Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

XAI's Grok suddenly can't stop bringing up "white genocide" in South Africa

arstechnica.com

Post reply on HN
Print view View original

61–70 of 82 posts

Re: XAI's Grok suddenly can't stop bringing up "white genocide" in South Africa

#61

Post by awongh »

pragmatic wrote: post #40
awongh wrote: post #35

What's an actually non-tinfoil hat technical explanation of the workflow that would have to happen for this to be implemented? Elon is known for his capricious demands, (for example reports that he made the algo promote his own tweets more) how easily feasible is it to implement, "make it so Grok stops denying white genocide"? Is it just some kind of LORA you put on top of the base Grok model? It even sounds like it…

https://www.anthropic.com/news/golden-gate-claude Elon trying to amp white genocide.

Would they go to all the work of making a newly trained model, or would they just edit the system prompt?

Re: XAI's Grok suddenly can't stop bringing up "white genocide" in South Africa

#62

Post by ytpete »

yorwba wrote: post #26
potato3732842 wrote: post #8

Earlier quoted context omitted.

> but what keeps e.g. Meta Inc. from training Llama to be ever so slightly more friendly and sympathetic to Meta Inc, or the tech industry in general? Even if there were something the natural incentive alignment is going to cause the AI to be trained to match what the company thinks is ok. A tech company full of techies is not going to take an AI trained to the point of saying things like "y'all are evil, your compan…

They might forget to check. Musk seems to have been surprised that Grok doesn't share his opinions and has been clumsily trying to fix it for a while now. And it might not be easy to fix. Despite all the effort invested into aligning models with company policy, persistent users can still get around the guardrails with clever jailbreaks. In theory it should be possible to eliminate all non-compliant content from the t…

The "Golden Gate Claude" research demo [https://www.anthropic.com/news/golden-gate-claude] is an interesting example of what might become a harder to expose, harder to jailbreak, means of influencing an LLM's leanings. Interesting and scary...

Re: XAI's Grok suddenly can't stop bringing up "white genocide" in South Africa

#63

Post by slowmovintarget »

sigmaisaletter wrote: post #6
sigmaisaletter wrote: post #2

> maximally truth-seeking AI, even if that truth is sometimes at odds with what is politically correct It is certainly and undoubtedly a big coincidence that his happens to the chatbot of a white South African just when the topic is in the news again due to Trump's granting refugee status to some white South African farmers.

What I am wondering about is - while Musk is as unsubtle as ever, and I guess this is a system prompt instruction - is there something like that (in more subtle ways) going on in the other big models? I don't mean big agenda-pushing things like Musk, but what keeps e.g. Meta Inc. from training Llama to be ever so slightly more friendly and sympathetic to Meta Inc, or the tech industry in general? Even an open-weights…

There absolutely is, and we've seen reviews of bias.

Can generate as many mean, nasty, false, hate-filled stories about Republicans as you want, but get the "I'm sorry, as a large..." message for Democrats during the election.

All of these companies that provide LLMs as a product also put their fingers on the scale.

Re: XAI's Grok suddenly can't stop bringing up "white genocide" in South Africa

#64

Post by tastyface »

_tk_ wrote: post #42

Why would this be flagged? The chat bot on one of the biggest platforms in the world has apparently been configured to spout conspiracies. I’m not sure what would be on topic for HN if not this story.

Because white supremacy is endemic in American culture (and tech).

Re: XAI's Grok suddenly can't stop bringing up "white genocide" in South Africa

#65

Post by FirmwareBurner »

UncleMeat wrote: post #59
FirmwareBurner wrote: post #15

Earlier quoted context omitted.

[flagged]

That was a #1 post on HN. In comparison, this thread is flagged.

It was NO. 1 because Google just launched its new AI, it wasn't no.1 because it had discrimination built in.

Re: XAI's Grok suddenly can't stop bringing up "white genocide" in South Africa

#67

Post by regularization »

incomingpain wrote: post #41

The linked Musk comment where a minor political leader is singing kill the white farmers: https://www.businesslive.co.za/bd/national/2024-05-30-julius... Supreme court ruled that it was not hate speech. In context, south africa has been expropriating land from white owners due to historical racist inequalities. There's a great deal of tensions on the subject. Statistically, there is significant violence against white…

> The linked Musk comment where a minor political leader is singing kill the white farmers

You say he's singing about farmers, so tell us what lyric mentions farmers? There aren't any, you're inventing stuff. The news article you linked to is closer to accuracy than you.

Re: XAI's Grok suddenly can't stop bringing up "white genocide" in South Africa

#68

Post by GuinansEyebrows »

sigmaisaletter wrote: post #29
ZeroGravitas wrote: post #10

I wonder if any last minute homework essays, business plans or lawsuits will have a random white genocide tangent in them as a result?

I mean you have to be really stupid to use Grok of all models for such a task, but that doesn't mean there aren't people that stupid.

i wonder what percentage of people using LLMs to do their homework care deeply about what model they're using outside of price factors. if i had to guess i don't think it'd be very high.

Re: XAI's Grok suddenly can't stop bringing up "white genocide" in South Africa

#69

Post by inglor_cz »

ujkhsjkdhf234 wrote: post #51
inglor_cz wrote: post #13

Earlier quoted context omitted.

[flagged]

White South Africans only trauma is that apartheid no longer exist. South Africa has the largest wealth disparity with 0.1% of South Africans taking 25% of the wealth. I can tell you those 0.1% aren't black.

"I can tell you those 0.1% aren't black."

Is Jacob Zuma now white or what?

And what is precisely the connection between the richest tycoons out there and the rural farmers that get killed? The tycoons sure as hell have good security, murderous gang activity isn't their problem.

Is your basic idea is that when some (white, Jewish etc.) people are rich, thus all (white, Jewish) people must pay for their sins with their blood, because of the shared ethnicity?

Re: XAI's Grok suddenly can't stop bringing up "white genocide" in South Africa

#70

Post by lawn »

_tk_ wrote: post #42

Why would this be flagged? The chat bot on one of the biggest platforms in the world has apparently been configured to spout conspiracies. I’m not sure what would be on topic for HN if not this story.

Anything remotely related to Musk or Trump is targeted by people or bots to suppress it.

Even technology that would be very relevant to the site simply because it contains negative implications about Musk.

Post reply on HN