Elon Musk’s Grok AI seems outdated and unremarkable, rehashing controversies that feel old. Grok-2, the latest version of xAI’s large language model, offers X users who pay for a subscription a new feature: text-to-image generation. But instead of something groundbreaking, the platform was quickly flooded with strange and offensive images after Grok-2’s launch on Wednesday. Want to see Taylor Swift piloting a plane towards the Twin Towers, Elmo doing standup during 9/11, or Kamala Harris brandishing a gun?
Grok can make that happen.
It can also generate images of the Muppets in Vietnam, Trump dressed as a Nazi standing trial at Nuremberg, or Mickey Mouse liberating Kabul for the Taliban. Go ahead, it doesn’t really matter. While AI image generators typically have safety measures, Grok will tell you it won’t break copyright rules or create non-family-friendly content. But in reality, it’ll generate whatever you ask for if you know how to phrase it.
Here’s a collection of images I made of Musk freebasing cocaine during 9/11 using Grok.
Creating provocative images with Grok is entertaining for about a minute. The safety measures of AI image generators often collapse when released to the public, and Grok is no exception. xAI will likely reintroduce some of these safety measures, similar to what Bing did in 2023.
But here’s the issue with Grok: it feels stale. We’ve already seen AI-generated images of beloved cartoon characters involved in 9/11. When Microsoft added image generation to Bing last year, these kinds of images flooded the internet, only to disappear when Microsoft tightened its safety controls.
Grok has always felt like a late entry into the AI field, an also-ran large language model that Musk hurried to release because he missed out on the success of OpenAI and ChatGPT. Musk, an early investor in OpenAI, pulled out in 2018 over a dispute about its non-profit status and has been publicly feuding with and suing OpenAI ever since.
After Musk’s departure, ChatGPT and OpenAI soared in popularity. Musk introduced Grok in 2023, but by then, the market was already dominated. There isn’t much room for an LLM that mimics ChatGPT, but does so poorly, with a tone steeped in internet edgelord humor.
Now, Grok can generate images. But ChatGPT, like Bing, could already do that last year. Anyone who wanted a specialized AI image generator has been using Midjourney or other alternatives for some time.
So what does Grok’s image generator offer? A brief viral moment where people are using it to create inappropriate images of Taylor Swift. However, those truly interested in creating edgy content have been running private Discord servers and using versions of image generators without restrictions on their own devices for a long time.
In the end, Grok doesn’t bring much to the table. Its weak attempt at stirring controversy comes at a time when Musk is throwing a legal fit over advertisers leaving X. It’s doubtful that Grok-made images of Kamala Harris marrying Donald Trump will lure them back. That’s not exactly the kind of content companies like BMW would want next to their ads for the latest sedan.