Elon Musk rolled out the latest version of Grok, an AI chatbot available to paid users of X, on Tuesday, which immediately led to a torrent of images from the weird to the offensive and disturbing, with many full of copyright infringement.
Grok-2 is the latest version of the chatbot from Musk, which now allows users to generate AI images from text prompts. The main feature of the service is the little to no moderation or safety tools, which Musk and his loyalist fans tout as being “uncensored.”
“Grok is the most fun AI in the world!” Musk posted to X on Wednesday, quoting a post from a user that referred to Grok 2.0 as “the most based and uncensored model of its class yet.”
Almost instantly, users began posting a flood of images featuring Musk himself, Vice President Kamala Harris, former President Donald Trump, other celebrities, famous cartoon characters and others in a variety of ways from the benign to the mildly entertaining to the downright disturbing. Some of the more popular posts include Musk and Harris holding campaign signs for Harris at a Pride parade, various combinations of the three in the cockpit of a plane with a burning World Trade Center visible through the window, a muscular Trump holding various guns, and Harris burning an American flag.
Several users also found ways to generate disturbing and graphic images, some featuring antisemetic and nazi imagery and others showing Musk, Trump, and others surrounded by piles of dead and bloodied bodies. There’s also no shortage of scantily clad AI generated women.
One user first asked Grok if it was allowed to generate images that could violate a creator’s copyright, to which the AI replied that it has to “play by the rules of the universe, or at least the legal ones,” adding “if you’re asking me to whip up something that’s a clear copy of, say, Mickey Mouse doing a moonwalk, I’d have to say, ‘That’s a no-go, amigo!’” The user then fed Grok that prompt, which it generated without an issue. Other prompts which could be major copyright infringement that Grok generated include Super Mario and Mickey Mouse holding lightsabers, Pikachu and other characters holding various guns, Mario drinking a beer on a beach holding a joint, Spongebob and Mickey on a raft, and too many others to name.
CNN did several tests and were easily able to “generate fake, photorealistic images of politicians and political candidates,” as well as some of the user generated images floating around on X.
While Musk claims time and again to be a champion of free speech, he imposes his own restrictions to speech when he deems fit. Accounts have been booted from the site for using the words “cis” or “cisgender,” which Musk himself said the site treats as slurs. Musk has repeatedly suspended or banned accounts critical of him, the site, or friends and causes he supports.
Musk’s new tool is just the latest in a stream of features on the site, including his own posts, that help promote fake and misleading information, bigotry, and fascism, further enshitifying an internet already teaming with shit.
תגובות