Hacker News new | ask | show | jobs
by tossandthrow 538 days ago
I am not aware of any non western models that are not under censorship.

Ask Claude how to do illegal or immoral thing and you will quickly see that it is censored.

I didn't mean to problematize censorship. Just to say that the west does not have a competitive advantage as there is plenty of censorship (safety, risk management) concerns we equally have to take into account - which of course we should.

2 comments

I asked an LLM to implement a gender guessing library for python, and it outright refused saying it was a safety issue.

It's not just an illegal or immoral thing, it's broad strokes to potentially catch illegal or immoral things, by certain people who decide what those morals are.

Trying to equate government mandated censorship to private company policy censorship is a wholly dishonest sleight of hand.
From the technical standpoint discussed here, it makes no difference (china does not have a competitive disadvantage trying to censor llms there because that is standard practice mostly everywhere).
Both in the EU and the US there is plenty of regulation that mandates these types of censorship - and with reason.

In the US there is 18 U.S.C. § 842(p).

In the EU there is the entire AI Act.

But I am sure you can yourself chat your way through to figure out what legislation companies like OpenAI and Anthropic are under.

18 U.S.C. § 842(p). criminalizes bomb instructions when taught with the intent of committing crimes.

TM 31-210 Improvised Munitions Handbook is readily available.

Yep, anthropic has to comply with that.
Is any of these equivalent in nature to, for example, censoring information about Tiananmen square events?
It's possible that China censors info about Tiananmen square because so much of what was published came from Western news orgs - and the West has form for using the "news" to attack other nations. Another example might be the supposed "genocide" of the Uyghur people - the MSM pushed the genocide narrative hard, while radicalising, funding and arming Uyghur Islamic extremists, so they could control the narrative. And of course, it largely worked.
This is more a political discourse that a business or technical one.

You sure can establish that there is a qualitative difference on the type of censorship carried out - congrats.

The main point I spelled out is that there is no comparative advantage (technical or business wise) on working on these products in the west as you have to implement and operationalize the same amount of censorship / safety.

> equating [censorship] to [censorship] is dishonest

Someone who would use this obvious of a red herring is dishonest. The point was not that the censorship is identical, but that the effect of censorship is in both cases to lobotomize the models.

Why on earth would it be better? Trillion dollar corpos in the turbo-capitalist West are already far more powerful than most states.