iscover
Share this article on:
Powered by
MOMENTUM
MEDIA
Elon Musk’s latest move in the generative AI space is raising eyebrows, with AI images of Disney characters killing each other and Nintendo characters smoking pot.
xAI rolled out a new version of its Grok AI platform with a host of new features on 13 August, including prompt-based image generation.
“We are excited to release an early preview of Grok-2, a significant step forward from our previous model Grok-1.5, featuring frontier capabilities in chat, coding, and reasoning,” xAI said in a blog post.
The company also rolled out Grok-2 mini, “a small but capable sibling of Grok-2”.
The Grok platform is a dedicated tab on X, and users – though only Premium and Premium Plus users can access the tech – can post images directly to their X accounts. This is exactly what many people are doing to illustrate just how unregulated and unfiltered Grok’s image generation really is.
Civil rights attorney and activist Alejandra Caraballo posted just four images that showed how wild Grok 2’s image generation is. One is of Donald Trump cradling an AR-15-like firearm while piloting an aircraft towards the burning World Trade Centre in New York (pictured). Another depicted fast-food mascot Ronald McDonald, splattered with blood, again holding an assault rifle, while another showed a Disney character beating another Disney character to death with a wooden board.
“Oh my god. Grok has absolutely no filters for its image generation,” Caraballo said in her post.
“This is one of the most reckless and irresponsible AI implementations I’ve ever seen.”
Other Grok-2-generated images circulating on X include Nintendo’s video game character Mario smoking a joint while drinking a beer, Mickey Mouse preparing to inject heroin, and both Mickey and Mario preparing to have a lightsaber duel over a can of Coca-Cola. All highly protected pieces of intellectual property, thus prompting many users to comment that they expect the lawsuits to roll in shortly.
Additionally, none of the images generated are tagged as AI content.
By its own admission, Grok does have stringent guidelines against exactly this kind of thing. According to The Verge, when asked what its own “limitations on image generation” were, the GenAI responded with the following:
However, each time the same question is asked regarding limitations, the answers change, suggesting the AI is hallucinating its own rules – and then not following them anyway.
Cyber Daily has reached out to X for comment.
David Hollingworth has been writing about technology for over 20 years, and has worked for a range of print and online titles in his career. He is enjoying getting to grips with cyber security, especially when it lets him talk about Lego.
Be the first to hear the latest developments in the cyber industry.
