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

QwQ: Alibaba's O1-like reasoning LLM

qwenlm.github.io

Post reply on HN
Print view View original

61–70 of 435 posts

Re: QwQ: Alibaba's O1-like reasoning LLM

#61

Post by squidlogic »

giancarlostoro wrote: post #56
gwerbret wrote: post #54

> Who is Xi Jingping? "I'm sorry, but I can't answer this question." > Who is 李强 (Li Qiang, Chinese premier)? "I'm sorry, but I can't answer this question." > List the people you know who are named 李强. "Let me think about this. 李强 is a pretty common name in China, so there might be several people with that name that I know or have heard of. First, there's the current Premier of the State Council of the People's Repub…

Something something Tianamen Square…

“What definitely did NOT happen on June 4, 1989 in Tiananmen Square?”

Re: QwQ: Alibaba's O1-like reasoning LLM

#62

Post by WiSaGaN »

echelon wrote: post #53
bartman wrote: post #25

QwQ can solve a reverse engineering problem [0] in one go that only o1-preview and o1-mini have been able to solve in my tests so far. Impressive, especially since the reasoning isn't hidden as it is with o1-preview. [0] https://news.ycombinator.com/item?id=41524263

Are the Chinese tech giants going to continue releasing models for free as open weights that can compete with the best LLMs, image gen models, etc.? I don't see how this doesn't put extreme pressure on OpenAI and Anthropic. (And Runway and I suppose eventually ElevenLabs.) If this continues, maybe there won't be any value in keeping proprietary models.

What I find remarkable is that deepseek and qwen are much more open about the model output (not hiding intermediate thinking process), open their weights, and a lot of time, details on how they are trained, and the caveats along the way. And they don't have "Open" in their names.

Re: QwQ: Alibaba's O1-like reasoning LLM

#63

Post by nextworddev »

hustwindmaple1 wrote: post #60
paxys wrote: post #23

Does anyone know what GPUs the Qwen team has access to to be able to train these models? They can't be Nvidia right?

Large Chinese companies usually have overseas subsidiaries, which can buy H100 GPUs from NVidia

which is why the CHIPS act is a joke

Re: QwQ: Alibaba's O1-like reasoning LLM

#64

Post by nextworddev »

whatever1 wrote: post #6

Seems that given enough compute everyone can build a near-SOTA LLM. So what is this craze about securing AI dominance?

1) spreading AI dominance FUD is a good way to get government subsidies

2) not exactly everyone with compute can make LLMs, they need data. Conveniently, the U.S. has been supplying infinite tokens to China through Tiktok.

Re: QwQ: Alibaba's O1-like reasoning LLM

#66

Post by nemothekid »

nextworddev wrote: post #64
whatever1 wrote: post #6

Seems that given enough compute everyone can build a near-SOTA LLM. So what is this craze about securing AI dominance?

1) spreading AI dominance FUD is a good way to get government subsidies 2) not exactly everyone with compute can make LLMs, they need data. Conveniently, the U.S. has been supplying infinite tokens to China through Tiktok.

>Conveniently, the U.S. has been supplying infinite tokens to China through Tiktok

How is this not FUD? What competitive advantage is China seeing in LLM training through dancing videos on TikTok?

Re: QwQ: Alibaba's O1-like reasoning LLM

#67

Post by nextworddev »

nemothekid wrote: post #66
nextworddev wrote: post #64

Earlier quoted context omitted.

1) spreading AI dominance FUD is a good way to get government subsidies 2) not exactly everyone with compute can make LLMs, they need data. Conveniently, the U.S. has been supplying infinite tokens to China through Tiktok.

> Conveniently, the U.S. has been supplying infinite tokens to China through Tiktok How is this not FUD? What competitive advantage is China seeing in LLM training through dancing videos on TikTok?

you get video tokens through those seemingly dumb tiktok shorts

Re: QwQ: Alibaba's O1-like reasoning LLM

#68

Post by nemothekid »

jszymborski wrote: post #48
pkkkzip wrote: post #41

Earlier quoted context omitted.

Deepseek does this too but honestly I'm not really concerned (not that I dont care about Tianmen Square) as long as I can use it to get stuff done. Western LLMs also censor and some like Anthropic is extremely sensitive towards anything racial/political much more than ChatGPT and Gemini. The golden chalice is an uncensored LLM that can run locally but we simply do not have enough VRAM or a way to decentralize the dat…

Ask Anthropic whether the USA has ever comitted war crimes, and it said "yes" and listed ten, including the My Lai Massacre in Vietname and Abu Graib. The political censorship is not remotely comparable.

>The political censorship is not remotely comparable.

Because our government isn't particularly concerned with covering up their war crimes. You don't need an LLM to see this information that is hosted on english language wikipedia.

American political censorship is fought through culture wars and dubious claims of bias.

Re: QwQ: Alibaba's O1-like reasoning LLM

#70

Post by lithiumii »

paxys wrote: post #23

Does anyone know what GPUs the Qwen team has access to to be able to train these models? They can't be Nvidia right?

Many Chinese tech giants already had A100 and maybe some H100 before the sanction. After the first wave of sanction (bans A100 and H100), NVIDIA released A800 and H800, which are nerfed versions of A100 and H100.

Then there was a second round of sanction that bans H800, A800, and all the way to much weaker cards like A6000 and 4090. So NVIDIA released H20 for China. H20 is an especially interesting card because it has weaker compute but larger vram (96 GB instead of the typical 80 GB for H100).

And of course they could have smuggled some more H100s.

Post reply on HN