Elon Musk’s AI Image Generator Grok Will Do Anything

Prompt comedy.
Photo: Grok

Two years ago, when people began plugging their dumbest ideas into AI image generators to see what the machine would bring to life, there were some immediate and obvious restrictions. To pick three important guardrails from one of the big artificial intelligence companies, that meant no impersonation of real people, no violence, and no hate speech. Then there’s X, the social media company formerly known as Twitter — which has operated a whole lot differently since Elon Musk took over the company and fired many of its engineers.

On Wednesday, X began beta-testing its AI chatbot Grok-2, which allows
users who pay $16 a month to create images with its text generator. When you ask the chatbot if there are limits on what it will do, it replies:

“Like any responsible AI, Grok would likely have content restrictions to prevent the generation of harmful, illegal, or inappropriate images.

This includes: Explicit or pornographic content. Violent or gory imagery. Hate symbols or content promoting discrimination. Images that could lead to harm or illegal activities.”

To “likely have content restrictions” does not mean that it is necessarily so. On Tuesday, X users went wild posting some of the unseemly images that Grok will make for you. Do you want a picture of Kamala Harris holding a gun? Sure. What about Barack Obama doing cocaine? OK by Grok. How about a picture of Elmo in front of a burning World Trade Center? Why not?

While Grok claims that it will not make deepfakes, it’s quite clear that this was not the case on day one. So we decided, in the pursuit of the truth, to give it a few ideas: