MENLO PARK, CA—After a week of headlines touting artificial intelligence’s latest triumphs over human intelligence in chess, medicine, and software development, one skeptical Facebook commenter, Jack Wilkinson, 43, has stepped forward to demand an even greater test: spelling the word “restaurant” without autocorrect assistance.
“Look, I don’t care if your fancy computer can find a cure for cancer or write a program that runs the stock market. If it can’t spell ‘resteraunt’ right the first time, it ain’t smarter than me,” wrote Wilkinson in a comment thread under a news article about OpenAI’s latest breakthrough.
AI researchers, who have spent years training machine-learning models to surpass human capabilities in logic, strategy, and complex problem-solving, were reportedly taken aback by this new standard for intelligence.
“Honestly, we never thought to program that in,” admitted Dr. Emily Zhao, an AI researcher at MIT. “We figured if AI could perform 100 quadrillion calculations per second, it could probably handle basic spelling. But Jack raises an important point: Can intelligence even be real if it doesn’t struggle with common English words like ‘definitely’ or ‘bureaucracy’?”
In a dramatic response, OpenAI’s latest model, GPT-X, was put to the test in a livestreamed spelling challenge. When prompted to spell “restaurant,” the AI correctly spelled it in under 0.02 seconds. However, Jack remained unimpressed.
“Yeah, but it probably looked it up first,” he scoffed. “I wanna see it spell it from memory, like a real person who just stares at the word and guesses until it feels right.”
In an effort to meet Wilkinson’s rigorous intellectual standards, scientists at Google DeepMind have now begun developing an AI model that will misspell words three different ways before arriving at the correct answer, much like a typical human. They are also reportedly working on a neural net that insists it knows the spelling, argues with autocorrect, and then shamefully googles it when no one is looking.
Meanwhile, Jack has moved on to more pressing concerns, questioning why AI should be trusted to drive self-driving cars if it has never personally experienced road rage. “If a robot ain’t slammed its fist on the dashboard screaming ‘COME ON, GO ALREADY!’ at a green light, then I don’t trust it to drive,” he wrote in a follow-up post.
AI developers remain hopeful that their technology will soon meet Jack’s high standards. However, sources indicate that Jack himself remains unable to spell “definitely” without assistance.
English
Leave a Reply