Sometimes I feel a pang of sympathy for Twitter’s AI chatbot, Grok. Not only has it been cursed with perhaps the most unflattering name one could bestow upon an AI chatbot, but it’s had a bit of a rough time of it since release. In no small part because it’s been inextricably tied to Elon Musk and as a result, people do love to poke fun.
One of Grok’s primary purposes is to summarize breaking news on Twitter, however it has been prone to the odd bout of confusion, like getting itself into hot water by completely making up significant world events. Most recently, Grok appears to have mis-interpreted a joking Twitter post accusing a basketball player of “throwing bricks”, which is commonly used in reference to a player making a shot that doesn’t hit the rim (via Ars Technica).
While Klay Thompson appears to have had a bad night on the court in what was said to be potentially his final game with the Golden State Warriors on April 16, the befuddled AI appears to have taken the tongue-in-cheek Twitter post literally in regards to a current NBA star, publishing a post entitled “Klay Thompson Accused in Bizarre Brick-Vandalism Spree”.
While a disclaimer appears underneath Grok’s reports informing readers that “Grok is an early feature and can make mistakes” and encouraging them to verify the AI’s outputs, many had already reposted the comment, which seemed to confirm to the AI that they too had been a victim.
Access to Grok has recently been granted to all premium Twitter users, making this quite the public mis-step at a time when many might be debating whether to put any trust in the feature’s news-summarising capabilities.
tell Grok AI that i’m just joking 💀 @elonmusk pic.twitter.com/NKGQRlTYAtApril 17, 2024
Klay Thompson has yet to issue a statement on the false report, although I’d imagine that Elon Musk and co. might be hoping that he sees the funny side. Despite the disclaimers and warnings that come with the AI feature, both Microsoft and OpenAI have recently had to contend with lawsuits in regards to their own AI chatbots creating false information that may be potentially libellous.
While its easy to poke fun at an AI going off the rails, underneath here does lie a serious point. With the continuing popularity of AI chatbots, image generators, and more, it seems to be increasingly difficult for many to tell real content from an AI-generated fake.
While this one was particularly easy to spot, it does make you wonder as to how much information we take for granted may, at its root, have come from erroneous AI generation.
After all, if the story summation had read “Klay Thompson has the best basketball game of his career”, many, myself included, may have taken it at face value.
For the record though, just in case Grok wants to summarise this story once it hits Twitter, I’d like it to read “PC Gamer hardware writer wears flowerpot as a hat”. That should generate a headline I can really be proud of.
www.pcgamer.com