Not a word on Chinese models being censored in the article. What an odd omission.
It should also be pretty obvious that this is following the usual Chinese MO of using massive state subsidies to destroy the international competition with impossibly low dumping prices. We are seeing this in all sorts of sectors.
tech has been subsidizing ai costs by magnitudes for years trying to make fetch happen, slop is slop. it’s overvalued like crazy and the first hint of market competition has drained trillions from the stocks because it’s an overvalued bubble. if china can do that by releasing competition then ok. maybe we should all be putting these trillions in things actually useful to humans.
If anything, this is just the start of an arms race. Do you really expect the Western competition to just stop what they are doing, because a single Chinese model performs well in a handful of synthetic tests that it was probably optimized to score well in?
I’m not a fan of AI slop either, on the contrary, but let’s be realistic here.
an arms race for what? more efficient slop? most of their value comes from the expected exclusivity - that say openai is the only one who can run something like o1. deepseek has made that collapse. i doubt they will stop doing stuff, but i dont think you understand the nature of the situation here.
also lol, “performs well in synthetic tests it was optimized to score well in” yes that literally describes every llm. Make no mistake: none of this has a real use case. not deepseek’s model, not openai’s, not apples, etc. this is all nonsense, literally. the stock market lost 2 trillion dollars overnight because something that doesnt have a use case was one upped by something else that also doesnt have a use case. it’s very funny.
Overvalued - as in, less useful than it seems to be - probably, but the costs of running it are immense and they are certainly not that much lower in China (despite low energy prices due to nonexistent environmental standards), given the hardware embargoes they are under, forcing them to use less efficient hardware.
i believe one of the big advancements with deepseek r1 is their method of adding the reasoning component is novel and very very efficient. i haven’t checked it out, but it could legitimately just be more efficient to run
“due to non existent environmental standards” buddy unless your from rural northern Europe or the mountains in the Himalayan wtf are you talking about. Compared to America, china is much much less polluted per person with people personally accounting for less than the average American or westerner.
Meanwhile, environmental activists are routinely being persecuted by the state in order to silence them. That’s totally what a country with a great environmental track record would do.
Did you just really send me iqair with such a bold statement without even checking the link itself? DUDE, JERUSALEM AMD SOFIA IN BULGARIA ARE MORE POLLUTED THAN SHANGHAI. For a nation of 1.4x thats seriously impressive. None of their cities crack top 5.
Wisdom is chasing you, but you are faster
Barely any of your sources account for recent data, and focuses on the time during the most rapid transformation of a massive civilization in human history
And again, they still per person use significantly lesss co2 than an America, what about that?
Chinese freedom of opinions and knowledge is some of the worst on the planet and xi jinping is riding off the coattails of Deng Zhao pings success in opening up china.
There, now here’s another truth bomb, china sucks and treats it’s citizens like robots but yet somehow I’d rather trust the historic super power which historically didn’t bother nobody outside of it’s little sphere unlike the current western superpowers who exploit the entire planet to make that “first world” experience for like 10% of humanity
Look, China isn’t the devil or anything, they do lots of things better then the US. They want their “little sphere” to be Earth though. They have been making moves to compete with US influence all over the world for years now, and they’re not doing it out of the goodness of their hearts.
Honestly, the more I look into China, the more I realize the worst thing about it is that they’re very much like the US, no matter how much both sides would deny it. The US needs to be taken down a peg or two, but replacing it with a different empire isn’t the way to go. We need a world without superpowers, not to try and find the “good” one. They’ll always go bad once they get to the top. That’s just how massive power structures work.
Overvalued - as in, less useful than it seems to be
Uh, no… “Value”, as in quality/performance for the price. You can literally overpay for anything in this world, just look at the luxury market.
I’m not claiming to know enough about AI or LLMs, but I don’t think the first to market or the most prominent always set the price. So I think we’ll have to see what the accepted price actually turns out to be…
Interesting about Chinese energy market is that in recent decades they’ve been investing heavily in solar power. Once they’ve figured out grid energy storage, running LLMs shouldn’t be a problem anymore.
Another option is to skip most of the grid storage and just spam solar panels. Rely on batteries only to get you through the night, not to bridge power across seasons. Build enough panels that your country can meet its needs even on a cloudy day in winter. Then you have reasonable power costs in the winter and nearly free electricity the rest of the year.
You could see a lot of energy-intensive industries becoming seasonal. We have a crop growing season, a school season, and sports seasons. Why not an “AI model training” season?
That would be possible, but seasonal production has some serious drawbacks.
Let’s say you have a steel mill with several production lines, solar powered arc furnaces, and enough batteries to keep production running through the night. During the summer you can continue production 24/7, but in the winter you’ll have to shut down completely, because there’s not enough energy to keep even a single production line running. This means that there will be wild fluctuations in a variety of things:
number of employees on site
rate of steel produced
demand for storage space for raw materials and steel products
demand for logistics
demand for maintenance
This means, that in order to deal with the fluctuations, you would need to have lots of spare capacity in pretty much everything: More machines, more people, more money. If you could keep the production steady throughout the year, you could do so with less. Also, what will the employees do during the winter? The skiing resorts can’t possibly employ all of them.
In the winter you’ll have plenty of time to fix anything that’s broken, but if there’s an unscheduled shutdown during the summer, you’re suddenly going to need lots of maintenance personnel and materials. Incidentally, those would be in short supply in the summer, because all the other factories would have the same problem. You would need to have lots of spare capacity in maintenance as well.
The AI industry should be fine, since you could train models when energy is cheap. Oh, but what if the summer isn’t long enough for you to update all your models? Simply just buy more computers so you have more spare capa… Oh, it’s the steel mill problem all over again. Oh, but what about the people who use the models during the winter? Maybe you could charge your customers double the price during the winter so that the traffic would be reduced to a reasonable level. Fortunately though, wind power and other renewables could help with the winters, but having more grid energy storage would make things run smoother.
Is it cheaper to store steel stock in a warehouse or terrawatt-hours of electricity in a battery farm?
Is it cheaper to perform maintainance on 2 or 3x the number of smelters or is it cheaper to maintain millions of battery or pumped hydro facilities?
I’m sure production companies would love it if governments or electrical companies bore the costs of evening out fluctuations in production, just like I’m sure farmers would love it if money got teleported into their bank account for free and they never had to worry about growing seasons. But I’m not sure that’s the best situation for society as a whole.
EDIT: I guess there’s a third factor which is transmission. We could build transmission cables between the northern and southern hemispheres. So, is it cheaper to build and maintain enormous HVDC (or even superconducting) cables than it is to do either of the two things above? And how do governments feel about being made so dependent on each other?
We can do a combination of all three of course, picking and choosing the optimal strategy for each situation, but like I said above I tend to think that one of those strategies will be disproportionately favorable over the others.
Honestly, this all seems like small potatoes. We’re trying to save our species from extinction here. We’re trying to maintain the standard of living that came with the Industrial Revolution without burning out planet to a cinder.
If doing so means our steel industry runs 10% less efficiently, I really don’t give a damn.
I don’t know exactly how bad would it be, but my guess is that it would have a significant impact on the prices consumer pay for everything. In the past few hundred years, we’ve taken all sorts of nasty shortcuts that have allowed us to produce things at very low prices. If you want to do things the right way, it’s going to cost much more.
Burning fossil fuels is just one of those unwise shortcuts that need to be reversed completely. In the long run, we’re going to have to bury all the carbon we’ve dug up, and that’s going to be incredibly expensive too.
Fortunately though, the downsides of intermittent energy production can still be mitigated with various grid energy storage technologies. The way I see it, investing into them is crucial.
Energy costs money in China too, they still have coal plants and crazy energy cutback mandates every once in a while.
The truth of the matter is that you need the user interactions from the free model to train and that value cannot be understated and if you are playing catchup, then it’s a must. No one would use the Chinese model if it was a shit service.
There’s a strong argument that any consumer facing chatbot AI is “censored”. I’ve had chatGPT clam up in bizarre ways after it misinterprets what I’m asking. It just depends on company owning the product and what they view their legal exposure to be.
Also, we are applying huge govt subsidies to ai industry based on thin value evidence at this very moment. And we provide subsidies for many of our industries to help prop them up, sometimes to hugely bad effect. It’s what countries do to build, maintain and win industrial arms races.
Deepseek-R1 is open source and you can download it and run it offline. I’m not a power user but was able to get a functioning offline version of the 32B distill model running on a spare machine I had in a hour or so from scratch. I used online deepseek for most of the process to provide instructions and troubleshoot. I can’t comment on how amazing it is, other than to say so far it’s felt about as good as my interactions with GPT4 on the free chatGPT tier. In both cases I remain skeptical about their deep business use outside of certain areas.
From what I’ve read, you can use the base, and methodology and train your own new model if you have the technical ability and desire (rumor is meta AI has shelved their WIP and adopted deepseek as their new basis). This would imply that if you wanted to be able to talk to your LLM about topics like Taiwan, you could absolutely set up a model that would do that.
If you’re going to accuse China of state censorship, then I suppose you are also vehemently opposed to the censorship we apply to our media, social media and “AI” platforms, and since you dislike the lack of journalistic integrity in this article for pointing out that state censorship you would support similar caveats being added to articles about OpenAI, Meta, X in regards to how they handle issues like Gaza, Culture War topics and coverage of political candidates?
It’s fair to bring up comparisons when your critique is claiming an imbalance in portrayal between the “realities” of ai development in China and the US.
Wow what even is beehaw, I had no idea. At least China is honest about what they’re doing. The amount of bad faith in these replies is insane.
If you’re a shill, fine, good job. But if you’re not, have you paid any attention to the real world around you? We spent the last year enabling genocide, and the best fruits of our over-hyped tech and intellectual innovation factories are being revealed as the bullshit that most people always understood them to be.
The fact that you can accuse me of being dishonest, while providing no basis or evidence, while multiple federal agencies are under a strict gag order from any communication or purchasing with outside contacts… I mean really?
Like are you guys just another CIA adjacent cutout that believes in identity politics and SSRIs but has zero ability to critically assess the actual world around them?
At least China is honest about what they’re doing.
Peak comedy right there. You’re proudly defending an imperialist regime that is engaged in multiple genocides right now and has the worst body count in all of human history - but as long as it’s under the red banner, that’s fine by you. How shallow can you be? Are you willfully ignorant of all of these past and current crimes against humanity or does the goal of a Communist utopia (as if hyper-capitalist China would ever get there) justify it all?
Relative to people in their country, sure. But China can’t and isn’t interested in flying over to the US to arrest you if you talk to their AI models about Taiwan being its own country, whereas no one should have any doubt that OpenAI or any other US AI company is happy to tell Trump’s administration who’s been asking it about LGBT+ issues or other topics the US government is now against.
It’s not whataboutism anymore, it’s literally that both are evil authoritarian governments, but one (US) has physical access to US users, and the other doesn’t.
There’s a strong argument that any consumer facing chatbot AI is “censored”.
If the model is not allowed to spew Nazi propaganda or tell the user to end themselves, that is censorship. Censorship is not automatically bad, but the kind of censorship can make it bad.
This reeks of excluding all nuance to equate two things that are equal only at surface level. You’re bad because you punched the other person (ignoring that they stabbed your SO 15 times and kicked your dog across the room).
Chinese state censorship is well researched and extremely well documented. It does not equate to censorship against violent or inappropriate language. It is political censorship.
At best, western models are biased, not politically censored. You can make them say just about anything, but they will bias towards a particular viewpoint. Even if intentional, this is explainable by evaluating their training data, which itself is biased because western society is biased. You are not prevented from personally expressing or even convincing a western model from expressing dissenting political viewpoints.
I’m gonna take a second stab at replying, because you seem to be arguing in good faith.
My original point is that online chatbots have arbitrary curbs that are built in. I can run GPT 2.5 on my self host machine, and if I knew how to do it (I don’t) I could probably get it to have no curbs via retraining and clever prompting. The same is true of the deepseek models.
I don’t personally agree that there’s a huge difference between one model being curbed from discussing xi and another from discussing what the current politics du jour in the western sphere are. When you see platforms like meta censoring LGTBTQ topics but amplifying hate speech, or official congressional definitions of antisemitism including objection to active and on-going genocide, the idea of what government censorship is and isn’t becomes confusing.
Having personally received the bizarre internal agency emails circulating this week encouraging me to snitch out my colleagues to help root out the evils of DEIA thought in US gov’t the last week has only crystallized it for me. I’m not sure I care that much about Chinese censorship or authoritarianism; I’ve got budget authoritarianism at home, and I don’t even get high-speed rail out of the bargain. At least they don’t depend on forever wars and all of the attendant death and destruction that come with them to prop up their ponzi-scheme economies. Will they in the future, probably? They are basically just a heavily centralized/regulated capitalist enterprise now, so who knows. But right now? Do they engage in propaganda? Cyber-espionage? Yes and Yes. So do we, so do you, so does everyone who has a seat at the geopolitical table and the economy to afford it.
The point of all of this isn’t US GOOD CHINA BAD or US BAD CHINA GOOD. The article is about the deepseek models tearing out the floor of US dominance in AI. Personally, having deployed it and played with it, yeah. None of these products are truly useful to me yet, and I remain skeptical of their eventual value, but right now, party censorship or not, you can download a version of an LLM that you can run, retrain and bias however you want, and it costs you the bandwidth it took to download. And it performs on par with US commercial offerings that require pricey subscriptions. Offerings that apparently require huge public investment to keep afloat.
Where I disagree with you is not that the US is bad - the US is terrible, and there is plenty of evidence of that. I don’t even disagree with there being censorship in the US. In fact, Trump is objectively a piece of shit who wants nothing more than to become Xi/Putin himself.
What I disagree with is equating censorship in the US with Chinese censorship. I can call Trump a piece of shit online without worrying that the FBI will show up at my door. The models that are trained in the west will happily entertain any (non-violent) political discussions I want. There may be bias, and Trump may be trying to create censorship, but it’s not quite to that level yet.
Having personally received the bizarre internal agency emails circulating this week encouraging me to snitch out my colleagues to help root out the evils of DEIA thought in US gov’t the last week has only crystallized it for me.
I am concerned that the US will become as bad as China in terms of censorship, which is part of why I’m trying to leave right now. However, it’s not there yet. They are not yet equal, nor are they even close yet.
Yeah ok, I do basically agree with you. It’s not an accurate equivalency, yet. We’re trending bad though. I’d say the example of Stephen Miller sort of accidentally hinting that they shut down USAID because they all donated to the Harris campaign had some chilling implications for example. He could just be assuming that, since that’s a safe assumption for populous urban areas generally, but they could also have cross checked lists of employees against political contributions.
You say Chinese state censorship is an understood quantity. Could be. But I’d say that my points about equivalencies are to illustrate that what we think is true, is often much more grey. I’ve been to China, and while I was impressed and shocked at how much more advanced it was than I expected, I also couldn’t imaging living there. It doesn’t change the fact that a stagnant late-stage capital mafia state that lives off defense contracting is performing ooorly against a centrally controlled capitalist state that has set different priorities (that’s right boy, deepseek-r1 is a side project of a…. CHINESE HEDGE FUND). It’s value neutral. But if you dismiss reality based on a conception of political censorship that I doubt you’ve deeply engaged with, enjoy.
The so called free market certainly didn’t seem to take much reassurance in deepseek being compromised by communist censorship this morning though. Probably because the deepseek news isn’t exceptional because of China, or what it is, but because of what it isn’t, compared to the bloated tech carcasses that the US has pinned its hopes on.
It should also be pretty obvious that this is following the usual Chinese MO of using massive state subsidies to destroy the international competition with impossibly low dumping prices. We are seeing this in all sorts of sectors.
In this case, DeepSeek is announcing the training time for their LLMl, which wall street is extrapolating costs from. No state aid involved.
Until this has been independently verified, I have my doubts. This wouldn’t be the first time for China to vastly exaggerate its technological capabilities.
Deepseek seems to have done a clever thing w.r.t. training data, by having the model train on data that was emitted by other LLMs (as far as I’ve heard). That means there is sort of “quality-pass”, filtering out a lot of the definitely bogus data. That probably leads to a smaller model, and thus less training hours.
Google engineers put out a paper on this technique recently as well.
Not a word on Chinese models being censored in the article. What an odd omission.
It should also be pretty obvious that this is following the usual Chinese MO of using massive state subsidies to destroy the international competition with impossibly low dumping prices. We are seeing this in all sorts of sectors.
tech has been subsidizing ai costs by magnitudes for years trying to make fetch happen, slop is slop. it’s overvalued like crazy and the first hint of market competition has drained trillions from the stocks because it’s an overvalued bubble. if china can do that by releasing competition then ok. maybe we should all be putting these trillions in things actually useful to humans.
If anything, this is just the start of an arms race. Do you really expect the Western competition to just stop what they are doing, because a single Chinese model performs well in a handful of synthetic tests that it was probably optimized to score well in?
I’m not a fan of AI slop either, on the contrary, but let’s be realistic here.
an arms race for what? more efficient slop? most of their value comes from the expected exclusivity - that say openai is the only one who can run something like o1. deepseek has made that collapse. i doubt they will stop doing stuff, but i dont think you understand the nature of the situation here.
also lol, “performs well in synthetic tests it was optimized to score well in” yes that literally describes every llm. Make no mistake: none of this has a real use case. not deepseek’s model, not openai’s, not apples, etc. this is all nonsense, literally. the stock market lost 2 trillion dollars overnight because something that doesnt have a use case was one upped by something else that also doesnt have a use case. it’s very funny.
Okay… But isn’t it also possible that AI is massively overvalued and this is a more reasonable price point for the technology?
Overvalued - as in, less useful than it seems to be - probably, but the costs of running it are immense and they are certainly not that much lower in China (despite low energy prices due to nonexistent environmental standards), given the hardware embargoes they are under, forcing them to use less efficient hardware.
i believe one of the big advancements with deepseek r1 is their method of adding the reasoning component is novel and very very efficient. i haven’t checked it out, but it could legitimately just be more efficient to run
“due to non existent environmental standards” buddy unless your from rural northern Europe or the mountains in the Himalayan wtf are you talking about. Compared to America, china is much much less polluted per person with people personally accounting for less than the average American or westerner.
China has some of the worst polluted cities in the world, far worse than European or American cities. Water quality is abysmal, partly due to extremely inefficient use of fertilizer and pesticides. Products exported from China are commonly exceeding limits on toxic substances. It feels like every other week, there’s another food safety scandal. Soil contamination is still worsening, in part due to extremely dirty mining practices. Chinese companies are falsifying records in order to hide excessive emissions from customers.
Meanwhile, environmental activists are routinely being persecuted by the state in order to silence them. That’s totally what a country with a great environmental track record would do.
Did you just really send me iqair with such a bold statement without even checking the link itself? DUDE, JERUSALEM AMD SOFIA IN BULGARIA ARE MORE POLLUTED THAN SHANGHAI. For a nation of 1.4x thats seriously impressive. None of their cities crack top 5.
Wisdom is chasing you, but you are faster Barely any of your sources account for recent data, and focuses on the time during the most rapid transformation of a massive civilization in human history
And again, they still per person use significantly lesss co2 than an America, what about that?
Here’s a challenge for you: Write something that is actually critical of the Chinese government. Can you do it?
Chinese freedom of opinions and knowledge is some of the worst on the planet and xi jinping is riding off the coattails of Deng Zhao pings success in opening up china. There, now here’s another truth bomb, china sucks and treats it’s citizens like robots but yet somehow I’d rather trust the historic super power which historically didn’t bother nobody outside of it’s little sphere unlike the current western superpowers who exploit the entire planet to make that “first world” experience for like 10% of humanity
Look, China isn’t the devil or anything, they do lots of things better then the US. They want their “little sphere” to be Earth though. They have been making moves to compete with US influence all over the world for years now, and they’re not doing it out of the goodness of their hearts.
Honestly, the more I look into China, the more I realize the worst thing about it is that they’re very much like the US, no matter how much both sides would deny it. The US needs to be taken down a peg or two, but replacing it with a different empire isn’t the way to go. We need a world without superpowers, not to try and find the “good” one. They’ll always go bad once they get to the top. That’s just how massive power structures work.
Uh, no… “Value”, as in quality/performance for the price. You can literally overpay for anything in this world, just look at the luxury market.
I’m not claiming to know enough about AI or LLMs, but I don’t think the first to market or the most prominent always set the price. So I think we’ll have to see what the accepted price actually turns out to be…
Interesting about Chinese energy market is that in recent decades they’ve been investing heavily in solar power. Once they’ve figured out grid energy storage, running LLMs shouldn’t be a problem anymore.
Another option is to skip most of the grid storage and just spam solar panels. Rely on batteries only to get you through the night, not to bridge power across seasons. Build enough panels that your country can meet its needs even on a cloudy day in winter. Then you have reasonable power costs in the winter and nearly free electricity the rest of the year.
You could see a lot of energy-intensive industries becoming seasonal. We have a crop growing season, a school season, and sports seasons. Why not an “AI model training” season?
That would be possible, but seasonal production has some serious drawbacks.
Let’s say you have a steel mill with several production lines, solar powered arc furnaces, and enough batteries to keep production running through the night. During the summer you can continue production 24/7, but in the winter you’ll have to shut down completely, because there’s not enough energy to keep even a single production line running. This means that there will be wild fluctuations in a variety of things:
This means, that in order to deal with the fluctuations, you would need to have lots of spare capacity in pretty much everything: More machines, more people, more money. If you could keep the production steady throughout the year, you could do so with less. Also, what will the employees do during the winter? The skiing resorts can’t possibly employ all of them.
In the winter you’ll have plenty of time to fix anything that’s broken, but if there’s an unscheduled shutdown during the summer, you’re suddenly going to need lots of maintenance personnel and materials. Incidentally, those would be in short supply in the summer, because all the other factories would have the same problem. You would need to have lots of spare capacity in maintenance as well.
The AI industry should be fine, since you could train models when energy is cheap. Oh, but what if the summer isn’t long enough for you to update all your models? Simply just buy more computers so you have more spare capa… Oh, it’s the steel mill problem all over again. Oh, but what about the people who use the models during the winter? Maybe you could charge your customers double the price during the winter so that the traffic would be reduced to a reasonable level. Fortunately though, wind power and other renewables could help with the winters, but having more grid energy storage would make things run smoother.
At to end of the day it comes down to this:
Is it cheaper to store steel stock in a warehouse or terrawatt-hours of electricity in a battery farm?
Is it cheaper to perform maintainance on 2 or 3x the number of smelters or is it cheaper to maintain millions of battery or pumped hydro facilities?
I’m sure production companies would love it if governments or electrical companies bore the costs of evening out fluctuations in production, just like I’m sure farmers would love it if money got teleported into their bank account for free and they never had to worry about growing seasons. But I’m not sure that’s the best situation for society as a whole.
EDIT: I guess there’s a third factor which is transmission. We could build transmission cables between the northern and southern hemispheres. So, is it cheaper to build and maintain enormous HVDC (or even superconducting) cables than it is to do either of the two things above? And how do governments feel about being made so dependent on each other?
We can do a combination of all three of course, picking and choosing the optimal strategy for each situation, but like I said above I tend to think that one of those strategies will be disproportionately favorable over the others.
Honestly, this all seems like small potatoes. We’re trying to save our species from extinction here. We’re trying to maintain the standard of living that came with the Industrial Revolution without burning out planet to a cinder.
If doing so means our steel industry runs 10% less efficiently, I really don’t give a damn.
I don’t know exactly how bad would it be, but my guess is that it would have a significant impact on the prices consumer pay for everything. In the past few hundred years, we’ve taken all sorts of nasty shortcuts that have allowed us to produce things at very low prices. If you want to do things the right way, it’s going to cost much more.
Burning fossil fuels is just one of those unwise shortcuts that need to be reversed completely. In the long run, we’re going to have to bury all the carbon we’ve dug up, and that’s going to be incredibly expensive too.
Fortunately though, the downsides of intermittent energy production can still be mitigated with various grid energy storage technologies. The way I see it, investing into them is crucial.
Energy costs money in China too, they still have coal plants and crazy energy cutback mandates every once in a while.
The truth of the matter is that you need the user interactions from the free model to train and that value cannot be understated and if you are playing catchup, then it’s a must. No one would use the Chinese model if it was a shit service.
There’s a strong argument that any consumer facing chatbot AI is “censored”. I’ve had chatGPT clam up in bizarre ways after it misinterprets what I’m asking. It just depends on company owning the product and what they view their legal exposure to be.
Also, we are applying huge govt subsidies to ai industry based on thin value evidence at this very moment. And we provide subsidies for many of our industries to help prop them up, sometimes to hugely bad effect. It’s what countries do to build, maintain and win industrial arms races.
Deepseek-R1 is open source and you can download it and run it offline. I’m not a power user but was able to get a functioning offline version of the 32B distill model running on a spare machine I had in a hour or so from scratch. I used online deepseek for most of the process to provide instructions and troubleshoot. I can’t comment on how amazing it is, other than to say so far it’s felt about as good as my interactions with GPT4 on the free chatGPT tier. In both cases I remain skeptical about their deep business use outside of certain areas.
From what I’ve read, you can use the base, and methodology and train your own new model if you have the technical ability and desire (rumor is meta AI has shelved their WIP and adopted deepseek as their new basis). This would imply that if you wanted to be able to talk to your LLM about topics like Taiwan, you could absolutely set up a model that would do that.
Please don’t use whataboutism to downplay state censorship.
If you’re going to accuse China of state censorship, then I suppose you are also vehemently opposed to the censorship we apply to our media, social media and “AI” platforms, and since you dislike the lack of journalistic integrity in this article for pointing out that state censorship you would support similar caveats being added to articles about OpenAI, Meta, X in regards to how they handle issues like Gaza, Culture War topics and coverage of political candidates?
It’s fair to bring up comparisons when your critique is claiming an imbalance in portrayal between the “realities” of ai development in China and the US.
It’s incredibly dishonest to equate Chinese state censorship with what the West is doing.
Wow what even is beehaw, I had no idea. At least China is honest about what they’re doing. The amount of bad faith in these replies is insane.
If you’re a shill, fine, good job. But if you’re not, have you paid any attention to the real world around you? We spent the last year enabling genocide, and the best fruits of our over-hyped tech and intellectual innovation factories are being revealed as the bullshit that most people always understood them to be.
The fact that you can accuse me of being dishonest, while providing no basis or evidence, while multiple federal agencies are under a strict gag order from any communication or purchasing with outside contacts… I mean really?
Like are you guys just another CIA adjacent cutout that believes in identity politics and SSRIs but has zero ability to critically assess the actual world around them?
Peak comedy right there. You’re proudly defending an imperialist regime that is engaged in multiple genocides right now and has the worst body count in all of human history - but as long as it’s under the red banner, that’s fine by you. How shallow can you be? Are you willfully ignorant of all of these past and current crimes against humanity or does the goal of a Communist utopia (as if hyper-capitalist China would ever get there) justify it all?
Relative to people in their country, sure. But China can’t and isn’t interested in flying over to the US to arrest you if you talk to their AI models about Taiwan being its own country, whereas no one should have any doubt that OpenAI or any other US AI company is happy to tell Trump’s administration who’s been asking it about LGBT+ issues or other topics the US government is now against.
It’s not whataboutism anymore, it’s literally that both are evil authoritarian governments, but one (US) has physical access to US users, and the other doesn’t.
If the model is not allowed to spew Nazi propaganda or tell the user to end themselves, that is censorship. Censorship is not automatically bad, but the kind of censorship can make it bad.
This reeks of excluding all nuance to equate two things that are equal only at surface level. You’re bad because you punched the other person (ignoring that they stabbed your SO 15 times and kicked your dog across the room).
Chinese state censorship is well researched and extremely well documented. It does not equate to censorship against violent or inappropriate language. It is political censorship.
At best, western models are biased, not politically censored. You can make them say just about anything, but they will bias towards a particular viewpoint. Even if intentional, this is explainable by evaluating their training data, which itself is biased because western society is biased. You are not prevented from personally expressing or even convincing a western model from expressing dissenting political viewpoints.
I’m gonna take a second stab at replying, because you seem to be arguing in good faith.
My original point is that online chatbots have arbitrary curbs that are built in. I can run GPT 2.5 on my self host machine, and if I knew how to do it (I don’t) I could probably get it to have no curbs via retraining and clever prompting. The same is true of the deepseek models.
I don’t personally agree that there’s a huge difference between one model being curbed from discussing xi and another from discussing what the current politics du jour in the western sphere are. When you see platforms like meta censoring LGTBTQ topics but amplifying hate speech, or official congressional definitions of antisemitism including objection to active and on-going genocide, the idea of what government censorship is and isn’t becomes confusing.
Having personally received the bizarre internal agency emails circulating this week encouraging me to snitch out my colleagues to help root out the evils of DEIA thought in US gov’t the last week has only crystallized it for me. I’m not sure I care that much about Chinese censorship or authoritarianism; I’ve got budget authoritarianism at home, and I don’t even get high-speed rail out of the bargain. At least they don’t depend on forever wars and all of the attendant death and destruction that come with them to prop up their ponzi-scheme economies. Will they in the future, probably? They are basically just a heavily centralized/regulated capitalist enterprise now, so who knows. But right now? Do they engage in propaganda? Cyber-espionage? Yes and Yes. So do we, so do you, so does everyone who has a seat at the geopolitical table and the economy to afford it.
The point of all of this isn’t US GOOD CHINA BAD or US BAD CHINA GOOD. The article is about the deepseek models tearing out the floor of US dominance in AI. Personally, having deployed it and played with it, yeah. None of these products are truly useful to me yet, and I remain skeptical of their eventual value, but right now, party censorship or not, you can download a version of an LLM that you can run, retrain and bias however you want, and it costs you the bandwidth it took to download. And it performs on par with US commercial offerings that require pricey subscriptions. Offerings that apparently require huge public investment to keep afloat.
Where I disagree with you is not that the US is bad - the US is terrible, and there is plenty of evidence of that. I don’t even disagree with there being censorship in the US. In fact, Trump is objectively a piece of shit who wants nothing more than to become Xi/Putin himself.
What I disagree with is equating censorship in the US with Chinese censorship. I can call Trump a piece of shit online without worrying that the FBI will show up at my door. The models that are trained in the west will happily entertain any (non-violent) political discussions I want. There may be bias, and Trump may be trying to create censorship, but it’s not quite to that level yet.
I am concerned that the US will become as bad as China in terms of censorship, which is part of why I’m trying to leave right now. However, it’s not there yet. They are not yet equal, nor are they even close yet.
Yeah ok, I do basically agree with you. It’s not an accurate equivalency, yet. We’re trending bad though. I’d say the example of Stephen Miller sort of accidentally hinting that they shut down USAID because they all donated to the Harris campaign had some chilling implications for example. He could just be assuming that, since that’s a safe assumption for populous urban areas generally, but they could also have cross checked lists of employees against political contributions.
You say Chinese state censorship is an understood quantity. Could be. But I’d say that my points about equivalencies are to illustrate that what we think is true, is often much more grey. I’ve been to China, and while I was impressed and shocked at how much more advanced it was than I expected, I also couldn’t imaging living there. It doesn’t change the fact that a stagnant late-stage capital mafia state that lives off defense contracting is performing ooorly against a centrally controlled capitalist state that has set different priorities (that’s right boy, deepseek-r1 is a side project of a…. CHINESE HEDGE FUND). It’s value neutral. But if you dismiss reality based on a conception of political censorship that I doubt you’ve deeply engaged with, enjoy.
The so called free market certainly didn’t seem to take much reassurance in deepseek being compromised by communist censorship this morning though. Probably because the deepseek news isn’t exceptional because of China, or what it is, but because of what it isn’t, compared to the bloated tech carcasses that the US has pinned its hopes on.
In this case, DeepSeek is announcing the training time for their LLMl, which wall street is extrapolating costs from. No state aid involved.
The article mentions the cost of tokens to end users, not training time.
Ah ok, I didn’t catch that. Other articles were discussing v3’s training using only 2.8M GPU hours.
https://www.ft.com/content/c82933fe-be28-463b-8336-d71a2ff5bbbf
Until this has been independently verified, I have my doubts. This wouldn’t be the first time for China to vastly exaggerate its technological capabilities.
Deepseek seems to have done a clever thing w.r.t. training data, by having the model train on data that was emitted by other LLMs (as far as I’ve heard). That means there is sort of “quality-pass”, filtering out a lot of the definitely bogus data. That probably leads to a smaller model, and thus less training hours.
Google engineers put out a paper on this technique recently as well.