(Photo Illustration by Jaque Silva/SOPA Images/LightRocket via Getty Images)
Premium X users got access to a new version of Grok AI today, Grok-2, and have discovered it will readily create images of famous figures.
Some of the examples spreading on social media include Mario drinking beer on a beach, Elmo reporting on the 9/11 attacks, and Donald Trump holding Vice President Kamala Harris's pregnant belly. The Trump images seem to be particularly realistic, while those of Harris are not. But they all have one thing in common: outlandish visuals that make you think, "Is that real?"
Other AI image generators, such as OpenAI's Dall-E and Midjourney, no longer accept requests to create images of famous figures. Midjourney banned it ahead of the 2024 elections, citing an inability to implement proper content moderation controls (though the CEO admitted creating these types of images is "fun.")
Google also programmed its Gemini chatbot to refuse election-related inquiries due to risk of hallucinations and spreading false information.
When Elon Musk launched Grok on X in November 2023, he positioned it as renegade alternative to competitors that would "answer spicy questions that are rejected by most other AI systems." But when combined with the platform's wide reach, these images could inadvertently spread misinformation, violate copyright, and ruin reputations in the name of entertainment.
Grok has already been found to spread misinformation on several occasions through its text-based chatbot answers. Its image-generation capabilities could offer another avenue for doing so without the proper controls.