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

Qwen 3.8

twitter.com

Post reply on HN
Print view View original

391–400 of 793 posts

Re: Qwen 3.8

#391

Post by moffkalast »

villish wrote: post #363
moffkalast wrote: post #345

Earlier quoted context omitted.

Do you have proof it's not? There are no laws they have to follow in regards to it, and it's practically impossible for them to go against their core self interest. More data is literally a direct component to better models and more revenue. If someone proves it's fake worst they get is a month of bad PR and then people will move on, as they always do.

> If someone proves it's fake worst they get is a month of bad PR and then people will move on. You can say they stole from everyone to train their models in the first place and that's valid, but this isn't that. You are saying they are actively ex-filtrating data from any company using their services and lying about it. Google/Apple/Microsoft or all of the dozen trillion dollar companies in the US would absolutely c…

I'm sure in the handful of cases where they have a genuine bulk contract on paper with a large firm that states they'll be in breach if they train on it, they do abide by it.

But unless you are one of those you mentioned and a few others you probably aren't notable enough to care about. Everyone who uses their services directly, paying or not, is surely ignored in that sense. I wouldn't be surprised if there's a team of their own lawyers ready to interpret their EULA in fascinating ways.

And out of those three I'd only probably assume Apple is the only one who doesn't use the data given that they've built up privacy as a selling point, MS and Google probably train their own models on it themselves.

Re: Qwen 3.8

#392

Post by SwellJoe »

redman25 wrote: post #241
storus wrote: post #58

Earlier quoted context omitted.

DeepSeek V4 hallucinates like crazy and often forgets explicitly mentioned parts of the context. I guess compressing tokens and cherry-picking attention comes at a cost.

Deepseek is one of the worst in terms of hallucination rate according to artificial analysis' benchmark: https://artificialanalysis.ai/?omniscience=omniscience-hallu...

That's interesting. The "best" current models, Fable and especially GPT 5.6, are also lying liars that lie all the time. Seems like we're going the wrong way on hallucination.

Re: Qwen 3.8

#393

Post by SwellJoe »

dannyw wrote: post #43
SwellJoe wrote: post #34

Qwen is the most censored of the Chinese models in my testing, which makes me wonder in what other ways it is compromised. Open weights doesn't really reveal what's in there. And, in my tests, existing Qwen models are not at the pareto frontier of any metric; DeepSeek V4 Pro is better, faster, and much cheaper than Qwen 3.7 Max. (DeepSeek is also among the least censored of the Chinese models.) I guess we'll see if t…

What censorship? ;) https://github.com/p-e-w/heretic

You still don't know what's going on in there.

Re: Qwen 3.8

#394

Post by elmer2 »

grommz wrote: post #298
andsoitis wrote: post #285

Earlier quoted context omitted.

Then why doesn’t he give it away for free?

It's your job to vote for a government that gives you cheap or free AI. Europe is building AI gigafactories so that small businesses can have access to cheap AI. At least that's the plan.

Easy to do, when you don't need to put billions of dollars into research, and just have the end product.

It reminds me of socialized healthcare: wait for US companies to develop the drugs, buy a cheap generic, and somehow point to it as a superior model.

Re: Qwen 3.8

#395

Post by watwut »

mceachen wrote: post #343
TheDong wrote: post #335

Earlier quoted context omitted.

I feel like there could also be a simpler explanation. Why does a debian contributor make debian free, why do they work on this thing anyone can use? Is it because linux and debian hate windows and iOS and want to see american fail? No, it's because most debian contributors believe software source code, information, should be free, users should be free to modify the code they use, and that they're building a thing th…

Please don't conflate a volunteer effort with no expected economic gain with a very well funded company (or fleet of companies). With the CCP's highly successful track record with subsuming other markets, Occam's razor applies to why they're doing this.

1.) Majority of debian contributions are from people paid for it.

2.) It is scary because they do what sillicon valley did for decades? While it bragged about disruption being the goal?

Re: Qwen 3.8

#396

Post by jchw »

nylonstrung wrote: post #390
jchw wrote: post #380

Earlier quoted context omitted.

Well of course nobody believes that, there are a large number of great Chinese open source projects, and plenty of great Chinese contributors to open source projects. I have zero doubts that Chinese people are at least equally as capable of embracing open source as anyone else. It would be strange to suggest otherwise. That said, though, I do have trouble believing the long-term story for open weights, anywhere. We d…

The CCP encourages open source contributions from companies, you can see this in how much Alibaba open sources across their entire stack Chinese tech leans much more heavily towards build vs. buy than the SaaS dominated West (where programmers are more expensive) so the positive externalities on their tech industry are more pronounced

Sure, when it comes to the big corporations I don't deny CCP's influence on their strategy. Hell, in some cases, it's easy to root for their strategy, because sometimes our tech industry sucks. Don't have to love them to occasionally agree with them.

But, in terms of individuals, of course, we're really not so different.

Re: Qwen 3.8

#397

Post by organsnyder »

SwellJoe wrote: post #34

Qwen is the most censored of the Chinese models in my testing, which makes me wonder in what other ways it is compromised. Open weights doesn't really reveal what's in there. And, in my tests, existing Qwen models are not at the pareto frontier of any metric; DeepSeek V4 Pro is better, faster, and much cheaper than Qwen 3.7 Max. (DeepSeek is also among the least censored of the Chinese models.) I guess we'll see if t…

I had a fun conversation with Qwen 3.5 a while back. I ended up getting it to admit that it was complicit in the coverup of the Tiananmen Square massacre. This was running locally—it wouldn't surprise me if they had additional safeguards for their hosted service.

Re: Qwen 3.8

#398

Post by matheusmoreira »

hodgehog11 wrote: post #14

The "second only to Fable 5" comment is pretty telling here. I remember early on when a lot of naysayers were saying that Fable was barely an improvement on Opus. Like it or not, Anthropic have a genuine moat right now with that model, provided they continue to allow people to use it. It will be genuinely exciting when an open model is able to beat it.

What's the point of Fable if we can't use it? I get to prompt it like 5 times on my subscription before it gets cut off, and even then I'm constantly fighting the insufferable safety classifier.

I'll switch to OpenAI soon because of this. I also can't wait for the day it becomes feasible to run these awesome open weight models on my own hardware.

Re: Qwen 3.8

#399

Post by throwaway-11-1 »

elmer2 wrote: post #382
TheDong wrote: post #335

Earlier quoted context omitted.

I feel like there could also be a simpler explanation. Why does a debian contributor make debian free, why do they work on this thing anyone can use? Is it because linux and debian hate windows and iOS and want to see american fail? No, it's because most debian contributors believe software source code, information, should be free, users should be free to modify the code they use, and that they're building a thing th…

The Chinese government is so nice and giving. In fact, they have so much love in their hearts for the Uyghur people, they created a special mobile app for them, just to make sure nothing bad happens to them.

You do realize the US was fighting Uyghur terrorists alongside China 20 years ago in ago in Pakistan’s and Afghanistan for their support and cooperation with Al Queda? Like I know everyone is supposed to hate China now or whatever but can you guys show a little consistency?

Re: Qwen 3.8

#400

Post by watwut »

kelnos wrote: post #324
gardnr wrote: post #6

Earlier quoted context omitted.

It's hard to say what their motivation is. The Chinese firms seem to be working hard to commoditize intelligence which may be the most effective way to debase American frontier labs. And yeah: it also happens to be really good for humanity.

> It's hard to say what their motivation is. Feels pretty easy to me. They want to turn LLMs into a commodity, and watch the US AI labs crash and burn. There will still be plenty of customers who will pay them to host the models and run inference, even if the weights are open and others can offer competing products. (If necessary, the Chinese government can ban use of foreign inference services by Chinese citizens an…

> They want to turn LLMs into a commodity, and watch the US AI labs crash and burn.

I want to watch that too.

If they take Meta and Musk with them, all the better, but that is just dreaming I am afraid.

Post reply on HN