Grok: Elon Musk's AI can now understand pictures and jokes

Grok, the AI chatbot in X, can explain jokes, says Elon Musk. The AI model has been given image analysis capabilities.

listen Print view
A man in front of a computer laughs.

Thanks to Grok, he also got the joke.

(Image: fizkes/Shutterstock.com)

2 min. read

The AI chatbot from xAI can now also understand and analyze images. Grok, which is also the name of the underlying AI model, is available to paying subscribers of X. Images can now be uploaded there, which can be explained by the AI. According to X and xAI boss Elon Musk, this even works with jokes, which can be displayed in comic form or other images – including text.

Humor is known to be very important to Musk, even when it comes to AI. Grok has hardly been given any guidelines, which means that the chatbot and integrated image generator can be used to create almost any text and image imaginable. Other providers do not allow photorealistic images, for example, or refuse to allow images that show violence and sexual depictions as well as images that show politicians in situations that could influence voting decisions. Musk believes that such images should also be allowed and taken with humor.

The example that Musk posts on X also shows a joke about dead scientists. He has also linked the chatbot's explanation of why the comic strip is funny. However, Musk also says that the function is still in its early stages but will improve quickly.

Videos by heise

It was only in August that xAI released Grok-2, the integrated image generator based on the Flux.1 model from AI provider Black Forest Labs. This was accompanied by the announcement that Grok would soon be multimodal, meaning it would be able to process text as well as images and, if necessary, audio.

At X, someone immediately criticized that Grok would still not be able to process all file formats, such as PDFs. Musk replied that this would soon change. "We can do in months what others have taken years," he writes in his famously modest manner. It is unclear which years he is referring to, as the other major AI providers have also significantly expanded their models to include functions and multimodality within just a few months. OpenAI is now even talking about an omnimodel, which they have launched on the market with GPT-4o.

(emw)

Don't miss any news – follow us on Facebook, LinkedIn or Mastodon.

This article was originally published in German. It was translated with technical assistance and editorially reviewed before publication.