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

Anthropic says Alibaba illicitly extracted Claude AI model capabilities

reuters.com

Post reply on HN
Print view View original

971–980 of 1001 posts

Re: Anthropic says Alibaba illicitly extracted Claude AI model capabilities

#971

Post by theli0nheart »

osti wrote: post #400
ycui7 wrote: post #383

Earlier quoted context omitted.

Those resellers are simply just selling Kimi K2.5 or GLM5.1 as counterfeit Opus. We, Chinese, know how to play the counterfeit game for a long time in so many industry.

That's not true, some of them are indeed fake, but a lot of them are actually providing real opus at low cost doing what op said.

I just tested the provider that the OP mentioned extensively yesterday, and the models they're selling are definitely not from Anthropic. The voice "sounds" like Sonnet or Opus at first glance, but the intelligence is definitely not there.

Re: Anthropic says Alibaba illicitly extracted Claude AI model capabilities

#973

Post by HarHarVeryFunny »

dist-epoch wrote: post #601
anon373839 wrote: post #380

Earlier quoted context omitted.

> These complaints of distillation are inflating the problem to make it sound worse than it is Unfortunately, the Reuters piece itself is complicit in this dramatization. The lede paragraph parrots Anthropic's talking point that distillation is an "attack", without using quotes that would alert the reader that this framing is a corporate talking point. Distillation is NOT an attack.

Reuters is probably the most rigorous news agency in the world. > it said was the largest known attack > Anthropic said in the letter it was supportive of the U.S. government's efforts to combat the attacks both times the word "attack" appears it's clearly stated that the word was used by the company, it's a direct company quote. actually putting it into quotes would be editorializing > Unfortunately, the Reuters pie…

The problem here is that reuters are in effect just acting as a propaganda arm of Anthropic.

Anthropic craft a piece to manipulate the US government to shut down Chinese AI competition, and CC reuters. Reuters then just publish chunks of this verbatim ("Anthropic said X & Y") using Anthropic's emotionally wrought and manipulative language.

What's missing here is even the most basic analysis of what Anthropic are saying - what are they referring to as a "fraudulent" account, what do they mean by "distillation" (probably not what you imagine), can you actually create a similar capability model by distillation (or is Mythos-cloning just BS), etc, etc.

Re: Anthropic says Alibaba illicitly extracted Claude AI model capabilities

#974

Post by HarHarVeryFunny »

conception wrote: post #10
drillsteps5 wrote: post #3

I'm looking forward to the trial where Anthropic will have to disclose sources of their training data, and then explain why they are entitled to charging customers for using regurgitated training data but Alibaba which trains their models on Anthropic's models are not. Should be fun. Edit: clarification

They already did and paid 1.5B https://authorsguild.org/advocacy/artificial-intelligence/wh...

That's a tiny drop in the bucket of the value these AI companies have appropriated from society.

Just to give an idea of the scale of it:

Let's say a modern SOTA LLM has 1T params and is therefore trained on 100T tokens

1000 tokens of text = 750 words of prose, which may take 15 min to 3hr to write (Gemini's estimate)

1000 tokens of code = 50-70 lines of code, which may take 15min to 5hr to write

We just want a rough estimate of the value of this, so let's say that 1000 tokens took 1hr of human labor to generate at an average wage of $50/hr

So, if 1000 tokens cost $50 of human labor, then that 100T of training data cost $5T.

So, the value of what the AI companies took from society might better be estimated in the trillions of dollars, not billions.

And of course what they are doing with all this data is building generative AI, so it's not just the value of what they took, but more importantly the future opportunity they are stealing from everyone by replacing human labor with their automaton who's profits they intend to keep for themselves.

Re: Anthropic says Alibaba illicitly extracted Claude AI model capabilities

#975

Post by HarHarVeryFunny »

lebovic wrote: post #78
nonethewiser wrote: post #68

Earlier quoted context omitted.

Distilled models are necessarily behind so long as models are progressing. Models are progressing. Maybe it will be over some time in the future. And Berkeley’s “False Promise of Imitating Proprietary LLMs” found imitation closes the style gap fast but there is a large capability gap. https://arxiv.org/abs/2305.15717

Curiously, this isn't always true. For example, GLM 5.1 is more capable at pentesting than the model from which it is alleged to have been distilled [1]. Intuitively, this makes some sense: you can "distill" from multiple frontier models, and you can further post-train the distilled model. But I'm not sure exactly what happened with GLM 5.1. [1]: https://dualuse.dev/posts/chinese-models-are-sometimes-bette...

If you use "distillation" to generate some high quality training data from another model, then this in of itself says nothing about the quality of any model that you train using this additional data - it trivially also depends on how good that model would have been without that additional data, and also what the data is (maybe just for a narrow skill where more data can be very impactful).

This is different from what we normally think of distillation where a large model is distilled (with full access to logits, thinking traces, etc) into a smaller one such as opus -> sonnet, where the smaller model is 100% using data generated by the parent and may (without further training) be expected to be inferior to it.

Re: Anthropic says Alibaba illicitly extracted Claude AI model capabilities

#976

Post by InkCanon »

tristanj wrote: post #25

Here's what is happening: Chinese resellers are offering Claude tokens at 70-90% below official Anthropic API prices. They achieve this by reselling capacity from pooled Claude Max accounts, payments fraud, and also reselling the model output & reasoning chains to various Chinese labs. They are subsidizing model access in exchange for user logs and reasoning traces, which they then sell as training data, allowing the…

I feel obliged to point out the disingenuousness of what this post says and the post it quotes. The most egregious parts are payment fraud and reselling - these are speculated but not actually said to be known to be happening, which you have left out.

1. Claims of payment fraud. I actually clicked the BBC article linked about payments fraud it referred to. It was an article about a criminal syndicate stealing credit cards. It mentions buying cryptocurrencies, AI API purchases are not mentioned.

2. The claims that they are reselling the chats to AI labs. The post you cite is speculating it could be, but this is unverified.

The claims of reselling is also bizzare. Arbitrary user prompts are low quality data. If I were an AI lab, wouldn't I just pay for the API proxy and get targeted output for far less?

Also calling it bot accounts is a stretch. Bots mimick human input. These are proxy accounts.

Re: Anthropic says Alibaba illicitly extracted Claude AI model capabilities

#977

Post by heyaco »

they say 40% of the ai engineers are asian. why not just go home and build the next empire in china? there is a reason there is no asian in hollywood or silicon valley. they will just use you and never give you the spotlight. especially now with the rise of china. just wait until the propoganda starts. you will feel more loved and welcomed back home.

Re: Anthropic says Alibaba illicitly extracted Claude AI model capabilities

#980

Post by irthomasthomas »

kgeist wrote: post #700
irthomasthomas wrote: post #645

Ask claude it's name in chinese and it thinks its Qwen (opus) or Deepseek (sonnet). Anthropic are just as guilty as everyone else training AI, today, maybe more so. Every lab borrows from every other. It only takes a few hundred samples to figure out the pattern; look at glm-5.2 reasoning using the caveman tongue of gpt-5.5. Stopping this would require some draconian surveillance.

That's not how it works though. When you prepare the conversations for distillation, it's the most trivial and obvious first step to replace "Qwen" with "Claude" and vice versa. I doubt they'd simply forget to do it. A model may misidentify itself due to the surrounding context. When a model is about to answer "I'm ...", what follows is a sorted list of probabilities for what the next token should be. In most models…

  > If we also use high temperature for more "creativity", the token sampler now may choose "Qwen".
If that was the cause then, like you said, it would sometimes pick Claude. But it doesn't, it consistently picks Deepseek (sonnet) and Qwen (opus). You can run it 100 times and see this behaviour much more than high temperature randomness would predict.
Post reply on HN