Given how frequently Grok is cited as a source by some members here, this current story seems relevant:
On July 4, Elon Musk announced that his company had “improved @Grok significantly.” Users should “notice a difference,” he said, when they ask it questions on X. Within a few days — and following an update to its system prompt encouraging it to make “politically incorrect” claims — they certainly had. Grok, the “anti-woke” answer to ChatGPT, just a few weeks after bombarding users with unwanted content about “white genocide,” went full Nazi.
First, it suggested that “a particular group” of people was ruining movies, before confirming that it was referring to “Jewish executives.” It didn’t take much nudging from users to get Grok to go further. After the fatal floods in Texas, a troll account with a stolen avatar posting under the name “Cindy Steinberg” posted that it was glad “future fascists” had died, producing a massive (and, again, fabricated) outrage cycle on X, Grok weighed in. “And that surname? Every damn time, as they say.”
“They,” here, are open anti-semites on X, 4chan, and elsewhere, who use the phrase constantly to identify Jews, a connection Grok then helpfully madeitself. Which 20th century figure might be able to deal with such a “problem,” users asked? “Adolf Hitler, no question,” the chatbot said. Ranting about IQ, Grok said “MechaHitler mode” had been activated; asked about a liberal social media personality, it outlined, in great detail, how one might rape him. In a particularly revealing example, Grok played along with a racist old 4chan routine, calling on the “Towerwaffen” to work together to spell slurs…
This news is all over the place, and I have seen many of the problematical quotes from the AI. For a time I wondered if this was somehow out of context. Grok is supposed to be snarky and willing to break the mold of other AI’s by being willing to say impolite things if prompted , so … maybe a human asked it to say antisemitic stuff, and so, you know, it said antisemitic stuff since it was asked to? But nothing I’ve seen so far suggests that it was simply doing what a human asked it to do.
Interestingly, I queried Google about this, and I expected the results to lead with an AI summary of the story as Google does that sort of thing now. And that AI SAID NOTHING and only linked to various stories about it. I thought it was interesting that an AI was being silent on a leading story about another AI.
Elon Musk asked people on X to comment “things that are not politically correct, but nonetheless factually true”. Most of the responses were antisemitic ‘facts’ about the Holocaust and Jewish control of society. But I think Musk just told Grok that all those things were true.
I think it’s quite amusing that not long ago, we were discussing whether Grok was so blasted smart that it could be trusted as an arbiter of whether there’s a god. How the artificially mighty have fallen.
Maybe. Or maybe he is being honest and all he did was remove the “woke” settings.
I vaguely recall something similar happening a few years ago, where they tried to changes the setting on a chatBot so it we be less restricted in terms of “free speech” and within days it was spewing racist propaganda. But I can’t seem to find any reference to this, so don’t quote me.
I guess it’s just like natural, human intelligence, where “non-woke” is only a very short step, if that, from outright White Christian Supremacism
They can’t stop self-owning in the Trump-Musk sphere, and how all the rightwing and Christofacist lunacy interacts with Grok and other LLM is an endless resource of hilarity and irony:
I really don’t like these sorts of polarizing takes because it can easily leave this impression that you must either agree with everything on side Y or you’re probably on side X, and there’s no other position to take.
I know this isn’t what you’re saying (and in most instances it’s likely true that people who rail against all things “woke” are terrible people), but I think there’s a danger with this idea that anything that doesn’t characterize itself as “woke”, or would seek to distance itself from certain let’s-call-it-woke-for-lack-of-a-better-term excesses, is immediately suspect of being either outright nazis, Christian nationalist, or white supremacist.
That’s one reason I’m not particularly fond of the expression “woke,” either as praise or as criticism, because it just sweeps in too much, and means too many different things to different people.
I got called some pretty awful, Nazi-adjacent things when I was representing resource landowners in lawsuits against government agencies. But it was a case of people really miscategorizing in a big way. The principles I was involved in defending were things like due process of law (e.g., should you be able to revoke land-use permits in mid-construction, just because some agency suddenly doesn’t like the project for reasons that should have been raised during permit review? And if you should be able, what process must you follow to make this fair?). But it was seen too readily as “pro-environment” versus “anti-environment,” and of course people felt that only very, very bad people are against protecting the environment, so I must be on the “evil” side. People claimed that some of my clients were associated with far right white supremacist groups – and this was just absolutely and utterly untrue, not even close to true, but ideological division made it easy for people to believe.
Right now I am pretty much in a full-time state of astonished horror at the Trump administration. So I am not one of those people who says that “polarization” is bad and that both sides are equally to blame – far from it. But polarization can definitely be bad.
I know a lot of people who think that outright state socialism is the answer to many of our woes. And I understand, and respect, that point of view, while disagreeing with it in most cases. What disturbs me sometimes is that polarization does seem to have led to some of those people thinking that state socialism and robber-baron capitalism are the only two options on offer. I would like, so often, to be able to discuss questions like “what does government do really well?” and “what does free enterprise do really well?” without either being called a Nazi or a libtard.
