There are few things more frustrating in the 21st century than so-called intelligent AI systems with the alleged power to understand human language. As has been demonstrated time and time again, they can’t. Earlier today, I asked Siri to set a reminder to pay my credit card. “Here’s what I found on the web for ‘set a reminder to pay my credit card,’” Siri replied. And no matter how many times I shout my full social security number in public, the AI-powered customer service line for my bank only replies, “Sorry, I didn’t quite get that.” It’s infuriating—but if you’re in the McDonald’s drive-thru line, you might instead find it hilarious.
Fast food restaurants have been tinkering with voice-activated order kiosks for a while now, but McDonald’s has even been enlisting AI at some of its drive-thrus since 2021. At the time, the company’s CEO claimed the technology could take orders with 85% accuracy. But several TikTokers have found themselves in that bottom 15%, and they’ve been sharing the results.
One TikToker alleges that after placing an order for a hash brown, a sweet tea, and a Coke, the AI mistakenly added a Diet Coke to the order after “hearing” someone in the adjacent drive-thru line say that to their own kiosk. When the TikToker told the AI to take the Diet Coke off the order, it did—but then added nine sweet teas to the order (certainly more than any car’s cup holders could handle). Overwhelmed, the customer simply drove off without a caffeine fix.
“I’m going to lose my shit,” says a user in another video as she exasperatedly reiterates her order—in a very similar manner to the way I shout when I’m on the phone with the bank, I might add—and the order screen continues to tack on random items. Over the course of the video, an order for a large water and a cup of ice cream appears to morph into a water, a caramel sundae (“no caramel!” she yells), four ketchup packets, and two pats of butter. As the customer repeats that she just wants water and a cup of ice cream, the bot tacks on... a third pat of butter.
Maybe these bots are smarter than we think, and they’re messing up orders this badly just to fuck with us. If there’s one thing we know about AI from every book, movie, show, or other artwork ever produced about it, it’s that AI is categorically evil.
Obviously, these blunders pose the question: Is it really worth it for these companies to deploy such error-prone AI? According to Forbes, the answer is yes. More accurate AI models are more costly and complex, and typically only marginally more accurate than their next best alternative. So unfortunately, it looks like this flawed vision of the future is here to stay. Hope you like butter.