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.

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 model can't be easily inspected, so this is likely to remain undetected.

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 company is evil, your industry is evil" and push it to prod.

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 training data, but that would most likely entail running all training data through an LLM, which would make the training process about twice as expensive.

So, in practice, companies have been releasing models that they do not have full control over.

So, for example, if a model was trained with no references to the Tiananmen Square massacre, I could see it just synthesizing commonalities between other massacres and inventing a new, worse Tiananmen Square Massacre. "That's not a thing that ever happened" isn't something most AIs are particularly good at saying.

I.e. even if you create an explicit Tiananmen Square massacre-shaped hole in your training data... your other training data implicitly includes knowledge of the Tiananmen Square massacre, so might leak it in subtle ways.

E.g. how there are many posts that reference June 4, 1989 in Beijing with negative and/or horrified tones?

Which at scale, an LLM might then rematerialize into existence.

More likely SOTA censorship focuses on levels above base models in the input/output flow (even if that means running cut-down censoring models on top of base models for every query).

Would be fascinated to know what's currently being used for Chinese audiences, given the consequences of a non-compliant model are more severe.

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.

The big labs do have evals for sensitive topics to make sure it demurs from weighing on, say, Mark Zuckerberg as a person

Zuckerberg doesn’t have a control panel where he can move sliders all by himself, any change in weight on the algorithm has to be implemented by a whole bunch of people, any of whom could leak to the press.

It’s not guaranteed it would happen by any means but it’s definitely something that would factor into a decision. Broadly I agree with you though, normally I’d say “extraordinary claims require extraordinary evidence” but I’m increasingly convinced the extraordinary claim here would be that they aren’t manipulating things to benefit themselves in some way or another.

They have their own DEI, affirmative action and cultural sensitivity teams who move the bias sliders based on their political viewpoints and on what management tells them depending which of the political groups they need to pander to: "Let's move the slider to the left to make sure Trump doesn't win; Oh shit Trump won, quick, move the slider to the right".

>any of whom could leak to the press

That's why they sign NDAs.

banks would ask international clients the origin of the money. but not if you are opening an account under refugee status. and then they only have to pay us tax on further income, not on fortune. all that money selling black market gems to russians will be squeak clean.

it's not just something to virtue signal to their bible belt electorate. they probably sold lot of trump coins for this deal.

I would expect Ukrainian AI to talk extensively about the Holodomor, Israeli AI to talk extensively about the Holocaust, Irish AI to talk extensively about the potato famine etc.

Of course the amount of "extensive" will vary. Musk is not a subtle person by any means, and neither is Grok.

I don't think this is anyone's idea, on the entire planet. Yes truly, 8 billion people and I'm not sure a single one thinks this way.

I would assume it's people who pay for Twitter pro or whatever it's called getting Grok as part of the bundle and not wanting to sign up for another subscription.

Another explanation is that there's another discussion. At least on the comments for that discussion the explanation is that this discussion exists...

https://news.ycombinator.com/item?id=43987266

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

Are you saying it’s crappy to be white in South Africa even if you’re not a poor farmer?

It's a bit like what happened in Zimbabwe that devastated that country 20 years ago when all the white farmers were kicked out, but in slower motion.

Land reform to repair the inequality from the previous apartheid state isn't exactly "kicked out". Zimbabwe became Zimbabwe from Rhodesia via a civil war that wrestled control away from the minority white government.

In South Africa today, 75% of the privately owned land is held by about 8% of the population(white folks), who were given legal preference multiple times in history to own all that land.

Even if you remove apartheid, you still have a similar situation to antebellum US where ex slaves were working on the same old plantations, now as share croppers. Of course in SA we're not talking about ex slaves, but you're maintaining the status quo one way or another.

Land reform is required if your country ever wants to know peace. The US never finished reconstruction and didn't redistribute those plantations and other ill gotten gains, and as a result is still struggling to move past slavery.

I don't know too much about Zimbabwe. From what I read about 4,000 white farmers lost house and land, and now 40 years later they are giving compensation. Why can't you say "kicked out" for that?

I do lack the knowledge to say if it was justified or not. But I consider that a kick on the way out.

Or do you mean that what happens in SA is not to be considered getting kicked out?

https://en.wikipedia.org/wiki/Black_Economic_Empowerment

So it may have unintended consequences but it seems to be as good a solution as the US could come up with.

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 could be as simple as changing the text of the system prompt.

I wonder how often this happens (to any big model) without anyone noticing.

Elon trying to amp white genocide.

[0] https://news.ycombinator.com/item?id=39465250

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.

Are you suggesting that most of those murdered are white farmers?

The "Third Quarter 2024/2025 Crime Stats Presentation" [pdf] from South African Police Service Crime stats [0] show between: October 2024 to December 2024 there were 6953 Murders in RSA.[pg 8] and of those there were 12 victims linked to the "farm/small holding" community [pg 32]. 5 farm dwellers, 4 employees, 1 Farmer, 1 Security, & 1 not specified.

12 people linked to farms (but only 1 Farmer murdered) in the last quarter of 2024 in a place with some of the highest murder rates in the world -- does Not white farmer genocide make, nor is it "significant violence against white farmers". There were 68 total "contact crimes" committed in Rural communities; and 187,892 "contact crimes" for RSA in that time.

Contact crimes being violent crimes (and not counting sexual offenses or robbery)

[0] https://www.saps.gov.za/services/crimestats.php

https://www.ushmm.org/m/pdfs/20110210-what-is-genocide.pdf

Especially this one can be applied to many things:

c. Deliberately inflicting on the group conditions of life calculated to bring about its physical destruction in whole or in part.

In this situation, it's specifically about farmers(large scale private landowners). The nature of them being targeted isn't purely because they're white, but that they hold 3/4 of the privately owned land in SA and likely wield that privilege and power.