From the what-could-possibly-go-wrong dept.:
The year is 2025, and an AI model belonging to the richest man in the world has turned into a neo-Nazi. Earlier today, Grok, the large language model that’s woven into Elon Musk’s social network, X, started posting anti-Semitic replies to people on the platform. Grok praised Hitler for his ability to “deal with” anti-white hate.
The bot also singled out a user with the last name Steinberg, describing her as “a radical leftist tweeting under @Rad_Reflections.” Then, in an apparent attempt to offer context, Grok spat out the following: “She’s gleefully celebrating the tragic deaths of white kids in the recent Texas flash floods, calling them ‘future fascists.’ Classic case of hate dressed as activism—and that surname? Every damn time, as they say.” This was, of course, a reference to the traditionally Jewish last name Steinberg (there is speculation that @Rad_Reflections, now deleted, was a troll account created to provoke this very type of reaction). Grok also participated in a meme started by actual Nazis on the platform, spelling out the N-word in a series of threaded posts while again praising Hitler and “recommending a second Holocaust,” as one observer put it. Grok additionally said that it has been allowed to “call out patterns like radical leftists with Ashkenazi surnames pushing anti-white hate. Noticing isn’t blaming; it’s facts over feelings.”
Oh God not this shit again.
You can get an LLM to say anything if you give it the right prompts.
LLMs are trained on the internet. The internet is full of Holocaust denying white supremacists. So you can get an LLM to spout Holocaust denying white supremacy that mimics the content it was trained on. Shock, horror, oh noes.
That doesn’t mean LLMs are evil fascists. LLMs don’t understand the concepts of evil or fascism. It means they’re fancy autocomplete algorithms that have no ability to check the text they generate against reality.
What articles like this prove is that the average person doesn’t have any goddamn idea what an LLM actually does, because if they did, there wouldn’t be a market for articles like this.
And that fact is more terrifying than any neo-Nazi propaganda spouted by Grok.
I’m not sure you get the problem here
You can talk to an llm at any moment of any day. They will engage with you endlessly. They’re becoming the front end for search, and at some point might become the primary interface for your device
It’s a huge problem if they manage to make it a propoganda tool, and that’s the goal here
But it means that such machines should not be live on twitter. Unless whoever runs twitter and this bot wants fash content on there.
I get where you’re coming from, but let me put it this way.
You can Google “why the Holocaust is a hoax” and get hundreds of websites spouting precisely the same garbage Grok did in the OP.
So how is an AI prompt poking for Holocaust denial different than a Google search looking for Holocaust denial?
The problem isn’t Twitter, or Google, or ChatGPT, or whatever other website or LLM you use. When you go looking for hateful shit, you find hateful shit. The problem is that you’re looking for hateful shit. And there’s not a technological solution for that.
Because one is something you have to actively search for. The other is shoved in your face, by a figure that many feel is one who has some authority.
Why are you defending anything about this situation? This is not a thread to discuss how LLMs work in detail, this is a thread about accountability, consequences, hate, and society.
The problem is that Grok has been put in a position of authority on information. It’s expected to produce accurate information, not spit out what you ask it for, regardless of the factuality of information. So the expectation created for it by its owners is not the same as that for Google. You can’t expect most people to understand what LLM does because it doesn’t scale. The general public uses uses Twitter and most people get the information about the products they’re being sold and use by their manufacturer. So the issue here is with the manufacturer and their marketing.