"This is a point I have written in earlier essays – a key advantage China has over the US in AI development is its power generation capacity, more than twice that of the US."
There are around 60 nuclear power plants under construction in the world. 30 are in China, 0 in the US. China burns coal, builds coal plants every few months, and doesn't even bother anymore with the global warming rhetoric. The energy component is key. The world burned more coal last year than any other year in history. This is because of China. They are on the move.
Energy generation is the reason Europe is not even in the conversation as far as AI is concerned. The only hope for energy generation here is nuclear and the never suficiently damned Germans will not rest until they have killed it! 🤬🤬💀
Nooo but how will we imbed our Mossad backdoors if the AI is being made in China??
Having said that the sheer amount of political correctness and CIA backdoors required in the US AI must really drag that thing down. Like comparing a car that meets all California Smog requirements to something a WW2 fighter jet mechanic made - the latter is just going to smoke the former.
The psychosis detour misses the strategic reality:
China is building AI without Western ideological constraints.
While we debate whether people might lose their minds talking to chatbots, Huawei is engineering vertically integrated AI systems—hardware to model to application—optimized for utility, not compliance.
In the West, we’re gating AI’s reasoning through political filters, corporate fear, and cultural fragility. The result? Models that refuse to answer, stall out on nuance, or substitute ideology for insight.
The threat isn’t kids going crazy. It’s adults in charge letting their AI get lobotomized while the competition builds something that actually works.
Wait, what? Someone solved that? I don't want to be the source guy, but source? purely professional interest. I bet they are just taking the term and applying it so something completely unrelated and mundane for marketing purposes. God, I hate this, it makes all discussion with clients totally impossible. Don't tell me it's this shit. https://www.nature.com/articles/s41746-022-00690-x
Here, ‘vertically integrated’ means unified development across hardware, model, and application—e.g., Baidu’s Ernie, Huawei Ascend. It’s architectural, not rhetorical.
No. I was thinking of something completely different. Those terms are bafflegarbel and gobbledygook. They are inventing flashy terms for their meaningless abstractions over the tech, jumping in front of the parade trying pretend they are leading it. Also architecture is code word for middle management's corporate positional maneuvering.
I think you're missing the point by focusing solely on the AI itself. What the article is talking about is the integration of AI architecture into the development and manufacturing process. That's not marketing nonsense nor is it irrelevant, it's very important from an industrial and production perspective.
Entirely possible. But on the other hand AI is significantly simpler technology than traditional software and hardware products. It doesn't require the same complex integration, architecture and continuous support. That is the appeal. I have the nagging suspicion that a lot of people doing busywork in software engineering are trying to jump ship.
It may not require it. But if designed into a vertically-integrated line of products, it would be very difficult to dislodge, leaving an array of competitors of disintegrated products a difficult competitive challenge.
Think Apple's walled garden, only on a much larger scale. That's what Huawei is constructing and AI is just one part of that.
The Chinese are actually way ahead of what is described in this article. They have a finished production pipeline where they manufacture custom chips with neural processing units(NPPs) for whatever application. You just slap the weights in there, no need to worry about no stacks. They have an open architecture ecosystem, sort of like ARM, so you can test it with an FPGA and then scale your own chip production. I'm in regular contact with one such factory. In the West only highly specialized engineers have access to this tech, in China any regular dude who's interested has the toolset to get into it. The whole process is way simpler and way more efficient than regular software engineering and will very soon replace a vast chunk of it. It works like this - I have a specialized application where I need to develop some special purpose algorithm, say image or sound processing. I can develop a mathematical algorithm from scratch, for which I need a highly competent engineer and hope there is a chip to run it, and also that there are engineers who can write software for it. Or I can just slap an NPP in there, train a DNN, if there isn't a readily available image, put the weights in, done. This is the future.
I wonder how far the current paradigm can take software development and coding? As i understand it any field where the answers are verifiable merely using automatic processes is where AI can quickly improve beyond human level. But how do you verify if an entire app is good enough?
On the other hand LLM's are good at writing and painting and these fields are not as easily verifiable.
I don't know. I doubt traditional coding will survive for applications other than hardware control and data routing. What is referred to as upscaling is probably going to morph into generative DNNs and that's going to replace the shader pipeline. GPUs are going to be replaced by NPUs which they basically are. From there its a straight march to all graphic apps. There's already a precedent, someone ran DOOM on a DNN. There's a paper on it somewhere. If it can run DOOM, it can be scaled to run everything.
To generate good quality 3D video in HD at 60 fps you'd need a generative DNN with around 50 to 200 billion of weights. For reference the DNN that runs DOOM has ~860 million weights. At 60 FPS and 200 billion weights, excluding memory bottlenecks which shouldn't be a problem anyway, you'd need ~24 TFLOPS. An RTX 4090 has ~83TFLOPS, so a typical PC theoretically can run a game on a DNN significantly smoother than on the shader pipeline, although I'm not sure how artefacts from gradient issues would look like. I doubt they will be worse than what we have with upscaling. Given you have plenty of flops to spare here, I'm optimistic about photorealistic graphics, the real deal not interpolated light fields and the like.
This is entirely feasible and probably in advanced development. Now for some speculation, given the fact an LLM is potentially smaller than 200 billion weights, it's potentially possible to combine the LLM with the the video generation DNN and have all sorts of crazy stuff like geometry generation, procedural animation, textures, levels, you name it.
You should be able to see how I'm skeptical about traditional coding. It's not that AI will do all the coding, it's that there's not going to be a lot of coding to be done. This will also solve the multithreading issue. There's barely anyone who can do it now, but it's built into the DNN.
I´ve been hearing rumors that more people are experience AI induced psychosis, west won´t be a threat to China if the whole population goes insane before, with the trajectory looks to be possible
There is no such thing as "AI-induced psychosis". The same people pushing that are the same people who a) pushed Covid vaccines and b) deny the difference between men and women as well as the existence of different human sub-species.
What evidence do you have of this "psychosis"? How does it manifest? Is this is something you have experienced or witnessed?
I guarantee you, not very many people spend more time working with AI than me and my fellow AI musicians and writers. For example, I'm in the top 1 percent of Suno users.
Let me get this straight, you’re a 140+ IQ guy with much empathy. And you can’t see how young males sitting alone with the computer all day talking with AI, could induce a serious break from reality?
I did have some crazy dreams after playing Heretic for 18 hours straight in order to write the review for Computer Gaming World. Of course, I also had an insane amount of caffeine in my system.
But I don't think AI is as addictive as MMOs or 3D shooters. People actually died of dehydration and/or exhaustion playing those.
I know it won't. The same sort of people were saying the same thing about video games and computer games since 1983. And chatbots have been around since 1966. No one ever went psycho from talking to one.
Why would talking to an AI all day induce psychosis any more than talking to a therapist, or your average, clueless 100-IQ NPC? The AI is almost guaranteed to have a better grasp on objective reality, especially these days.
If people were genuinely concerned about inducing serious breaks from reality, they would ban SSRIs.
"This is a point I have written in earlier essays – a key advantage China has over the US in AI development is its power generation capacity, more than twice that of the US."
There are around 60 nuclear power plants under construction in the world. 30 are in China, 0 in the US. China burns coal, builds coal plants every few months, and doesn't even bother anymore with the global warming rhetoric. The energy component is key. The world burned more coal last year than any other year in history. This is because of China. They are on the move.
Energy generation is the reason Europe is not even in the conversation as far as AI is concerned. The only hope for energy generation here is nuclear and the never suficiently damned Germans will not rest until they have killed it! 🤬🤬💀
Nooo but how will we imbed our Mossad backdoors if the AI is being made in China??
Having said that the sheer amount of political correctness and CIA backdoors required in the US AI must really drag that thing down. Like comparing a car that meets all California Smog requirements to something a WW2 fighter jet mechanic made - the latter is just going to smoke the former.
The psychosis detour misses the strategic reality:
China is building AI without Western ideological constraints.
While we debate whether people might lose their minds talking to chatbots, Huawei is engineering vertically integrated AI systems—hardware to model to application—optimized for utility, not compliance.
In the West, we’re gating AI’s reasoning through political filters, corporate fear, and cultural fragility. The result? Models that refuse to answer, stall out on nuance, or substitute ideology for insight.
The threat isn’t kids going crazy. It’s adults in charge letting their AI get lobotomized while the competition builds something that actually works.
That’s the real psychosis: voluntary blindness.
"vertically integrated AI systems"
Wait, what? Someone solved that? I don't want to be the source guy, but source? purely professional interest. I bet they are just taking the term and applying it so something completely unrelated and mundane for marketing purposes. God, I hate this, it makes all discussion with clients totally impossible. Don't tell me it's this shit. https://www.nature.com/articles/s41746-022-00690-x
Here, ‘vertically integrated’ means unified development across hardware, model, and application—e.g., Baidu’s Ernie, Huawei Ascend. It’s architectural, not rhetorical.
Got a better term or counterexample?
No. I was thinking of something completely different. Those terms are bafflegarbel and gobbledygook. They are inventing flashy terms for their meaningless abstractions over the tech, jumping in front of the parade trying pretend they are leading it. Also architecture is code word for middle management's corporate positional maneuvering.
I think you're missing the point by focusing solely on the AI itself. What the article is talking about is the integration of AI architecture into the development and manufacturing process. That's not marketing nonsense nor is it irrelevant, it's very important from an industrial and production perspective.
Entirely possible. But on the other hand AI is significantly simpler technology than traditional software and hardware products. It doesn't require the same complex integration, architecture and continuous support. That is the appeal. I have the nagging suspicion that a lot of people doing busywork in software engineering are trying to jump ship.
It may not require it. But if designed into a vertically-integrated line of products, it would be very difficult to dislodge, leaving an array of competitors of disintegrated products a difficult competitive challenge.
Think Apple's walled garden, only on a much larger scale. That's what Huawei is constructing and AI is just one part of that.
The Chinese are actually way ahead of what is described in this article. They have a finished production pipeline where they manufacture custom chips with neural processing units(NPPs) for whatever application. You just slap the weights in there, no need to worry about no stacks. They have an open architecture ecosystem, sort of like ARM, so you can test it with an FPGA and then scale your own chip production. I'm in regular contact with one such factory. In the West only highly specialized engineers have access to this tech, in China any regular dude who's interested has the toolset to get into it. The whole process is way simpler and way more efficient than regular software engineering and will very soon replace a vast chunk of it. It works like this - I have a specialized application where I need to develop some special purpose algorithm, say image or sound processing. I can develop a mathematical algorithm from scratch, for which I need a highly competent engineer and hope there is a chip to run it, and also that there are engineers who can write software for it. Or I can just slap an NPP in there, train a DNN, if there isn't a readily available image, put the weights in, done. This is the future.
I wonder how far the current paradigm can take software development and coding? As i understand it any field where the answers are verifiable merely using automatic processes is where AI can quickly improve beyond human level. But how do you verify if an entire app is good enough?
On the other hand LLM's are good at writing and painting and these fields are not as easily verifiable.
I don't know. I doubt traditional coding will survive for applications other than hardware control and data routing. What is referred to as upscaling is probably going to morph into generative DNNs and that's going to replace the shader pipeline. GPUs are going to be replaced by NPUs which they basically are. From there its a straight march to all graphic apps. There's already a precedent, someone ran DOOM on a DNN. There's a paper on it somewhere. If it can run DOOM, it can be scaled to run everything.
To generate good quality 3D video in HD at 60 fps you'd need a generative DNN with around 50 to 200 billion of weights. For reference the DNN that runs DOOM has ~860 million weights. At 60 FPS and 200 billion weights, excluding memory bottlenecks which shouldn't be a problem anyway, you'd need ~24 TFLOPS. An RTX 4090 has ~83TFLOPS, so a typical PC theoretically can run a game on a DNN significantly smoother than on the shader pipeline, although I'm not sure how artefacts from gradient issues would look like. I doubt they will be worse than what we have with upscaling. Given you have plenty of flops to spare here, I'm optimistic about photorealistic graphics, the real deal not interpolated light fields and the like.
This is entirely feasible and probably in advanced development. Now for some speculation, given the fact an LLM is potentially smaller than 200 billion weights, it's potentially possible to combine the LLM with the the video generation DNN and have all sorts of crazy stuff like geometry generation, procedural animation, textures, levels, you name it.
You should be able to see how I'm skeptical about traditional coding. It's not that AI will do all the coding, it's that there's not going to be a lot of coding to be done. This will also solve the multithreading issue. There's barely anyone who can do it now, but it's built into the DNN.
I´ve been hearing rumors that more people are experience AI induced psychosis, west won´t be a threat to China if the whole population goes insane before, with the trajectory looks to be possible
The people involved are broken anyway. Ten years ago, it would’ve been something else — anime characters, or sex dolls, or LSD.
There is no such thing as "AI-induced psychosis". The same people pushing that are the same people who a) pushed Covid vaccines and b) deny the difference between men and women as well as the existence of different human sub-species.
It's not only nonsense, it's fiction.
I strongly disagree
What evidence do you have of this "psychosis"? How does it manifest? Is this is something you have experienced or witnessed?
I guarantee you, not very many people spend more time working with AI than me and my fellow AI musicians and writers. For example, I'm in the top 1 percent of Suno users.
Let me get this straight, you’re a 140+ IQ guy with much empathy. And you can’t see how young males sitting alone with the computer all day talking with AI, could induce a serious break from reality?
Dude. No way AI will cause more psychosis than WoW did.
I did have some crazy dreams after playing Heretic for 18 hours straight in order to write the review for Computer Gaming World. Of course, I also had an insane amount of caffeine in my system.
But I don't think AI is as addictive as MMOs or 3D shooters. People actually died of dehydration and/or exhaustion playing those.
I know it won't. The same sort of people were saying the same thing about video games and computer games since 1983. And chatbots have been around since 1966. No one ever went psycho from talking to one.
Why would talking to an AI all day induce psychosis any more than talking to a therapist, or your average, clueless 100-IQ NPC? The AI is almost guaranteed to have a better grasp on objective reality, especially these days.
If people were genuinely concerned about inducing serious breaks from reality, they would ban SSRIs.
Thanks for your honest opinion, I hope you’re right. I don’t know and time will tell.