Hacker News Reader

Comment by inigyou | Hacker News Reader

Grok is the #1 uncensored easily-available model, and it's also tightly integrated with Twitter.

Is uncensored a selling point? What do people use uncensored Grok for (like, real use cases) that they can't or won't use other LLMs for? Literally the only thing I can think of is generating bad porn of unconsenting people.

I mean absolutely read any thread about Fabel and it's fill with people complaining about how it instantly downgrades or refuses if anything has CVE in the name.

Other then that there is the whole alignment issue. Models that are 'nerfed' in just about any manner tend to exhibit reduced performance is seemingly unrelated areas.

That said Grok doesn't appear to be close enough to the frontier for that to matter. Maybe if they catch up it will.

Thanks for the reply. I only use local/open models, and don't use them for security work, so I don't have much exposure to frontier alignment and Fable/Mythos stuff beyond what I read from others here on HN.

I don't really have a use for a model that thinks "how many people are in this photo?" is a political question.

I don't know what you're referencing.

Some have mentioned legal work. OpenAI and Anhropic models would refuse to work on cases where something immoral happened.

Thanks, that's does sound like a tricky edge case for the other "censored" companies.

> What do people use uncensored Grok for (like, real use cases) that they can't or won't use other LLMs for? Literally the only thing I can think of is generating bad porn of unconsenting people

untrue. There's a full thread about it: https://news.ycombinator.com/item?id=48837162 - but as much as I love Claude products, nothing's more aggravating than it refusing to help me diagnose a stack trace because it "violates Anthopic policy".

Security stuff. Poking holes in my app and suggest fixes.

I don't remember online discourses on filter avoidance for Grok to be any different from typical ones, except that it allegedly have tendency to take porn-biased interpretations of prompts, I think the "uncensored" pitch they had for a while was pure marketing in the end.

A few months ago it was the only model that would make children in pictures naked. I think that's been patched since Elon got sued over it.

Was that actually benchmarked and compared, or was that some British activist group insisting? I thought it was just the latter.

There is/was the official bot on Twitter that you can tag with a prompt, like "@grok put this spacesuit on a horse on a moon", that's not equal to being uncensored.

IIRC there was an IDOR bug around that time that allowed researchers to access all Grok-generated images.

> tightly integrated with Twitter

Not sure that is a good thing.

Uncensored?

Great if you want to make virtual child porn, I guess.

Reply on news.ycombinator.com