News for hackers. Stuff that matters.
Re:Grok answers unrelated queries with long paragraphs about... by rideontime on Wednesday May 14, 2025 @01:57PM (#43987416)
I feel a little less worried about Elon being able to tweak Grok for the benefit of his own propaganda goals now that we can see how blatantly obvious it is when it happens.
Re:Grok answers unrelated queries with long paragraphs about... by observationist on Wednesday May 14, 2025 @02:07PM (#43987495)
Similar things have happened to OpenAI and Claude - context gets leaked from somewhere it's not supposed to. In this case, the white refugees are trending; it's likely context is leaking in from grok checking the users feed and such.

Or you can pretend Elon Musk is a cartoon villain, whatever floats your boat.

Re:Grok answers unrelated queries with long paragraphs about... by rideontime on Wednesday May 14, 2025 @02:17PM (#43987609)
This very specific context? Multiple Grok replies suggest that it's being prompted with a particular image: https://x.com/grok/status/1922671571665310162

e: And since that reply is in the same thread, here's an example of it happening in a completely different one. Not difficult to find these. https://x.com/grok/status/1922682536762958026

Re:Grok answers unrelated queries with long paragraphs about... by burkaman on Wednesday May 14, 2025 @02:29PM (#43987707)
Yeah it really looks like someone added something about South Africa to the system prompt. Just scroll through its latest replies until you see one with an unprompted South Africa discussion, it won't take long: https://xcancel.com/grok/with_replies
Re:Grok answers unrelated queries with long paragraphs about... by amanaplanacanal on Wednesday May 14, 2025 @02:11PM (#43987555)
Nah, not cartoon.
Re:Grok answers unrelated queries with long paragraphs about... by wewtyflakes on Wednesday May 14, 2025 @02:11PM (#43987546)
Did we read the same thing? This seemed reminiscent of https://www.anthropic.com/news/golden-gate-claude, of which was an experiment done on purpose.
Re:Grok answers unrelated queries with long paragraphs about... by EnPissant on Wednesday May 14, 2025 @02:15PM (#43987590)
In addition, the reply doesn't even support Elon Musk's position. Clearly, this is either a bug, responding to a deleted tweet, or something else.
Re:Grok answers unrelated queries with long paragraphs about... by dinfinity on Wednesday May 14, 2025 @02:27PM (#43987685)
Except that it will trigger a lot of people to find that "Kill the Boer" song and will search for "south africa white genocide".

Pretty sure most people won't come out of that with a particularly nuanced view of the situation in South Africa.

Good manipulation is subtle.

Re:Grok answers unrelated queries with long paragraphs about... by Kivern7 on Wednesday May 14, 2025 @02:33PM (#43987764)
Excuse me, are you suggesting that any amount of "nuance" could make these acceptable? Or that people "finding" out about it is a bad thing?
Re:Grok answers unrelated queries with long paragraphs about... by dinfinity on Thursday May 15, 2025 @07:07AM (#43993762)
It's a matter of selection bias, presented as neutrality. Focusing most of the attention on one side of the matter lacks nuance.
Re:Grok answers unrelated queries with long paragraphs about... by EnPissant on Wednesday May 14, 2025 @02:42PM (#43987874)
We must have different definitions of subtle.
Re:Grok answers unrelated queries with long paragraphs about... by dinfinity on Thursday May 15, 2025 @07:09AM (#43993773)
Well, the random placement is not very subtle, but the feigned neutrality will fool most readers.
Re:Grok answers unrelated queries with long paragraphs about... by EnPissant on Thursday May 15, 2025 @01:03PM (#43997013)
Looks pretty neutral to me.
Re:Grok answers unrelated queries with long paragraphs about... by int_19h on Wednesday May 14, 2025 @09:27PM (#43990936)
It doesn't support Musk's position because Grok is smart enough to know when its system prompt is obvious bullshit.
Re:Grok answers unrelated queries with long paragraphs about... by subjectsigma on Wednesday May 14, 2025 @02:13PM (#43987571)
Elon Musk pretty much is a cartoon villain, and refugees are an important topic, but I think that’s almost irrelevant when considering the question at hand, which is whether or not the output from Grok is biased and inflammatory. I believe it is, but endless speculation about why is probably not a good idea when we’re talking about a literal nonsense generator. Nobody fucking understands why LLMs do half the things they do.

I think no matter the cause, users should demand better quality and/or switch to a different model. Or, you know, stop trusting a magical black box to think for them.

Re:Grok answers unrelated queries with long paragraphs about... by jrflowers on Wednesday May 14, 2025 @07:19PM (#43990200)
> Or you can pretend Elon Musk is a cartoon villain

What do you think villains do in cartoons

Re:Grok answers unrelated queries with long paragraphs about... by tastyface on Wednesday May 14, 2025 @08:28PM (#43990606)
This is just a stupid trial run. Eventually, this type of propaganda will become far more subtle and insidious.
Re:Grok answers unrelated queries with long paragraphs about... by empath75 on Wednesday May 14, 2025 @10:38PM (#43991292)
For whatever reason, all the LLMs of a certain size _seem_ to have a very strong sense of right and wrong. (I say "seem", because it's mostly consistent with what a person who had a strong sense of right and wrong would say, but who knows what is really going on inside.)