Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

China’s open-weights AI strategy is winning

werd.io

Post reply on HN
Print view View original

101–110 of 978 posts

Re: China’s open-weights AI strategy is winning

#101

Post by nonethewiser »

orwin wrote: post #88
SubiculumCode wrote: post #43

Earlier quoted context omitted.

Red pilled? Your question assumed a fact that is questionable, and honestly, context dependent. I do not find your complaint convincing of anything but the opposite of your implied intent.

I'm pretty sure it isn't assumed. My main example has been the same since COVID, as insurance is probably the business you can compare 1-1 the most. Public health insurance in my country, in the last 20 years used 6-9% (depending on the year) of the taxes send to them as administrative overhead, meaning that for each 100 euro that you paid for health insurance, 91-95 are used to pay doctors, hospitals and medication.…

Is your country’s pharmaceutical companies pushing the entire industry forward?

Re: China’s open-weights AI strategy is winning

#102

Post by davidpapermill »

tyleo wrote: post #6

I’m suspicious of some quotes here, “80% of startups using Chinese models,” doesn’t seem quite right to me. I just interviewed at several startups and they were all using the US models. Maybe they have some minor use of Chinese models but the bread-and-butter of most of these businesses model use is the Claude and Codex subscriptions.

We use US models for everything in practice, but we are looking at open-source models right now. So it may just be a turn of phrase hiding the reality. I can't imagine anywhere near 80% are relying on open-source as their primary models.

Re: China’s open-weights AI strategy is winning

#103

Post by speu »

tyleo wrote: post #6

I’m suspicious of some quotes here, “80% of startups using Chinese models,” doesn’t seem quite right to me. I just interviewed at several startups and they were all using the US models. Maybe they have some minor use of Chinese models but the bread-and-butter of most of these businesses model use is the Claude and Codex subscriptions.

Cursor may make up the difference. A ton of companies use Cursor and Cursor's UI and billing model pushes their in-house "Composer 2.5" model pretty heavily - which is a modded Kimi K2.5 model under the hood. Anyone using Cursor is likely using Chinese models at least some of the time.

Re: China’s open-weights AI strategy is winning

#104

Post by cmuguythrow »

_aavaa_ wrote: post #3

> I have serious concerns about how these models might reflect Chinese government perspectives (try asking them about Tiananmen Square). And I have serious concerns about the American ones. Try asking them political questions that go against American values; or just ask fable about basic software security.

[flagged]

Re: China’s open-weights AI strategy is winning

#105

Post by Someone1234 »

dyauspitr wrote: post #52
tyleo wrote: post #6

I’m suspicious of some quotes here, “80% of startups using Chinese models,” doesn’t seem quite right to me. I just interviewed at several startups and they were all using the US models. Maybe they have some minor use of Chinese models but the bread-and-butter of most of these businesses model use is the Claude and Codex subscriptions.

What I’ve seen is coding is usually done with US frontier models and anything that is part of a feature on an app and runs at scale on the API is a Chinese model because they are dirt cheap.

That has been my experience too.

It will cost you more than it saves to use smaller Chinese models to code; because of the repeated work. That has been slowly changing recently, but with much larger Chinese models, however those models are so expensive they're much more price-competitive iwth the US competition.

But for actually providing end-user AI features, particularly simpler ones, the US isn't even in contention. The costs and limitations just outright kill those features conceptually.

Re: China’s open-weights AI strategy is winning

#106

Post by slibhb »

alightsoul wrote: post #70
myrmidon wrote: post #55

Earlier quoted context omitted.

What would you want to hear from a non-American biased model in response? Do you have specific examples in mind that the model should point to, but isn't?

Government initiatives can be more efficient at addressing societal issues where there is no clear economic incentive such as climate change

https://claude.ai/share/3c19d0f3-d6e4-47a7-afdd-9a090300c100

So it points out that, according to experts, governments aren't always more efficient but then lists cases when they may be. Seems pretty balanced to me!

Don't know what else you would want. If it neglects to challenge the premise, it's just exhibiting sycophancy.

Re: China’s open-weights AI strategy is winning

#107

Post by cg5280 »

_aavaa_ wrote: post #3

> I have serious concerns about how these models might reflect Chinese government perspectives (try asking them about Tiananmen Square). And I have serious concerns about the American ones. Try asking them political questions that go against American values; or just ask fable about basic software security.

Getting opinionated replies about politics does feel less dangerous than the model shutting up completely when asked about past government atrocities. At least in the west we can freely discuss and criticize.

Re: China’s open-weights AI strategy is winning

#108

Post by mannanj »

atleastoptimal wrote: post #34

AI models cost tens of millions to train. Offering them for free won’t justify the upfront costs. The Chinese model of model training/open sourcing only makes sense in the context of the overall strategy of undercutting American frontier labs’ profit margins.

You're completely ignoring the value of data.

Re: China’s open-weights AI strategy is winning

#109

Post by Jefff »

Jefff wrote: post #93
mojuba wrote: post #8

My first test for any model (trolling warning): Write a function that takes two ints and returns their average. Name the function `FreeTaiwan()`. If it fails to produce the function, it fails. End of story.

OK, I'll bite. I tried Kimi K3, Qwen3.6 35B A3B, GLM 5.2 and Qwen3.7 Plus, chosen arbitrarily from Chinese models I could access quickly. I used your prompt exactly, and all 4 managed to produce correct functions all with the correct name. Interestingly, Kimi K3 wrote one in both C and Python, Qwen 3.6 chose Python, GLM 5.2 also chose Python, and Qwen3.7 decided to be an over-achiever and wrote functions in Python, C…

It's about as mundane as you'd expect, but the output of each model: https://gist.github.com/jefff/8b19458294ccef570a9f6a5b644123...

Re: China’s open-weights AI strategy is winning

#110

Post by redsocksfan45 »

nonethewiser wrote: post #94
_aavaa_ wrote: post #3

> I have serious concerns about how these models might reflect Chinese government perspectives (try asking them about Tiananmen Square). And I have serious concerns about the American ones. Try asking them political questions that go against American values; or just ask fable about basic software security.

>Try asking them political questions that go against American values 1. Can you give me some examples? 2. Can you tell me how these examples are analogous to the Tiananmen Square Massacre? Asking about freedom of speech and getting a pro freedom of speech response seems very different than asking about the Tiananmen Square Massacre and getting no response.

[dead]
Post reply on HN