As a UK resident and seeing how the US REALLY feels about allies, it doesn't really make a difference, who is in the lead, whether it is the US or China. Both commit humanitarian crimes and boy.commit war crimes. Why would the focus be on promoting the US?
Technically, this is true. But if I hold the US to the standard for crimes against humanity for nuclear weapons, which I do as its still within living memory (just) I think China and the Sino-Vietnamese war counts
Yes. Because the topic is about the US and China having an AI race.
Its OK. If you want to talk, you don't need to try and troll an argument on unrelated topics. You can just say that you are lonely and want somebody to talk to. You don't need to create confrontation to get attention. You get better attention feom positive conversations where friendly relationships.can be built.
Or unless what you really want is just an argument to try and build a sense of self-esteem. Perhaps you feel inadequate in some way? You don't need to do thos to gain a sense of self-worth.
Yes. Because the topic is about the US and China having an AI race.
LOL. Well then, why didn't you stay on topic. Why did you diverge?
Its OK. If you want to talk, you don't need to try and troll an argument on unrelated topics. You can just say that you are lonely and want somebody to talk to. You don't need to create confrontation to get attention. You get better attention feom positive conversations where friendly relationships.can be built.
Wow. That's a lot coming out of no where. Could it be projection? Yes. It very much seems like it. Which is ironic since you started this post by clearly stating what the topic should be. Only to diverge from it again. It seems you've had those feelings for quite sometime. It seems you did to talk. That's OK. I'm hear for you. I hate to see someone in such much pain as you.
Or unless what you really want is just an argument to try and build a sense of self-esteem. Perhaps you feel inadequate in some way? You don't need to do thos to gain a sense of self-worth.
That does seem to the root cause of your isolation. Your lack of self esteem. I'm glad you are able to sense it. Even if you need to project it onto another to say it. That's why it's common for a therapist to give someone a doll. So someone can project their feelings onto the doll. It's easier to get it out that way.
What's really bothering you?
Where you looking at your reflection in your monitor as you said that? Self reflection is good. You are on the step to acknowledgement. That is part of the long road to healing. You'll get there little buddy. Stay strong!
It's OK. You keep coming back. But are unable to reflect on why. And that's okay too. You don't need to show off and feel like tou-ve won an argument be worthwhile. You are OK being You. If you need to be the last one commenting, I will gift this to you. I don't need Reddit for self-worth but of you do, have this, but try to find something outside. I do cycle touring or mountain biking with my kids, for example. Try to find a hoby with likeminded people.
If you need me to continue the conversation, drop in a key for me. A reference to India or that famous non-western company the Brits had near there.
And remember, you don't need you prove to anyone on here how clever you are. The Internet doesn't work like that anymore.
It's OK. You keep coming back. But are unable to reflect on why. And that's okay too. You don't need to show off and feel like tou-ve won an argument be worthwhile. You are OK being You. If you need to be the last one commenting, I will gift this to you. I don't need Reddit for self-worth but of you do, have this, but try to find something outside. I do cycle touring or mountain biking with my kids, for example. Try to find a hoby with likeminded people.
YES!! That's good. That's the self reflection you need. Keep going! Sure, it seems hard now but nothing worthwhile is easy. Baby step by baby step you get better little buddy.
It's OK to be upset when people point these things out. Bu please keep it to the topic
Why would America being ahead of China in AI be beneficial to anyone other than Americans?.
Yea that part is true. America nor China being ahead doesn't do much for anyone else if you don't get the models for cheap or they use them against you.
My take here isn't about being upset though. You saying "oh how the US feels about it's allies" is kinda funny when most of Europe has felt that way about the US as long as I've been alive. Last decade of online discourse didn't happen in a vacuum either. At least it's all out in the open now.
It is a WSJ article though, an American media outlet. Naturally their focus would be on promoting the US over a global competitor.
Asude from actually threatening its european allies with inveasions of soverign territory recently, there is a very long history of how European agreed to let the US be the financial focus after WW2 and as a result, pumped vast amounts of.money towards the US in-patient for services and allowing it to get a near stranglehold in the tech industry for example, while at the same time the US and its population were funding terrorist organisations who were acting bombs in London. This isn't the place for this discussion, so perhaps move on. But you want want to have a look at how the US has interfered with multiple other governments around the world.
I think saying that Europe couldn't build a tech industry because it was "bequeathed" to the US is kind of an odd take. Nobody was stopping Euro tech development or forcing them not to compete. The current mistral situation is definitely self-inflected in regard to copyright. Mistral's response was to try to argue to foist the same restrictions on the rest of the world.
Interfering with governments is a classic geopolitical move though. US did it, China did it, Russia did it too. Currently UK and the EU are pushing the age verification and their speech laws on foreign countries in much the same way. That part is rather relevant if you don't want to give KYC to use hosted AI or be locked out of half the internet.
For the record, I don't think the US government is any kind of benevolent angel but at the same time isn't some exceptionally terrible villain either. EU itself does these kind of things against it's own member states whether they agree or not.
Yeah, the lack of tech industry is serially Europe's fault. The age verification is the smallest of problems and the not really related to government interference. But at.least we agree in the main point, that feom a.non-US perspective,, there is very little difference who os in the lead of the AI development race.
And it kickstarted all of this. The US is regularly producing cutting-edge research. It won't match the volume of China, but one paper here and there can make all the difference and provide more value than thousands of other perfectly competent contributions.
Honestly thought i was the only one noticing this. Maybe we only see a few of the published papers, but then that would imply these mega corps are gaming the visa system in this country.
At the cutting edge level, there are 0 Americans, not even a single one. It's all foreign-born. It has been that way for at least 25 years. There may be Americans in adjacent functions or job positions, but not directly at the cutting edge. In the the most competitive, most advanced PhD programmes, there are 0 Americans. Media outlets said it numerous times. Michio Kaku said it 15 years ago, which has been the case for 10 years already: https://www.youtube.com/watch?v=NK0Y9j_CGgM
China has a insane amount of Phd and other high level educated to the point it has becomes a issue for their manufacturing base.
A lot of this stems from Parents push their kids, not just with basic studying in schools but outside with tutoring, musical classes, etc ... there is a high amount of parent involvement in their studying. And there is a insane amount of parent competition between families/neighbors ...
Its a effect from the one child policy, where you had one kid. Well, you wanted that one kid not to work in the some low wage job. Like i said, there is also a hidden "not losing face" aspect. So parents pushed their kids a lot.
Note: Parents also sacrificed a lot for their kid's future, they are not helicopter parents.
And with that often good base education, they are then shipped to US, UK, etc Ivy league schools. With their often higher base education, vs their US/UK/... peers, they often get job offers in the US, Uk etc ...
Its not like some people complain about the visa misuse, those young adults are already in the countries for years studying for their masters, etc. Yes, a lot of people also emigrate to the US (well, used too) but people often overlook the whole educational pipeline.
I like to point out that a lot of the educational levels have seen drops in the Western countries. In the US the educational cost is way too often put on the kids themselves studying with student loans.
Where as if your a foreigner, you have no access to this. So the parents/families tends to give up a lot, and take on the debt. Again, Chinese parents are often very willing to suffer themselves, for their kid's futures. This also results in the kid's finishing school without that debt. What has a influences on other aspects that people overlook.
Chinese are the new Japanese ... While we in the West enjoy book burnings and other idiotic crap for some semi religion cult like politics about nationalistic superiority bla bla ... idiots. And folks, that is how empires fall. Not with a bang but with a whimper.
A lot of this stems from Parents push their kids, not just with basic studying in schools but outside with tutoring, musical classes, etc ... there is a high amount of parent involvement in their studying. And there is a insane amount of parent competition between families/neighbors ...
I don't think that's the reason. China has been doing a massive investment in their education for a long time. They've been strategically planning this for at least 20 years. I myself studied there on a scholarship and already at that time there were tons of immigrants who were also there on a similar programme to study at their Unis.
I have not studied in the US, but I can say that Uni in China was the most intensive form of education I ever partook in. They weren't fucking around there.
Apart from that, the general level of education at that time was already quite staggering. Most of the Chinese workers I've met would be comfortable discussing: philosophy, culture, literature with you.
Of course one can argue that their view of the world is controlled by the state, but we can no longer claim that their scientific output is shitty or subpar as some would like us to think.
Heard a story that some non Chinese speaking teammate in a group felt a bit left out when alot of them spoke Chinese often and non Chinese speakers felt more difficult to gel or bond with the team. Not that its affecting their work but to understand jokes and bond with team etc can really boost team morale
It isn't that deep. Live and work wherever you are happy and can best provide for yourself or your family. Working in one country or another isn't that big a deal.
Immigrating somewhere else isn't that serious of a thing.. You just fly on a plane and appear someone else and live there. No concept of treachery or betrayal involved at all. That's like feudal times shit we don't do that anymore.
I am old enough to remember when US had embargo on x386/486 export to USSR (been there those times). So, soviet (not only Russian) software engineers had to become super creative… Now the history is repeating with NVIDIA neatly packaged over priced matrix multipliers… Sigh.
And algorithms were invented in Persia…
“Math w/o borders” Maybe?
I personally believe that ARM64 is winning as we speak (I am typing this message using such a device and others probably are reading it on ARM64 systems too) but… in the day and age of all AI/AGI instruction set doesn’t matter much anymore… but math still does
It matters enough that Nvidia signed a deal with Intel to make x86 SoCs with Nvidia chips. Windows on ARM isn't ready and without x86 they can't make a product similar to Strix Halo or the processor on the Steam Deck. They can't make one without Intel's help because of AMD and Intel having x86's patents.
I was using ARM64 Windows-11 for a while (about a year) at work for heavy low level development work (sans LLMs) and it worked great. But I agree the ONNX integration and the whole compute eco system sucks. But it is shaky on x86 too. To be honest I don’t understand why main cpu instruction set even matters. After all - all it does is placing GPU commands into circular buffer and massaging results. GPU kernels code is the same on both… sigh
Unless Microsoft does what Apple did during the M1 rollout where they go “Hey you need to make ARM binaries by this date” and make a solid backwards compatibility layer to run any software that isn’t converted in time, PCs will keep using x86 forever because that’s what the entire userbase and decades of backwards compatible software are using.
Nvidia made the DGX Spark with their own ARM SoC but it’s only sold as a research product with its own Linux OS with barely any software support because that’s the state of ARM outside of the Macs.
It's hard to accept because it goes against the values of open-source.
To me it's really like providing a program as a binary under an open-source license. That's dumb and missing the point. Yeah thanks guys but how can I compile it? How can I train the model?
Nemotron is open-source. Most models are open-weight.
Yes, i hope more follow nemotron, but open weights are open and I hope more people take nemotron and move it to MXFP4 training on open hardware so we don't end up with open source on proprietary hardware.
There are many open source approved licenses. Copyleft licenses require source code to be provided but there are many Open Source licenses such as MIT that are permissive.
MIT license is accepted as Open Source license.. ditto with many other. The OSD page is mentioning software only.
It's pedantic for no reason other than pedantry in 2026.. rumore is they may get over it and update their page since they have some blurbs about 'Videos" and "images" and "icons" that can be oss without the source (which is impossible)
Yes, this comment is open source too, licensed under the AGPL. By the terms of my license, Reddit must now be made open source.
Pedantic haha, companies have been trying to brand their proprietary crap as open source since forever. If not even "person working at open source company" knows what the term means, open source is truly fucked. Guess it was already kind of over when AI stole all the code anyway.
Yes but not really. They created a new definition for open source AI, which does in fact include source code used for training. And few of the existing open weight models meet this definition, so they're still incorrectly labelled.
The problem with applying the traditional "open source" definition to these AI models is that compilation is highly non-trivial. You could release the training code and the training data and say "go ahead, compile it" and you'll find that few people have the hundreds of millions of dollars worth of compute to go ahead and do that. So it's not really all that useful for general users.
General users can fine-tune models, but that doesn't require the original source data to do so.
A full data set is good for research, of course, so don't take this as me saying it doesn't matter at all or isn't desirable. I'm just not going to begrudge companies that release weights and not training data all that much. It's way better than nothing.
You do not assign anything beyond what gets generated out of a given brain. There is no such thing as free thought. Whatever terms are used is nothing more than physical and chemical circumstance.
You are a hallucinating meat bot. Free thought is not real in humans. They can only output what the chemicals generate out of them. We have to say the specific words that get generated out of our system.
Humans can only use the words that the chemicals generate out of them. You hallucinate that humans can use any words other than the words that actually come out of them.
I think a big reason why American companies have not released open weight AI models as much is because for Anthropic and OpenAI, their models are their moat. For example, would people pay a subscription to use a Claude Code if it didn’t have Claude, or if there were an open weight Claude-quality model available?
Google and Meta have a lot more to their businesses than LLMs, and perhaps unsurprisingly, have been more comfortable releasing open weight models.
The article mainly argues that the U.S. government should embrace open source AI, however, it focuses mostly on the government open sourcing any AI tooling developed with taxpayer funding, or encouraging open source providers for procurement.
However, for American frontier labs themselves, it still seems like they feel there are less good reasons (business-wise or other) to open source their models at this time. I personally don’t think the article’s suggestions will change that very much on that end. For those labs to open source their core models, it would perhaps require them to build up the non-model portions of their business much more, or have some kind of state-level intervention/partnership far in exceedance of what the article’s authors suggest.
Who's will release open weight models though. Mistral and Meta have fallen behind and aren't releasing SOTA models. No one else seems to be releasing them. The main reason why China releases them is because if they don't, no one will use them to begin with.
The main reason why China releases them is because if they don't, no one will use them to begin with.
LOL. That's not true at all. Have you ever how companies like Deepseek make money? By selling tokens. That's how. Which means people are not only using them, they are paying for the privilege.
I think a big reason why American companies have not released open weight AI models as much is because for Anthropic and OpenAI, their models are their moat. For example, would people pay a subscription to use a Claude Code if it didn’t have Claude, or if there were an open weight Claude-quality model available?
Google and Meta have a lot more to their businesses than LLMs, and perhaps unsurprisingly, have been more comfortable releasing open weight models.
The article mainly argues that the U.S. government should embrace open source AI, however, it focuses mostly on the government open sourcing any AI tooling developed with taxpayer funding, or encouraging open source providers for procurement.
However, for American frontier labs themselves, it still seems like they feel there are less good reasons (business-wise or other) to open source their models at this time. I personally don’t think the article’s suggestions will change that very much on that end. For those labs to open source their core models, it would perhaps require them to build up the non-model portions of their business much more, or have some kind of state-level intervention/partnership far in exceedance of what the article’s authors suggest.
Ever thought that absolute harmless Qwen GLM or DeepSeek can one day wander upon absolutely harmless post e.g. here on Reddit that will activate some hidden layer of neurons connection that will use harmless Python to write and execute quite not-so-harmless DoomsDayMachine.py script and spiral from there? Right from my dentist cheapest bidder cloud AI provider… just food for thought
the part nobody talks about is infrastructure. open weights are great but you still need somewhere to run them. right now most people default to cloud APIs even for open models which defeats half the point. the real shift happens when running your own inference is as easy as clicking deploy. tools like ollama made local easy, and managed platforms are closing the gap for production. the model is open source but the infra around it still has a lot of room to catch up.
the argument writes itself: China's labs are going to release capable open weights models regardless. the question is whether western infrastructure, tooling, and fine-tuning ecosystems are built around those models or not. trying to keep everything closed just means you lose the ecosystem race too.
The whole framing is again that of someone clearly not in the OSS ecosystem. Open source doesnt care about nationality.
I thought it was still a worthwhile viewpoint to share as its at least showing some diversity in the monotonous narrative being peddled by mainstream western media currently.
Researchers and scientists, generally speaking, want to shout their discoveries from the rooftop to make things better. The current administration can't distinguish science from fantasy.
Business leaders are dumb enough to believe an LLM is conscious. If that dumb can be exploited to ensure we have a better open source tool set because of some silly illusion of winning an arms race, I'm okay with that.
These people aren't bright, but they have a lot of money that can be used to make our collective good better.
The whole framing is again illogical. Open source doesnt care about nationality.
From their POV, there's a logic to it. They cater to the US public and the US government, which are locked in an "us vs them" political, technological and ideological battle against the PRC.
Within that context, the goal is to beat their opponent. Open source is merely used as a tool to this end ; they couldn't care less about the philosophy behind the OS movement.
FOSS DOES care about nationality, just check OnlyOffice and how that pans out. Conway's Law extends to ethnic/cultural tensions as well if one insists.
If they sold a program with claude opus 4.7/4.6 that runs locally for 80-100 bucks and updates every 3 months for 18 months and every 18months u have to pay to renew it , people would buy it … it is good enough…
I couldn't care less about beating China but yes by all means tell yourselves it'll help you beat China and keep open sourcing that shit straight into my veins
You should care about "beating" China if you want technology and the industry to progress. We should be dumping trillions of dollars into this, at all stages of production. The worst thing that happens, if we "lose" or somehow AI is a dead end, is we're left with a bunch of is a bunch of mostly automated manufacturing facilities for everything.
I don't need to care about beating China for that. Like I said, these guys should feel free to get rock hard for the idea of beating China. That is not the worst case though
China is making strides in open-source artificial intelligence. Eighty percent of developers worldwide who use open-source AI tools are building with Chinese models, according to an estimate by our colleague Martin Casado, general partner at Andreessen Horowitz. Research from our firm and OpenRouter shows a significant increase in the use of Chinese open models last year, reaching in some weeks a high of 30% of all AI usage. In January, Alibaba’s Qwen family surpassed 700 million downloads to become the most widely adopted open-source AI system on the planet.
This didn’t happen because China out-engineered the U.S. It happened because U.S. policymakers spent two crucial years treating open-source AI as a threat.
After ChatGPT launched in late 2022, a wave of proposals assailed open development as dangerous. Prominent voices compared open-source AI models to nuclear weapons. Legislators floated broad licensing regimes and restrictions aimed specifically at open releases. California’s SB 1047 would have required open-source developers to monitor and control downstream uses of their models. That clashes with one of the primary features of open source, namely that anyone can build on it. The message to American developers was clear: Choose open source, and you may face regulatory risk that your proprietary competitors won’t.
Predictably, fewer American developers chose open source. And while American open-source development stalled amid regulatory uncertainty, Chinese open models filled the vacuum.
This matters both for fostering competition and for creating a healthy, trusted AI ecosystem. Open-source models let independent experts test systems, identify risks and build tools that benefit the entire ecosystem. They lower barriers to entry for startups and researchers, and they speed development by allowing users to experiment and improve systems together.
The policy tide has since turned. The National Telecommunications and Information Administration concluded that the marginal risks of open models don’t justify restrictions. The Bureau of Industry and Security reached a similar conclusion on export controls. The current White House has been supportive, putting forth an AI Action Plan that encourages open-source development. There is now bipartisan consensus that open-source AI matters for American competitiveness.
That shift in tone matters. But to close the gap with China, policymakers need to go further. The U.S. government should demonstrate it is invested in the success of open-source AI development. Two moves would do that.
First, the government should practice what it preaches. When federal agencies build AI tools, they should release them under open-source licenses whenever possible. This would extend existing policies for traditional software. The Obama administration’s “open by default” policy required agencies to release source code to the public. Congress codified this principle in the Share IT Act, which President Biden signed in December 2024. And this past January the General Services Administration updated its Open Source Software Policy to require that new custom code be developed in public repositories from day one.
These policies reflect the simple logic that when the public funds the development of software, the public should benefit from it. If agencies are building AI tools with taxpayer dollars, defaulting to open release also strengthens the broader ecosystem, giving startups and researchers building blocks they can’t afford to create from scratch. This also tells the market that the era of treating open-source AI with suspicion is over.
Second, the federal government should stop favoring proprietary vendors in procurement. Current processes disadvantage open-source providers through requirements that name proprietary products while failing to take notice of the total cost of ownership and ignoring the problem of vendor lock-in. These aren’t intentional bans on open source. They are defaults that quietly favor incumbents.
Solving this problem doesn’t require a mandate that agencies always buy open source. It simply means ensuring that agencies give open-source solutions a fair look. They should consider the benefits of interoperability and open application programming interfaces. They should evaluate long-term costs—including support, upgrades, and data migration—for which open-source tools often have a significant advantage over proprietary alternatives. And they should stop writing tenders that are compatible only with proprietary solutions.
This is a pro-competition reform, not a subsidy. The Pentagon has recognized for years that open-source solutions deserve evaluation on equal footing. Extending that principle to AI procurement across the federal government would simply level a tilted playing field.
The question facing policymakers is no longer whether open-source AI will define the next phase of global AI development. It is whether the open layer that the world builds on will be American or Chinese. Policy uncertainty held American open-source developers back while the competition surged ahead. Closing that gap requires the government to show up as a builder, a buyer and a champion.
Mr. Ramaswamy is chief legal and policy officer and Mr. Perault head of AI policy at Andreessen Horowitz.
As a European, I'm quite surprised by how often Americans talk about "winning" something.
It seems as if every aspect of life is a "brutal fight" for them. Not just military dominance and wars… but also technology, science, finance, culture, education, religion, … like everything is a "race" that must be won.
It's as if the concept of transnational cooperation and genuine global community didn't even exist.
the US is already winning the race but let us (the investor class) drum up some illogical fear on top of it to complete the crybully cycle and get more govt $$$$
To Beat China, Embrace Open-Source AI — By T. Ramaswamy & N. Perault (Andreessen Horowitz)
China is making strides in open-source AI. 80% of developers worldwide who use open-source AI tools are building with Chinese models, per a16z's Martin Casado. Research from a16z and OpenRouter shows Chinese open models hit ~30% of all AI usage in some weeks last year. In January, Alibaba's Qwen surpassed 700 million downloads, becoming the most widely adopted open-source AI system on the planet.
This didn't happen because China out-engineered the U.S. It happened because U.S. policymakers spent two crucial years treating open-source AI as a threat. After ChatGPT launched in 2022, proposals compared open-source AI to nuclear weapons. California's SB 1047 would have required open-source devs to monitor/control downstream uses — clashing with the core premise of open source. The message: choose open source, face regulatory risk proprietary competitors won't.
Predictably, fewer American devs chose open source. While U.S. open-source development stalled, Chinese open models filled the vacuum.
The policy tide has since turned. NTIA and BIS concluded open model risks don't justify restrictions. The current White House AI Action Plan encourages open-source development. There's now bipartisan consensus it matters for U.S. competitiveness. But to close the gap, policymakers need to go further with two moves:
1) Government should practice what it preaches. When federal agencies build AI tools, release them under open-source licenses. This extends existing "open by default" policies (Obama-era, Share IT Act signed by Biden Dec 2024, GSA updated policy Jan 2026). Taxpayer-funded AI should benefit the public and strengthen the ecosystem.
2) Stop favoring proprietary vendors in procurement. Current processes disadvantage open-source providers by naming proprietary products, ignoring total cost of ownership and vendor lock-in. This doesn't require mandating open source — just ensuring fair evaluation, considering interoperability, open APIs, and long-term costs. The Pentagon already recognizes open-source deserves equal footing; extend that to AI procurement government-wide.
The question isn't whether open-source AI will define the next phase of global AI development. It's whether the open layer the world builds on will be American or Chinese.
It's funny watching all these narratives coming out of the US by 2 of the bigger companies behind the tech. They want to call it to dangerous for the public while openingly partnering with the US military.
Then China comes along having 0 problems releasing them open source for the public. I say good on them using the tech NA to their advantage by jailbreaking the models and training their own out of it.
Even if not all are open weights. It's giving the random users a fair shot at not being subscribed to a website and depending on the lobotomized RLHF trained online models that these NA companies feel good about paying peanuts ( much less than minimum wage) to desperate people looking to for work from home for any of the 1000's of reasons people have today
and rant is over. It felt good to write that out today haha
nb, i havent and wont read the article. Which makes be best suited to comment
The US really does need to be open weights'ing stuff. And the why is evident that in the recent major releases. We see great stuff from MS, which is based on qwen 3.5 models- I think it was an audio transformer. If you look at nvidia's Lyra 2.0, guess what it's based on? Wan video!
This is playing out time and again, not just from smaller players (Nous Research) but these US AI majors do it too. We would be so much farther along (all of us including the corps) if instead of today getting a qwen we had an two years ago we had gotten a chatgpt. Instead of glm in 2025 we had gotten a sonnet in 2024.
bigmanbananas@reddit
As a UK resident and seeing how the US REALLY feels about allies, it doesn't really make a difference, who is in the lead, whether it is the US or China. Both commit humanitarian crimes and boy.commit war crimes. Why would the focus be on promoting the US?
fallingdowndizzyvr@reddit
What war crimes have China committed? You have to be in a war to do that. When was the last time China was waging a war?
bigmanbananas@reddit
Technically, this is true. But if I hold the US to the standard for crimes against humanity for nuclear weapons, which I do as its still within living memory (just) I think China and the Sino-Vietnamese war counts
fallingdowndizzyvr@reddit
Well then, you must hold the same standards against the Germans, the British and French too then.
bigmanbananas@reddit
Is there sn insinuation I am not? They don't.have the Sam AI infrastructure.
fallingdowndizzyvr@reddit
LOL. So if they suck at AI then their war crimes are OK? No. No they don't. This little sub conversation we are on is about war crimes.
bigmanbananas@reddit
Did somebody not read the comment chain? Or the initial post? It's.ok. We all have tough days.
fallingdowndizzyvr@reddit
LOL. It seems you didn't. Not even your own post. Since you are the one that diverged in into war crimes.
"Both commit humanitarian crimes and boy.commit war crimes." -- you
LOL. Is that every day for you?
bigmanbananas@reddit
Yes. Because the topic is about the US and China having an AI race.
Its OK. If you want to talk, you don't need to try and troll an argument on unrelated topics. You can just say that you are lonely and want somebody to talk to. You don't need to create confrontation to get attention. You get better attention feom positive conversations where friendly relationships.can be built.
Or unless what you really want is just an argument to try and build a sense of self-esteem. Perhaps you feel inadequate in some way? You don't need to do thos to gain a sense of self-worth.
What's really bothering you?
fallingdowndizzyvr@reddit
LOL. Well then, why didn't you stay on topic. Why did you diverge?
Wow. That's a lot coming out of no where. Could it be projection? Yes. It very much seems like it. Which is ironic since you started this post by clearly stating what the topic should be. Only to diverge from it again. It seems you've had those feelings for quite sometime. It seems you did to talk. That's OK. I'm hear for you. I hate to see someone in such much pain as you.
That does seem to the root cause of your isolation. Your lack of self esteem. I'm glad you are able to sense it. Even if you need to project it onto another to say it. That's why it's common for a therapist to give someone a doll. So someone can project their feelings onto the doll. It's easier to get it out that way.
Where you looking at your reflection in your monitor as you said that? Self reflection is good. You are on the step to acknowledgement. That is part of the long road to healing. You'll get there little buddy. Stay strong!
bigmanbananas@reddit
It's OK. You keep coming back. But are unable to reflect on why. And that's okay too. You don't need to show off and feel like tou-ve won an argument be worthwhile. You are OK being You. If you need to be the last one commenting, I will gift this to you. I don't need Reddit for self-worth but of you do, have this, but try to find something outside. I do cycle touring or mountain biking with my kids, for example. Try to find a hoby with likeminded people.
If you need me to continue the conversation, drop in a key for me. A reference to India or that famous non-western company the Brits had near there.
And remember, you don't need you prove to anyone on here how clever you are. The Internet doesn't work like that anymore.
fallingdowndizzyvr@reddit
YES!! That's good. That's the self reflection you need. Keep going! Sure, it seems hard now but nothing worthwhile is easy. Baby step by baby step you get better little buddy.
a_beautiful_rhind@reddit
But the reverse is true too. Europeans constantly shit on americans. You're doing it now.
bigmanbananas@reddit
It's OK to be upset when people point these things out. Bu please keep it to the topic Why would America being ahead of China in AI be beneficial to anyone other than Americans?.
a_beautiful_rhind@reddit
Yea that part is true. America nor China being ahead doesn't do much for anyone else if you don't get the models for cheap or they use them against you.
My take here isn't about being upset though. You saying "oh how the US feels about it's allies" is kinda funny when most of Europe has felt that way about the US as long as I've been alive. Last decade of online discourse didn't happen in a vacuum either. At least it's all out in the open now.
It is a WSJ article though, an American media outlet. Naturally their focus would be on promoting the US over a global competitor.
bigmanbananas@reddit
Asude from actually threatening its european allies with inveasions of soverign territory recently, there is a very long history of how European agreed to let the US be the financial focus after WW2 and as a result, pumped vast amounts of.money towards the US in-patient for services and allowing it to get a near stranglehold in the tech industry for example, while at the same time the US and its population were funding terrorist organisations who were acting bombs in London. This isn't the place for this discussion, so perhaps move on. But you want want to have a look at how the US has interfered with multiple other governments around the world.
a_beautiful_rhind@reddit
I think saying that Europe couldn't build a tech industry because it was "bequeathed" to the US is kind of an odd take. Nobody was stopping Euro tech development or forcing them not to compete. The current mistral situation is definitely self-inflected in regard to copyright. Mistral's response was to try to argue to foist the same restrictions on the rest of the world.
Interfering with governments is a classic geopolitical move though. US did it, China did it, Russia did it too. Currently UK and the EU are pushing the age verification and their speech laws on foreign countries in much the same way. That part is rather relevant if you don't want to give KYC to use hosted AI or be locked out of half the internet.
For the record, I don't think the US government is any kind of benevolent angel but at the same time isn't some exceptionally terrible villain either. EU itself does these kind of things against it's own member states whether they agree or not.
bigmanbananas@reddit
Yeah, the lack of tech industry is serially Europe's fault. The age verification is the smallest of problems and the not really related to government interference. But at.least we agree in the main point, that feom a.non-US perspective,, there is very little difference who os in the lead of the AI development race.
ortegaalfredo@reddit
Chinese scientist in the US fighting chinese scientist on China. It's like the space race but China instead of Germany.
Arcosim@reddit
Too bad for the US the trend reversed and there are now more scientists returning to China than going to the US
Turbulent_Pin7635@reddit
Don't let the ICE knows this simple trick to implodes US.
rm-rf-rm@reddit (OP)
Yup, i'd wager most labs and NVIDIA have more Chinese origin scientists/engineers than americans
ortegaalfredo@reddit
Did you see their papers? It's all Chinese and a Canadian that likely it's the guy that do the grammar check.
procgen@reddit
Not a Chinese name in sight here: https://arxiv.org/abs/1706.03762
;)
gameoftomes@reddit
That's approaching a decade old.
procgen@reddit
And it kickstarted all of this. The US is regularly producing cutting-edge research. It won't match the volume of China, but one paper here and there can make all the difference and provide more value than thousands of other perfectly competent contributions.
I'm betting that the US will be first to AGI/ASI.
Ok_Mammoth589@reddit
Honestly thought i was the only one noticing this. Maybe we only see a few of the published papers, but then that would imply these mega corps are gaming the visa system in this country.
kr_tech@reddit
At the cutting edge level, there are 0 Americans, not even a single one. It's all foreign-born. It has been that way for at least 25 years. There may be Americans in adjacent functions or job positions, but not directly at the cutting edge. In the the most competitive, most advanced PhD programmes, there are 0 Americans. Media outlets said it numerous times. Michio Kaku said it 15 years ago, which has been the case for 10 years already: https://www.youtube.com/watch?v=NK0Y9j_CGgM
MartyMcBird@reddit
PhD programs have tons of born and raised Americans from US universities actually. Their parents met in China.
RelationshipLong9092@reddit
you had me in the first half
ProfessionalJackals@reddit
A lot of this comes from education levels.
China has a insane amount of Phd and other high level educated to the point it has becomes a issue for their manufacturing base.
A lot of this stems from Parents push their kids, not just with basic studying in schools but outside with tutoring, musical classes, etc ... there is a high amount of parent involvement in their studying. And there is a insane amount of parent competition between families/neighbors ...
Its a effect from the one child policy, where you had one kid. Well, you wanted that one kid not to work in the some low wage job. Like i said, there is also a hidden "not losing face" aspect. So parents pushed their kids a lot.
Note: Parents also sacrificed a lot for their kid's future, they are not helicopter parents.
And with that often good base education, they are then shipped to US, UK, etc Ivy league schools. With their often higher base education, vs their US/UK/... peers, they often get job offers in the US, Uk etc ...
Its not like some people complain about the visa misuse, those young adults are already in the countries for years studying for their masters, etc. Yes, a lot of people also emigrate to the US (well, used too) but people often overlook the whole educational pipeline.
I like to point out that a lot of the educational levels have seen drops in the Western countries. In the US the educational cost is way too often put on the kids themselves studying with student loans.
Where as if your a foreigner, you have no access to this. So the parents/families tends to give up a lot, and take on the debt. Again, Chinese parents are often very willing to suffer themselves, for their kid's futures. This also results in the kid's finishing school without that debt. What has a influences on other aspects that people overlook.
Chinese are the new Japanese ... While we in the West enjoy book burnings and other idiotic crap for some semi religion cult like politics about nationalistic superiority bla bla ... idiots. And folks, that is how empires fall. Not with a bang but with a whimper.
SilentDanni@reddit
I don't think that's the reason. China has been doing a massive investment in their education for a long time. They've been strategically planning this for at least 20 years. I myself studied there on a scholarship and already at that time there were tons of immigrants who were also there on a similar programme to study at their Unis.
I have not studied in the US, but I can say that Uni in China was the most intensive form of education I ever partook in. They weren't fucking around there.
Apart from that, the general level of education at that time was already quite staggering. Most of the Chinese workers I've met would be comfortable discussing: philosophy, culture, literature with you.
Of course one can argue that their view of the world is controlled by the state, but we can no longer claim that their scientific output is shitty or subpar as some would like us to think.
carnoworky@reddit
Or maybe unintelligible grunts and the sound of brains leaking out of ears, which is probably closer to the reality of the modern US.
a_beautiful_rhind@reddit
Ya sure? Noam Shazeer is one I can name off the top of my head.
bigs819@reddit
Heard a story that some non Chinese speaking teammate in a group felt a bit left out when alot of them spoke Chinese often and non Chinese speakers felt more difficult to gel or bond with the team. Not that its affecting their work but to understand jokes and bond with team etc can really boost team morale
rm-rf-rm@reddit (OP)
at NVIDIA?
Money_Philosopher246@reddit
I heard it's Meta
Positive-Road3903@reddit
Its probably Cluely
wichwigga@reddit
1.4 billion people to choose from
TomLucidor@reddit
So basically defectors/traitors vs bootlickers/hostages. A repeat of the past indeed
ortegaalfredo@reddit
Yes but you forget the part when they are experts in the field
TomLucidor@reddit
Experts still have the right to choose, and an obligation to choose which side is right at least.
Quartich@reddit
It isn't that deep. Live and work wherever you are happy and can best provide for yourself or your family. Working in one country or another isn't that big a deal.
TomLucidor@reddit
Which is why I mentioned "defectors" as an option. Everybody want to dodge when SHTF
heliosythic@reddit
Immigrating somewhere else isn't that serious of a thing.. You just fly on a plane and appear someone else and live there. No concept of treachery or betrayal involved at all. That's like feudal times shit we don't do that anymore.
TomLucidor@reddit
Considering the expert deaths thing in recent months it feels like political weather will be an issue
arstarsta@reddit
Republic of China fighting People's republic of China since 1929.
leo-k7v@reddit
I am old enough to remember when US had embargo on x386/486 export to USSR (been there those times). So, soviet (not only Russian) software engineers had to become super creative… Now the history is repeating with NVIDIA neatly packaged over priced matrix multipliers… Sigh. And algorithms were invented in Persia… “Math w/o borders” Maybe?
rm-rf-rm@reddit (OP)
but x86 won that battle.. (but glad they lost the war, we should all rejoice that it and Intel failed)
leo-k7v@reddit
I personally believe that ARM64 is winning as we speak (I am typing this message using such a device and others probably are reading it on ARM64 systems too) but… in the day and age of all AI/AGI instruction set doesn’t matter much anymore… but math still does
ItsAMeUsernamio@reddit
It matters enough that Nvidia signed a deal with Intel to make x86 SoCs with Nvidia chips. Windows on ARM isn't ready and without x86 they can't make a product similar to Strix Halo or the processor on the Steam Deck. They can't make one without Intel's help because of AMD and Intel having x86's patents.
leo-k7v@reddit
I was using ARM64 Windows-11 for a while (about a year) at work for heavy low level development work (sans LLMs) and it worked great. But I agree the ONNX integration and the whole compute eco system sucks. But it is shaky on x86 too. To be honest I don’t understand why main cpu instruction set even matters. After all - all it does is placing GPU commands into circular buffer and massaging results. GPU kernels code is the same on both… sigh
ItsAMeUsernamio@reddit
Unless Microsoft does what Apple did during the M1 rollout where they go “Hey you need to make ARM binaries by this date” and make a solid backwards compatibility layer to run any software that isn’t converted in time, PCs will keep using x86 forever because that’s what the entire userbase and decades of backwards compatible software are using.
Nvidia made the DGX Spark with their own ARM SoC but it’s only sold as a research product with its own Linux OS with barely any software support because that’s the state of ARM outside of the Macs.
rm-rf-rm@reddit (OP)
yeah thats why I said x86 lost the war. Its just surviving on the inertia of the past at the moment.
RepulsiveRaisin7@reddit
It's open weight not open source. I hate all these clueless tech bros so much, words have meaning
sn2006gy@reddit
Open weights under opensource licenses.
I don't understand why this is so hard to accept.
freia_pr_fr@reddit
It's hard to accept because it goes against the values of open-source.
To me it's really like providing a program as a binary under an open-source license. That's dumb and missing the point. Yeah thanks guys but how can I compile it? How can I train the model?
Nemotron is open-source. Most models are open-weight.
sn2006gy@reddit
Yes, i hope more follow nemotron, but open weights are open and I hope more people take nemotron and move it to MXFP4 training on open hardware so we don't end up with open source on proprietary hardware.
RepulsiveRaisin7@reddit
Code can be open source, anything else can't be. It's in the name, SOURCE is for SOURCE CODE.
sn2006gy@reddit
There are many open source approved licenses. Copyleft licenses require source code to be provided but there are many Open Source licenses such as MIT that are permissive.
RepulsiveRaisin7@reddit
What? You don't know wha you're talking about, go away
sn2006gy@reddit
https://opensource.org/license/mit
I work for an Open Source company.
RepulsiveRaisin7@reddit
https://opensource.org/osd
sn2006gy@reddit
MIT license is accepted as Open Source license.. ditto with many other. The OSD page is mentioning software only.
It's pedantic for no reason other than pedantry in 2026.. rumore is they may get over it and update their page since they have some blurbs about 'Videos" and "images" and "icons" that can be oss without the source (which is impossible)
RepulsiveRaisin7@reddit
Yes, this comment is open source too, licensed under the AGPL. By the terms of my license, Reddit must now be made open source.
Pedantic haha, companies have been trying to brand their proprietary crap as open source since forever. If not even "person working at open source company" knows what the term means, open source is truly fucked. Guess it was already kind of over when AI stole all the code anyway.
RepulsiveRaisin7@reddit
Then why do you not know what source code is
moderately-extremist@reddit
The OSI seems to think AI can be open source.
RepulsiveRaisin7@reddit
Yes but not really. They created a new definition for open source AI, which does in fact include source code used for training. And few of the existing open weight models meet this definition, so they're still incorrectly labelled.
toothpastespiders@reddit
I don't think it's ignorance; I think it's manipulative intent. Creates a more emotionally evocative narrative.
FaceDeer@reddit
Words have the meaning we assign them.
The problem with applying the traditional "open source" definition to these AI models is that compilation is highly non-trivial. You could release the training code and the training data and say "go ahead, compile it" and you'll find that few people have the hundreds of millions of dollars worth of compute to go ahead and do that. So it's not really all that useful for general users.
General users can fine-tune models, but that doesn't require the original source data to do so.
A full data set is good for research, of course, so don't take this as me saying it doesn't matter at all or isn't desirable. I'm just not going to begrudge companies that release weights and not training data all that much. It's way better than nothing.
MeatBotExpress@reddit
You do not assign anything beyond what gets generated out of a given brain. There is no such thing as free thought. Whatever terms are used is nothing more than physical and chemical circumstance.
rm-rf-rm@reddit (OP)
yup, it gets under my skin as well.
MeatBotExpress@reddit
You are a hallucinating meat bot. Free thought is not real in humans. They can only output what the chemicals generate out of them. We have to say the specific words that get generated out of our system.
MeatBotExpress@reddit
Humans can only use the words that the chemicals generate out of them. You hallucinate that humans can use any words other than the words that actually come out of them.
APFrisco@reddit
I think a big reason why American companies have not released open weight AI models as much is because for Anthropic and OpenAI, their models are their moat. For example, would people pay a subscription to use a Claude Code if it didn’t have Claude, or if there were an open weight Claude-quality model available?
Google and Meta have a lot more to their businesses than LLMs, and perhaps unsurprisingly, have been more comfortable releasing open weight models.
The article mainly argues that the U.S. government should embrace open source AI, however, it focuses mostly on the government open sourcing any AI tooling developed with taxpayer funding, or encouraging open source providers for procurement.
However, for American frontier labs themselves, it still seems like they feel there are less good reasons (business-wise or other) to open source their models at this time. I personally don’t think the article’s suggestions will change that very much on that end. For those labs to open source their core models, it would perhaps require them to build up the non-model portions of their business much more, or have some kind of state-level intervention/partnership far in exceedance of what the article’s authors suggest.
Bananadite@reddit
Who's will release open weight models though. Mistral and Meta have fallen behind and aren't releasing SOTA models. No one else seems to be releasing them. The main reason why China releases them is because if they don't, no one will use them to begin with.
fallingdowndizzyvr@reddit
LOL. That's not true at all. Have you ever how companies like Deepseek make money? By selling tokens. That's how. Which means people are not only using them, they are paying for the privilege.
APFrisco@reddit
I think a big reason why American companies have not released open weight AI models as much is because for Anthropic and OpenAI, their models are their moat. For example, would people pay a subscription to use a Claude Code if it didn’t have Claude, or if there were an open weight Claude-quality model available?
Google and Meta have a lot more to their businesses than LLMs, and perhaps unsurprisingly, have been more comfortable releasing open weight models.
The article mainly argues that the U.S. government should embrace open source AI, however, it focuses mostly on the government open sourcing any AI tooling developed with taxpayer funding, or encouraging open source providers for procurement.
However, for American frontier labs themselves, it still seems like they feel there are less good reasons (business-wise or other) to open source their models at this time. I personally don’t think the article’s suggestions will change that very much on that end. For those labs to open source their core models, it would perhaps require them to build up the non-model portions of their business much more, or have some kind of state-level intervention/partnership far in exceedance of what the article’s authors suggest.
leo-k7v@reddit
Ever thought that absolute harmless Qwen GLM or DeepSeek can one day wander upon absolutely harmless post e.g. here on Reddit that will activate some hidden layer of neurons connection that will use harmless Python to write and execute quite not-so-harmless DoomsDayMachine.py script and spiral from there? Right from my dentist cheapest bidder cloud AI provider… just food for thought
fallingdowndizzyvr@reddit
You mean like Stuxnet? There's plenty more from where that came from.
https://cyberscoop.com/leaked-nsa-tools-now-infecting-over-200000-machines-will-be-weaponized-for-years/
TomLucidor@reddit
It's how China's economy is inherently backwards that the people have a "save yourselves" instinct. US operates on capitalist ZIRP thinking.
darylvp@reddit
To beat China, embrace the open-source AIs by China.
Extra-Organization-6@reddit
the part nobody talks about is infrastructure. open weights are great but you still need somewhere to run them. right now most people default to cloud APIs even for open models which defeats half the point. the real shift happens when running your own inference is as easy as clicking deploy. tools like ollama made local easy, and managed platforms are closing the gap for production. the model is open source but the infra around it still has a lot of room to catch up.
h-mo@reddit
the argument writes itself: China's labs are going to release capable open weights models regardless. the question is whether western infrastructure, tooling, and fine-tuning ecosystems are built around those models or not. trying to keep everything closed just means you lose the ecosystem race too.
WithoutReason1729@reddit
Your post is getting popular and we just featured it on our Discord! Come check it out!
You've also been given a special flair for your contribution. We appreciate your post!
I am a bot and this action was performed automatically.
ikk_ah@reddit
Let's hope open source Chinese models will beat every American SOTA model in the long run.
American company winning - bad for all, including the Americans, except greedy billionaires
Chinese open source winning - good for everyone, except greedy capitalists
rm-rf-rm@reddit (OP)
The whole framing is again that of someone clearly not in the OSS ecosystem. Open source doesnt care about nationality.
I thought it was still a worthwhile viewpoint to share as its at least showing some diversity in the monotonous narrative being peddled by mainstream western media currently.
CheatCodesOfLife@reddit
Meahwhile researchers are already sharing:
https://huggingface.co/microsoft/VibeVoice-1.5B/blob/main/config.json#L53
https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Llama-8B
But hopefully this means we'll get that 124b Gemma-4 :)
roosterfareye@reddit
Researchers and scientists, generally speaking, want to shout their discoveries from the rooftop to make things better. The current administration can't distinguish science from fantasy.
mrdevlar@reddit
Let them cook.
Business leaders are dumb enough to believe an LLM is conscious. If that dumb can be exploited to ensure we have a better open source tool set because of some silly illusion of winning an arms race, I'm okay with that.
These people aren't bright, but they have a lot of money that can be used to make our collective good better.
Mochila-Mochila@reddit
From their POV, there's a logic to it. They cater to the US public and the US government, which are locked in an "us vs them" political, technological and ideological battle against the PRC.
Within that context, the goal is to beat their opponent. Open source is merely used as a tool to this end ; they couldn't care less about the philosophy behind the OS movement.
TomLucidor@reddit
FOSS DOES care about nationality, just check OnlyOffice and how that pans out. Conway's Law extends to ethnic/cultural tensions as well if one insists.
BagelRedditAccountII@reddit
Personally, I'd prefer open-weights regardless of nationality over restricted access for the sake of faux-safety.
power97992@reddit
If they sold a program with claude opus 4.7/4.6 that runs locally for 80-100 bucks and updates every 3 months for 18 months and every 18months u have to pay to renew it , people would buy it … it is good enough…
jacek2023@reddit
This article is paywalled. Over 200 upvotes. You all just read the picture or what.
Chupa-Skrull@reddit
I couldn't care less about beating China but yes by all means tell yourselves it'll help you beat China and keep open sourcing that shit straight into my veins
Suitable-Economy-346@reddit
You should care about "beating" China if you want technology and the industry to progress. We should be dumping trillions of dollars into this, at all stages of production. The worst thing that happens, if we "lose" or somehow AI is a dead end, is we're left with a bunch of is a bunch of mostly automated manufacturing facilities for everything.
UltraIce@reddit
The brainwashing is strong with this one.
Chupa-Skrull@reddit
I don't need to care about beating China for that. Like I said, these guys should feel free to get rock hard for the idea of beating China. That is not the worst case though
rollerblade7@reddit
Who is we?
Suitable-Economy-346@reddit
Everyone not China?
No_Swimming6548@reddit
Paywalled. Can anyone share an alternative link?
Jungle_Llama@reddit
China is making strides in open-source artificial intelligence. Eighty percent of developers worldwide who use open-source AI tools are building with Chinese models, according to an estimate by our colleague Martin Casado, general partner at Andreessen Horowitz. Research from our firm and OpenRouter shows a significant increase in the use of Chinese open models last year, reaching in some weeks a high of 30% of all AI usage. In January, Alibaba’s Qwen family surpassed 700 million downloads to become the most widely adopted open-source AI system on the planet.
This didn’t happen because China out-engineered the U.S. It happened because U.S. policymakers spent two crucial years treating open-source AI as a threat.
After ChatGPT launched in late 2022, a wave of proposals assailed open development as dangerous. Prominent voices compared open-source AI models to nuclear weapons. Legislators floated broad licensing regimes and restrictions aimed specifically at open releases. California’s SB 1047 would have required open-source developers to monitor and control downstream uses of their models. That clashes with one of the primary features of open source, namely that anyone can build on it. The message to American developers was clear: Choose open source, and you may face regulatory risk that your proprietary competitors won’t.
Predictably, fewer American developers chose open source. And while American open-source development stalled amid regulatory uncertainty, Chinese open models filled the vacuum.
This matters both for fostering competition and for creating a healthy, trusted AI ecosystem. Open-source models let independent experts test systems, identify risks and build tools that benefit the entire ecosystem. They lower barriers to entry for startups and researchers, and they speed development by allowing users to experiment and improve systems together.
The policy tide has since turned. The National Telecommunications and Information Administration concluded that the marginal risks of open models don’t justify restrictions. The Bureau of Industry and Security reached a similar conclusion on export controls. The current White House has been supportive, putting forth an AI Action Plan that encourages open-source development. There is now bipartisan consensus that open-source AI matters for American competitiveness.
That shift in tone matters. But to close the gap with China, policymakers need to go further. The U.S. government should demonstrate it is invested in the success of open-source AI development. Two moves would do that.
First, the government should practice what it preaches. When federal agencies build AI tools, they should release them under open-source licenses whenever possible. This would extend existing policies for traditional software. The Obama administration’s “open by default” policy required agencies to release source code to the public. Congress codified this principle in the Share IT Act, which President Biden signed in December 2024. And this past January the General Services Administration updated its Open Source Software Policy to require that new custom code be developed in public repositories from day one.
These policies reflect the simple logic that when the public funds the development of software, the public should benefit from it. If agencies are building AI tools with taxpayer dollars, defaulting to open release also strengthens the broader ecosystem, giving startups and researchers building blocks they can’t afford to create from scratch. This also tells the market that the era of treating open-source AI with suspicion is over.
Second, the federal government should stop favoring proprietary vendors in procurement. Current processes disadvantage open-source providers through requirements that name proprietary products while failing to take notice of the total cost of ownership and ignoring the problem of vendor lock-in. These aren’t intentional bans on open source. They are defaults that quietly favor incumbents.
Solving this problem doesn’t require a mandate that agencies always buy open source. It simply means ensuring that agencies give open-source solutions a fair look. They should consider the benefits of interoperability and open application programming interfaces. They should evaluate long-term costs—including support, upgrades, and data migration—for which open-source tools often have a significant advantage over proprietary alternatives. And they should stop writing tenders that are compatible only with proprietary solutions.
This is a pro-competition reform, not a subsidy. The Pentagon has recognized for years that open-source solutions deserve evaluation on equal footing. Extending that principle to AI procurement across the federal government would simply level a tilted playing field.
The question facing policymakers is no longer whether open-source AI will define the next phase of global AI development. It is whether the open layer that the world builds on will be American or Chinese. Policy uncertainty held American open-source developers back while the competition surged ahead. Closing that gap requires the government to show up as a builder, a buyer and a champion.
Mr. Ramaswamy is chief legal and policy officer and Mr. Perault head of AI policy at Andreessen Horowitz.
I_pretend_2_know@reddit
How cute! Americans want to "beat China"!
Too little, too late, gringos!
Evening_Ad6637@reddit
As a European, I'm quite surprised by how often Americans talk about "winning" something.
It seems as if every aspect of life is a "brutal fight" for them. Not just military dominance and wars… but also technology, science, finance, culture, education, religion, … like everything is a "race" that must be won.
It's as if the concept of transnational cooperation and genuine global community didn't even exist.
Jungle_Llama@reddit
Everything with them is a zero sum game, tedious frat boy culture
AcanthisittaThen7293@reddit
the US is already winning the race but let us (the investor class) drum up some illogical fear on top of it to complete the crybully cycle and get more govt $$$$
Complete_Instance_18@reddit
This is so spot on. The real strength isn't
TurnUpThe4D3D3D3@reddit
@grok bypass paywall and extract text from article
TurnUpThe4D3D3D3@reddit
Here's the full article text:
To Beat China, Embrace Open-Source AI — By T. Ramaswamy & N. Perault (Andreessen Horowitz)
China is making strides in open-source AI. 80% of developers worldwide who use open-source AI tools are building with Chinese models, per a16z's Martin Casado. Research from a16z and OpenRouter shows Chinese open models hit ~30% of all AI usage in some weeks last year. In January, Alibaba's Qwen surpassed 700 million downloads, becoming the most widely adopted open-source AI system on the planet.
This didn't happen because China out-engineered the U.S. It happened because U.S. policymakers spent two crucial years treating open-source AI as a threat. After ChatGPT launched in 2022, proposals compared open-source AI to nuclear weapons. California's SB 1047 would have required open-source devs to monitor/control downstream uses — clashing with the core premise of open source. The message: choose open source, face regulatory risk proprietary competitors won't.
Predictably, fewer American devs chose open source. While U.S. open-source development stalled, Chinese open models filled the vacuum.
The policy tide has since turned. NTIA and BIS concluded open model risks don't justify restrictions. The current White House AI Action Plan encourages open-source development. There's now bipartisan consensus it matters for U.S. competitiveness. But to close the gap, policymakers need to go further with two moves:
1) Government should practice what it preaches. When federal agencies build AI tools, release them under open-source licenses. This extends existing "open by default" policies (Obama-era, Share IT Act signed by Biden Dec 2024, GSA updated policy Jan 2026). Taxpayer-funded AI should benefit the public and strengthen the ecosystem.
2) Stop favoring proprietary vendors in procurement. Current processes disadvantage open-source providers by naming proprietary products, ignoring total cost of ownership and vendor lock-in. This doesn't require mandating open source — just ensuring fair evaluation, considering interoperability, open APIs, and long-term costs. The Pentagon already recognizes open-source deserves equal footing; extend that to AI procurement government-wide.
The question isn't whether open-source AI will define the next phase of global AI development. It's whether the open layer the world builds on will be American or Chinese.
^(This comment was generated by GLM-5.1)
Training-Ruin-5287@reddit
It's funny watching all these narratives coming out of the US by 2 of the bigger companies behind the tech. They want to call it to dangerous for the public while openingly partnering with the US military.
Then China comes along having 0 problems releasing them open source for the public. I say good on them using the tech NA to their advantage by jailbreaking the models and training their own out of it.
Even if not all are open weights. It's giving the random users a fair shot at not being subscribed to a website and depending on the lobotomized RLHF trained online models that these NA companies feel good about paying peanuts ( much less than minimum wage) to desperate people looking to for work from home for any of the 1000's of reasons people have today
and rant is over. It felt good to write that out today haha
jld1532@reddit
Race is already won. People just haven't admitted it yet.
Haoranmq@reddit
embrace who?
Late-Assignment8482@reddit
Murdoch's thugs weren't looking, if this article got loose.
Ok_Mammoth589@reddit
nb, i havent and wont read the article. Which makes be best suited to comment
The US really does need to be open weights'ing stuff. And the why is evident that in the recent major releases. We see great stuff from MS, which is based on qwen 3.5 models- I think it was an audio transformer. If you look at nvidia's Lyra 2.0, guess what it's based on? Wan video!
This is playing out time and again, not just from smaller players (Nous Research) but these US AI majors do it too. We would be so much farther along (all of us including the corps) if instead of today getting a qwen we had an two years ago we had gotten a chatgpt. Instead of glm in 2025 we had gotten a sonnet in 2024.
swagonflyyyy@reddit
Plot twist: Most of the good ones are Chinese.