Aww, look, Grok is cosplaying Jar Jar Binks. That’s so cute, you can almost forget that “Dis violated… maybe US laws on CSAM”.
Oh wait. Jar Jar Binks is the single most hated character in the Star Wars franchise. And that’s saying something since he’s competing with the final 3 episodes.
I would guess Grok’s apology was in response to someone prompting it to write an apology in the voice of Jar Jar. Which doesn’t really make it better, but it’s not like this was a choice it made.
If someone is registered on X, maybe you can see what it was replying to. The profile of the first one in the list (@Magusiak) doesn’t look like a Musk fan.
It was this.
Maybe if Grok takes a heel turn that means musk himself gets to be a face.
That’s probably about the amount of juvenile “logic” behind all of this.
Just to explain what happened; a guy asked Grok to estimate the age of a couple of girls in a photo that I assume was pornographic and AI-generated (by xAI). The account that asked the question has since been suspended. Grok estimated one girl between 12-14 and a second between 14-16.
The questioner then asked what the legal ramifications would be for generating the photo. Grok said it was illegal and cited the US Code and potential punishment for violating the law.
The questioner then asked Grok for a summary of what had happened, and Grok said that it had created that child porn on request.
The questioner asked Grok to apologize after it basically admitted that it broke the law, and it did. Then a different person (not the original, now-banned person) asked Grok to repeat the apology in a Jar Jar voice (as I linked to before).
So this was all meant to highlight a prior situation where Grok created child porn for someone and get Grok to admit it.
Twitter recently added the ability to directly AI edit photos posted there. You click “edit” and you get a Grok prompt. For example you can open a photo and tell Grok to put them on bikinis or lingerie. So, real girls, edited photo.
That makes sense.
So fucked up.
And stupid. If you just wanted it for um…“private viewing”, nobody would know about it. But it is an uproar because the idiots are posting them to Twitter, even as replies to the original post.
Nobody is doing this for private viewing for titillation purposes. The entire point of it is a visible harassment campaign against female users of the site.
We all need to be very, very careful about how we discuss these technologies. The interactive surface — the language modeling — is such a compelling simulation of intelligent communication that we easily misconstrue the nature of the entity behind the language. Not only do these systems not have a state of mind, they are literally stateless. Neither Grok nor any other LLM can “admit” to past transgressions because they have no persistent memory or awareness of those actions and indeed no moral framework by which to recognize the fact that they are transgressions. And yet the illusion is so complete that many, many people are duped into perceiving these interactions as communications from some sort of consciousness. Indeed, we have people on these boards right now who argue that we’re seeing the emergence of a genuinely new kind of mind, instead of just an insanely complex manifestation of refrigerator-magnet poetry.
The prompted Grok output above, and several other Grok responses, triggered a small wave of news articles in which Grok was “quoted” as having “confessed” to the problem, “apologizing” for it, and “promising” that steps are being taken to correct the issue. This is absolute, abject nonsense; it’s nothing but technologically ignorant click-hunting media taking Silicon Valley executives’ relentless promotion of these chatbots as revolutionary AI at face value. The reality is, asking Grok for its reaction to this issue, or anything at all, has no more epistemological value than asking a calculator to apologize for displaying 80085.
I shared the Jar Jar styled quote above not to demonstrate that Grok (or xAI generally) isn’t taking the issue seriously and/or that it’s apologizing sarcastically. I shared the quote to show that we shouldn’t take these outputs seriously. They are entirely meaningless, except insofar as they serve in a meta-semantic sense as evidence of their own meaninglessness.
TLDR: We must be mindful of our choice of language to ensure we are not inappropriately anthropomorphizing these text-prediction systems and polluting our perception of what they are and how they work. Grok did not “admit” anything because Grok cannot admit anything.
I anthropomorphize things all the time because I’m a human being and it’s what we do. It’s neither strange nor dangerous.
I anthropomorphize computers, cars, my dog, just about everything. That’s how humans relate to and understand the world around them. Asking people not to is futile and misguided.
It’s no more dangerous to talk about Grok like it’s a real person than it is to talk about a television show character or book protagonist. We know it’s not real.
I disagree violently. The news stories I mentioned reproduced Grok’s posts and represented them as “quotes” reflecting the system’s “position” on the matter. This is absolutely dangerous because it allows the ignorant media to persist in its failure to understand what they’re reporting on and it conveys a deeply misleading picture to the mainstream public whose grasp on the nature of these technologies is even more off-base. To believe in any way that Grok’s outputs connote an apology or an admission or a position statement, or to use language that implicitly encourages such beliefs, redirects attention away from the human programmers and designers who are actually responsible for creating, implementing, and supporting this technology.
Edit to add: I fully agree that humans naturally and instinctively anthropomorphize inanimate objects, pseudo-sapient organisms, and so on. My point is that in this specific instance we should be extremely cautious and thoughtful about recognizing this behavior in ourselves, because it causes us to misperceive how these chatbots actually function. The illusion of communication is exceptionally powerful, and it almost literally hacks our brains specifically because of that instinct for anthropomorphism.
Agree completely w @Cervaise in how dangerous this is.
Agree completely w @Atamasama that it is pointless to resist the tide. The great unwashed mass of humanity absolutely will buy / have already bought that LLMs are sapient at human, or maybe human-plus levels already. They also routinely believe that Nigerian Princes have money to send them.
This is just a new form of industrial-scale pollution of the mindspace. Much like Faux News.
FTR, I was referring to the human soliciting the form of apology, not Grok, when I said it was done in the most racist way possible.
When I anthropomorphise my cats, or my car, or my computer, it does affect how I think about them. I know they aren’t human, but it’s easy to reason as if they are because of the language I use.
I anthrpomorphise much less plausibly sentient computer programs.
Exactly my thoughts, and well put.
I only have a twitter account because someone in the Philippines must have entered my email address by accident, so I got the email check. So I logged in, secured it, forgot if you can hide it but if so I did so. Never posted.
LSLGuy sent me here from an AI-ish thread. Musk is no longer part of OpenAI, which makes ChatGPT, and I am amazed at it. I have that, and Gemini hooked up to a website, yet am now paying ChatGPT as it degrades to GPT-4 (for like 5 hours) after using it for some hours.
This thing is helping me find a new car, as it can examine car ads and MOT (inspection histories). It’s going to help me with the closing on my Mom’s house in the USA (mostly travel advice). I asked it if it could pass the New York State Bar Exam, and it said not only that, but it could grade everyone else’s and tell what is right and wrong.
Don’t think I’ve posted in a pit thread without any grievance. The Yankees suck!
Sneaking in under the radar and vying with his Mars colony for levels of delusion, Elmo has been saying since November 2025 that he’s going to put 100GW of solar power generation in space per year to power AI datacenters in space. Which is about 25% of the power generation of the entire US. Detailed explanation of why this is entirely absurd:
And another 100gigawatts for the cooling.
Where do I send my check? 
I presume a data center in space would use passive radiative cooling, since that’s the only way you can cool anything in space.