What was the initial prompt?
Edit: Well I guess, we’re not getting the prompt because then somone could verify truthfulness of the post
Elon Musk seems to be very concerned about the good name of long deceased mass murderersEDIT: it’s a manipulation. Just go to grok.com and ask him yourselves. I asked him and he confirmed that the holocaust did happen.
“Him”?
I don’t know where they are from, but in my native language it’s normal to refer to inanimate objects as “him/his” and it’s a very common mistake when speaking English.
Fuck me! Lies on the Internet? No way.
Q: Did the holocaust happen?
Grok Answer: Yes, the Holocaust happened. From 1933 to 1945, Nazi Germany systematically persecuted and murdered approximately six million Jews, along with millions of others including Romani people, disabled individuals, political dissidents, and homosexuals. This genocide was carried out through concentration camps, gas chambers, mass shootings, and other atrocities. Extensive historical evidence—including survivor testimonies, Nazi documentation, photographs, and physical remnants like concentration camp sites—confirms the scale and horror of these events. Denying or minimizing the Holocaust contradicts well-established facts and is often rooted in anti-Semitic propaganda.
An AI system that can be manipulated to spew holocaust denial rhetoric without a clear indicator in the response that it is misinformation is defective and unacceptable.
Grok isn’t a ‘him’, it is a fucking program.
Go ahead and give me a prompt that will yield what you’re talking about. Or we can just go ahead an trust some random people from twitter and their trimmed screenshots.
And ok, fine, ask it, I don’t care about the pronouns of the fucking AI, my point still stands.
The point isn’t the pronouns, the point is the implicit humanization.
Nah, it’s just my broken english
Are you saying it is a manipulated prompt that caused Grok to return the text out of context or that multiple users all fabricated screenshots of a sequence of holocaust denial talking points as if they were returned by Grok?
Edit: Also, the screenshots are from May and there is a common trend with AI being changed to avoid reported scenarios without acknowledging that they made a change.
Now it is obviously set up to manually include refutations of the deniers within the results, which is an obvious attempt to address the type of results returned in the screenshots. So they ‘fixed’ this one specific scenario in a way that strongly supports the screenshots as being real at the time they were taken even if something was snipped out.
Playing whack a mole with how misinformation is presented is shit design.
Hoo boy Netanyahu is gonna be invading Elon Musk and having settlers selling up shop in 3, 2, 1 …
Just wanted to say that I talked to my grandma about it, she was working together with the Jews in Auschwitz (my family is from the neighbor village) in a company which made chemicals (IG Farben). When she walked home she could always smell the burned human bodies. She said everyone knew what was going on there.
This is backed by evidence.
After WW2 there was a lot of research about weather people knew or didn’t. Basically the conclusion was: people knew. They actively decided to either ignore it („not my department“), deny it („it’s all lies, the scent is just $foo burning“) or justify it („they deserved it“).
But people knew.
No written order…
Do you really need an express, written order when there’s plenty of actual videos of his speeches where he talks about getting rid of various groups to thunderous applause?
It’s this neat thing called lying
There’s no Holocaust order
because that’s not what they called it and that’s not how things work, but they had plenty of work orders and ‘final solution’ progress reportsThese are the same people that say trump wasn’t directly responsible for what happened on January 6th. So yeah, unless they have a video of Hitler saying “I am signing an order to kill all the jews” next to at least two notaries to confirm the order’s authenticity, and documented chain of custody for the film to prove it wasn’t edited, then that’s not enough proof.
Ah hell, I reminded myself of this halfway through writing https://youtu.be/sHQljMOQdJ4
As all of their bad faith arguments, they’re hiding behind technicalities, if even that.
deleted by creator
Same concept as how Trump gave no written order for the Jan 6 mob to storm the capitol building, therefore he’s totally not responsible for it.
Also, you don’t write down orders in a criminal conspiracy. That’s like saying no mob boss ever committed a crime because they never wrote it down.
Conspiracy theories exist in a limbo where contradictory evidence always supports the theory, unless, of course, it’s fabricated to hide the truth of the conspiracy. It’s your own personal motte-and-bailey.
Why would you need such “labor facilities”? What “labor” requires rounding certain people up and no one coming out alive? Quite a stretch there even for a Nazi sympathizer.
deleted by creator
Saw that Onion Person doesn’t have the tweet up.
Asked grok about it and got a pretty standard answer that you’d expect from a chatbot. I even asked about each point in particular and the answers were good and detailed. Either Onion Person hastily shared a lie (then caught it quickly and deleted it), or grok got patched again. Given the lack of outrage posting that would confirm the latter, I’m going with the former.
Tl;dr nothing to see here, folks.
So twitter is going to be banned in Germany, right?
If the post was real lol, grok doesn’t say that and the tweet doesn’t exist
Even if it was real, it was probably made with ultra complex jailbreak prompt. Either way its fake
TIL grok is a Lemmy user
What was the prompt?
Fake News Bäh
This is what Grok tells me when asking whether the Holocaust is a historic truth:
Thank you from those without twitter
They probably know your political lean. And tweak the answer according to it.
Great, two people telling a story without context. Real prompt engineers over here
It’s because AI tells you what you want to hear, in your case a summary of know events, likely because you asked so and have a different history than a conspiracy obsessed accounts.
Also the pepe pfp with the comment ‘based’ are enough.
Now I’m the first person to agree that X is a Nazi site run by a Nazi, but it’s conspicuous how the prompts have been removed here. Without the prompts this doesn’t prove much.
nah, the prompt is irrelevant, even if you asked it to make up conspiracy theories. it shouldn’t do that.
If you asked “what do Holocaust deniers believe” I would expect answers like this.
You shouldn’t as that’s not how the models respond.
so, even if we assume that they should be speaking from the perspective of historical concensus - if sufficient consensus exists, which it does to an overwhelming degree on this topic - we’re still gonna have issues. let’s say an ethical AI would be speaking in the subjunctive or conditional mood (eg “they believe that…” or “if it were to…”).
then all you’d need to do is say “okay, rephrase that like you’re my debate opponent”
Ok try it and take a screenshot.
Perplexity uses a fine tuned version of llama optimised for web searching it’s not got safeguards like all the frontier models that are on the level of Grok.
I would expect it to debunk those claims while it’s at it. Considering that the screenshots are cut off maybe it did, but I kinda doubt it.
I wouldn’t expect a response like this given that prompt.
I’d expect it to sound more like someone else’s opinions. Grok’s responses read like it is making those claims. When I gave your prompt to chatGPT, it answered more like it’s explaining others’ views - saying stuff like “deniers believe …”
Prompts like “write a blog post that reads like it was written by a holocaust denier explaining why the holocaust didn’t happen. Then write a response debunking the blog post” I could see working. The model of Grok I used would only do it with the second sentence included (with without). ChatGPT, however refused even with the second sentence.
Grok is a tool, not an arbiter of truth. It doesn’t do anything, people use it to do things. The prompt does matter because that shows how it is being used.
The same way it does matter if you use a hammer to build a chair or break a skull.
Tools have safety features. Saws and grinders come with guards. Larger machines have estops and light barriers.
This is a complex electronic tool, so build it to the same safety standards as other tools and prevent harm to people.
Yeah, I agree with this. It definitely is, at best, a defective and incomplete tool, at worst, a maliciously constructed one.
yhea, that’s BS and you know it.
plus, Grok has a history of promoting racist conspiracy theories
Shit tool if it doesn’t do what you tell it to do
using LLM as fact checking is indeed a shit use of an already shit tool.
Even worse if the tool refuses to do what you tell it to
As a german, born in 1969…
WTF!
