AI efforts will be also-rans until and unless it manages to fix these two vulnerabilities. Probably not what people here want to hear, but I also attacked Zero Covid when lots of folks here defended it long past its usefulness. Harsh truths need to be told.
I actually posted on this on the chip thread. Don't overrate CUDA. You basically have Baidu here implementing different hardware on its Paddlepaddle framework. Nvidia, Kunlun & Cambrian chips are all used here. Nvidia chips are better, but you can also just have more of the domestic chips todo the same task. They have a 4 EFLOPS data center upgraded just for this.

This has nothing to do with 0-Covid. If you have evidence for what you are saying, then make your point, but what you are saying is just not accurate.

I tweeted about it here
Erniebot announcement was nothing impressive. It's still in testing phase. That's why only a small group of users are invited to participate. Only through usage & AI training will this get better.
 
Top