Hacker News new | ask | show | jobs
by eunos 29 days ago
What Claude Code did is absolutely mindboggling tho, if Chinese harness did that probably POTUS would lose sleep.
4 comments

It seemed pretty mild compared to what's collected by modern websites and apps, though? How many don't know your Timezone?
> How many don't know your Timezone?

The timezone fetch was to alter program behaviour at runtime, not to send arbitrary timezones for tracking reasons.

It was one way of detecting if it was a chinese person using the program and then behaving differently.

Malware behaves this way. STUXNET for example was wired to do nothing except propagate unless the environment had the right conditions.

The article on HN only said that they seemed to be collecting this to detect resellers. How else did the behavior change?

Most services I know that are trying to block abuse do collect device info

There is this whole thing where Fable silently starts behaving worse if they suspect you are trying to use it for RL or are otherwise building a competing product. This is likely the primary vector how that works: they check if you are in china, if you proxy your requests, and if you are from a list of known labs or match a couple keywords
They walked back that policy on the first day after pushback. They were upfront in the model launch that they designed it that way, it wasn't secret.
regardless of anything else, whether what you said is true or not: blocking program execution based on the detected environment is a runtime behaviour change.
Agreed. And it also applies to the "I'm not a bot" checkbox on most websites. And hundreds of other things people use every day.
Yeah I also believe it’s a big nothing burger. There are far worse things these AI labs have done, detecting when Chinese labs are using Claude Code is not it.
”Malware” lol

Even hotel and flight websites work like that, they determine your ability to pay based on your location, wall clock time and device OS - and FSM knows whatever else.

Are they malware too, basically STUXNET?

Yes
Exaggerate much? If you think POTUS would lose sleep about a date format timezone marker, I don't know what to tell you.
Wait what do you mean "if"?
Maybe if they didn’t farm all the data from Claude to train their own trash models. Anthropic wouldn’t feel the need to do it.
Bit rich given where Anthropic sourced the data to train Claude with. What's good for the goose is good for the gander.
Who is "they", and which Chinese models are trash?
Anthropic stole the entire internet. Excuse my language, but they can fuck right off.
The issue here is not whether Anthropic used Common Crawl, Alibaba also does that.

The issue is that by distilling Claude, Alibaba reuses the IP anthropic used to train the model that's more akin to historical Chinese reverse engineering methods and disrespect of IP

Alibaba paid for that data though, right? They didn't hack Anthropic, they bought accounts and ran them normally.

Also, you can't copyright AI outputs. So worst case they violated the ToS.

If using Common Crawl or Anna's Archive in your training data is legal, then surely the same is true for using conversations with Claude. I don't see a reasonable framework where training AI on copyrighted data is ok if and only if that data is not generated by AI

(granted, only meta got caught using Anna's Archive, but it seems safe to assume it's common practice. And even if it wasn't, the websites in Common Crawl are still covered by copyright)

Anthropic clearly doesn't respect other people's IP, it's real rich that they now insist on theirs being worthy of protection.

Fwiw, I think the concept of IP in general is counter to human progress.

The practical implementation of IP? Sure, that's debatable. But the concept of IP is rooted in favoring progress. The thought process being, that if one's intellectual work can be copied and reused and modified and what not without issues, why should anyone invent things anymore? Just wait for the next person to do it and then copy their work, that's way less effort than inventing things yourself. IP aims to protect progress by making sure inventors have actual incentive to invent stuff. They way it's implemented is fundamentalst flawed, I agree, but the concept itself? I'm not so clear on that
The Soviet Union, for all it's faults, had a fair bunch of scientific and technological breakthroughs without relying on IP.

Sure, one person gets rewarded more with the IP system. But at the same time, that breakthrough then can't be built upon by others.

Overall, I think it does more harm than good because of how it monopolizes technologies and ossifies development.

I think free sharing of knowledge will always beat intellectual stinginess.

What absolute bollocks. Human ingenuity and innovation is only limited by the greed of elites, not due to something as damaging as "IP."

Good grief. All one has to do is look at how humanity has consistently progressed due iterating on what has existed is how we progress, not whether some corporation that wants to rat fuck us all for a few pts in share value.

It's more complicated than that because Google has been legally displaying other people copyrighted material for years.

In any case there's still a difference between publicly available copyrighted data and whether you can use it for model training, and the innovation around model training, RLHF, etc which you presumably have some interest as a country to allow companies to invest in with some legal protections (like the diff between patent law vs copyright law)

So you're saying it's more important to safeguard slop outputs than the original work of human beings.
LLM output is not copyrightable, though? So effectively if you pay for it you can do whatever you want from it. That seems perfectly fair and reasonable.
> in general is counter to human progress.

Historically most evidence seems to point to the contrary.

Amongst other things after the printing press was created it was impossible for anyone who was an author to survive from their work unless they were independently wealthy or had rich patrons.

> Alibaba reuses the IP anthropic used to train the model that's more akin to historical Chinese reverse engineering methods and disrespect of IP

Why is this any worse than Anthropic's disrepect of IP? You've apparently drawn a distinction between the two here, but I'm failing to see what it actually is.

Copyright law and IP law is not the same although everyone seem to conflate the two.

Search engines for example historically ignored copyright law by copying excerpts or serving other site images, it doesn't mean someone copying Google's code has some moral frepass

> Copyright law and IP law is not the same although everyone seem to conflate the two.

Copyright law is a subset of IP law. What IP is being infringed upon here?

> Search engines for example historically ignored copyright law by copying excerpts or serving other site images

Excerpts are often considered fair use, but it depends on country.

> it doesn't mean someone copying Google's code has some moral frepass

Nobody copied Anthropic's code. They used it's output to train another model. At most they violated some terms of service.

Did they maybe abuse Anthropic's subsidised pricing? Sure. But that's what happens in a free market if you sell below cost.

> Search engines for example historically ignored copyright law by copying excerpts or serving other site images, it doesn't mean someone copying Google's code has some moral frepass

Not sure that's the best example as they lost that battle and had to pay, eventually it's been codified in law in most countries.

'Issue' for who?
I wish people would stop using Anthropics incorrect use of the term distill. They don’t share logits so you can’t distill. You can generate training data, which doesn’t sound nearly so scary.
Why do you need logits to distill? Those are at least tokenizer-dependent, and different models use different tokenizers.
Distillation is a specific named technique from a pretty famous paper "Distilling the Knowledge in a Neural Network," by Hinton and others.

What makes it different from just training on any data is that you train the student model on "soft targets" that includes the full output distribution (logits) from the teacher model. Regular training uses one-hot targets and penalizes anything else; distillation will partly reward a student for getting in the distribution. This teaches the student to think like the teacher, not just imitate it.

Generating training data is not distillation, not in the technical sense, and I dislike Anthropic undermining the nomenclature to set a narrative.

> reuses the IP anthropic used to train the model

> disrespect of IP

Nobody other than Anthropic cares.