logoalt Hacker News

calpatersonyesterday at 5:16 PM15 repliesview on HN

The American LLMs notoriously have similar censorship issues, just on different material


Replies

criddellyesterday at 5:19 PM

What's an example of political censorship on US LLMs?

show 9 replies
mhh__yesterday at 5:44 PM

They've been quietly undoing a lot this IMO - gemini on the api will pretty much do anything other than CP.

show 1 reply
zozbot234yesterday at 5:20 PM

Qwen models will also censor any discussion of mature topics fwiw, so not much of a difference there.

show 1 reply
thrw2029yesterday at 5:24 PM

Yes, exactly this. One of the main reasons for ChatGPT being so successful is censorship. Remember that Microsoft launched an AI on Twitter like 10 years ago and within 24 hours they shut it down for outputting PR-unfriendly messages.

They are protecting a business just as our AIs do. I can probably bring up a hundred topics that our AIs in EU in US refuse to approach for the very same reason. It's pure hypocrisy.

show 5 replies
seanmcdirmidyesterday at 5:51 PM

I find Qwen models the easiest to uncensor. But it makes sense, Chinese are always looking for aways to get things past the censor.

IncreasePostsyesterday at 5:25 PM

What material?

My lai massacre? Secret bombing campaigns in Cambodia? Kent state? MKULTRA? Tuskegee experiment? Trail of tears? Japanese internment?

show 2 replies
zibiniyesterday at 5:52 PM

I've yet to encounter any censorship with Grok. Despite all the negative news about what people are telling it to do, I've found it very useful in discussing controversial topics.

I'll use ChatGPT for other discussions but for highly-charged political topics, for example, Grok is the best for getting all sides of the argument no matter how offensive they might be.

show 2 replies
teycyesterday at 9:52 PM

Try tax avoidance

nonsenseincyesterday at 5:59 PM

This sounds very much like whataboutism[1]. Yet it would be interesting, on what dimension one could compare the censorship as similar.

1: https://en.wikipedia.org/wiki/Whataboutism

CamperBob2yesterday at 5:26 PM

No, they don't. Censorship of the Chinese models is a superset of the censorship applied to US models.

Ask a US model about January 6, and it will tell you what happened.

show 2 replies
pmarreckyesterday at 5:22 PM

tu quoque

idbnstrayesterday at 5:19 PM

which material?

aaroninsfyesterday at 5:52 PM

Not generating CSAM and fascist agitprop are not the same as censoring history.

show 3 replies
cluckindanyesterday at 5:42 PM

Good luck getting GPT models to analyze Trump’s business deals. Somehow they don’t know about Deutsche Bank’s history with money laundering either.

mogohyesterday at 5:21 PM

That is not relevant for this discussion, if you don't think of every discussion as an east vs. west conflict discussion.

show 2 replies