Hacker News Reader

Comment by pandoro | Hacker News Reader

All frontier models have been trained without any regards for IP protection laws. I don't see how anyone can argue in good faith that distillation is not fair game and does not ultimately "benefit humanity™"

View on HN

Does distilling actually violate any IP protection laws? Sure, it’s against their terms of service.

I get what you're saying and two wrongs do not make a right but the irony, and why people are even talking about this, is that the thing being distilled clearly, and knowingly, violated copyright & terms across the entire internet.

Yeah, everyone is saying that but I don't find the irony very interesting.

If it does then their original training did as well.

For me, I don't really care about the theft aspect but when people are claiming that these open models are better value or going to overtake anthropic/openai models, the implication that the open models are training of distilled data means all the "progress" they are making is just mimiced from the closed models.

It's a bit interesting how the open models are able to keep pace with the closed models except whole maintaining a steady following time.

Distillation is absolutely not the reason they're good. It's not necessarily even done on a more capable model. It can even be done on itself and still bring improvement, or on a weaker model as well (see GLM and Gemini, which is definitely true because it repeats Deepmind's injections).

If distillation doesn't make them better, why do it?

If using Anthropic models for distillation doesn't make them better, why do it?

It does, it's just not the reason. Most of the work is done before that point, and as I said z.ai used a weaker model. Besides, this is all strictly one-sided, as nobody knows how much Anthropic and OpenAI borrowed from Chinese labs' open research and weights (and they innovated a lot, to put it mildly, starting with first reasoning models worth talking about long before OAI did the same). Chinese labs are also severely restricted on hardware.

This entire story makes certain American AI shops look cartoonishly evil and Chinese ones relatively sane. Not only they want to grab without giving anything back, they also want to sabotage everyone else's AI research and do plenty of terrible things like media manipulation on the global scale and getting in bed with the government. This can't possibly end well, for the Americans in the first place.

It's important to note that even if these open models are distilled, they are showing genuine improvements in their architecture, which enables inference costs to be several factors below what equivalent closed models have.

The interesting question is: will Anthropic release a Fable like model with an architecture similar to Kimi, and get the inference cost gains? They should surely beat Kimi because they can internally distill as much as they want.

One mentions distillation to intimate that one is in fact still the leader. It has zero to do with fairness.

Training a model is very expensive and creates something no individual rights-holder could. Distilling a model copies this value add and captures it without bearing the cost that created it.

What of the costs for creating the data that was used to train the model being distilled?

At least 1.5B by stealing books, per a recent ruling.

I don’t feel sorry for the model companies

They only had to pay for storing the books on a server for later possible use. They did not have to pay anything for the training which was declared fair use.

This actually undermines the argument that distilling is harmless because its founded on the idea that Anthropic did the same thing and didn’t have any repercussions.

Do you people seriously believe that Moonshot and the Chinese personal-cult-state dont possess and train on the same torrents?

They aren’t hypocritically crying foul about it, so no we don’t care if they do.

Well of course. But they dont hate Chinese model companies.

An incredibly ironic comment.

I'm going to hope this was sarcasm and if it is, it's great.

the outputs of the model have no property protections, and training a model on the outputs of another model does the exact same thing - its expensive and creates new value over what was in the input - a set of documents.

Could you give an example of the value that only training a model can create but none of the rights-holder could? I feel like if you got a direct, instant communication channel to any of the rights-holder that created the content in the training set of those models, you'd get more value than what the LLM could ever give you on any specific subject.

>Could you give an example of the value that only training a model can create but none of the rights-holder could?

Yeah.

LLMs

Yes. I bet Moonshot paid for API access as opposed to pirating like Anthropic did.

I understand you think Anthropic should have paid for the information it trained the models on. But im talking about all the costs to build a model. Do you think Anthropic didnt spend money to build these models? Did you not know that its actually very expensive?

Moonshot pay for access at the price point Anthropic sees fit, potentially more, since they likely had to jump through multiple hoops?

It's hard to feel sorry about the breach of their ToS, which ultimately is all Anthropic can argue, when they are constantly being sued by countless IP owners

A lot of those arguments could be said about writing a book, or a decent forum guide.

There is a lot more one has to be good at to make a fable level model. Distillation won't get you there, it will help refine some at the end

Reply on news.ycombinator.com