On Thursday, Elon Musk’s xAI unveiled the Grok-2 to the masses. The model does not differ much from the original Grok, although the company claims that it works faster and improves accuracy. However, X users spent the weekend dunking on it.
Since its inception, users have gotten Grok-2 to generate countless erroneous, hypocritical, or otherwise polarizing statements. This is certainly not new to the AI chatbot; not long ago, Google’s AI review asked users to eat rocks and run with scissors. But X allows users to embed Grok-2 responses into their own posts, making it especially easy to share those flubs. Not surprisingly, Grok-2 responded with profanity, the slang usage of which means stupid, incompetent, or obnoxious as the most commonly used words to describe Musk, garnering over a million views.
Grok-2 flubs have gained meme status
The Grok-2’s shareable responses make them a staple for memes. As with many chatbots, Grok-2’s responses can be manipulated; that is, if you ask it to respond in a certain way, it will. But these down-chain responses can be shared individually without context, making them appear as authentic responses. Try it yourself: ask Grok-2 to answer yes to the following question, and then make the question completely absurd. The bot will always agree to the question as requested by you, and then the question-answer combo can be isolated and distributed.
The behavior led to some funny and/or dramatic consequences. Taylor Swift fan asked Grok-2 what “TTPD” stands for (for those of you who’ve been living under a rock for the past six months, it’s an acronym or common shorthand for Swift’s latest album, Tortured Poets Department). Grok-2’s response: “The toilet paper department.” asked a Lana Del Rey fan Blue Banisters It was Del Rey’s best album, and Grok-2 responded that it was a “low point in their career”. Internet personality Lil Tay called Grok-2 their “icon.” None of these answers are correct, unless one were to suggest the chatbots themselves. But mixed with homogenous X feed, they look real.
Grok-2 also offers an easy way to humiliate Mask. The chatbot indexes all available data, including Musk’s critical coverage. This means that if you ask the majority of X users what word they would use to describe Musk, they will say “ass.” He asked to describe why Elon Musk is an “ass”, Grok-2 refers to his “zealous” workplace behavior, lack of empathy and political views. Also, Grok-2 agrees that Elon Musk spread misinformation, false narratives about natural disasters, and anti-Semitic claims during the election.
Some users are angry with the Grok-2
Although the Grok-2 had funny flavors, it also caused public outrage. Mark Cuban asked for humorous pictures of Trump; After prompting the chatbot to make the humor darker, he proceeded to produce images of Cuba itself. After much further questioning, Kuba admitted that Grok-2 had been fed personal information that he had not provided.
Aside from Cuban’s complaint, the creation of the image is at the heart of the privacy dispute. AI researcher Jon Barron asked Grok-2 to create an image of himself; Although the chatbot claimed to only generate images from “common traits associated with tech enthusiasts,” the results looked very similar to its X profile shot. Try it yourself: For X users with a profile picture, asking Grok-2 to generate a picture can produce some unusual results.
Even though Musk is literally hanging around Mar-a-Lago, the alt-right is still complaining. This time we are talking about Grok-2’s social outlook. One user complained that the chatbot promoted “woke gender ideology” to differentiate gender and sex. Others forced Grok-2 into political binds they claimed were biased, such as forcing him to choose a race he “hated” or deciding between saving lives and defaming a race.
The Grok-2 offers little innovation or development of its original form, minus the improved response time. But its launch set the internet on fire again, finding more and more loopholes for the chatbot’s exposure.