Why Google's AI can't spell Google (or anything else) | TechCrunch

Google’s AI Overviews in Search show multiple spelling mistakes, including incorrect counts of letters in words and misspellings in names. The errors include claims about the number of “P”s in “Google,” incorrect letter counts in “poop” and “journalism,” and a misspelling of the last name of the U.S. president. Similar problems have appeared in earlier AI Overviews, including citations of satirical content. Some issues have been patched, such as a search for “disregard” returning an assistant-style reply instead of a definition. The underlying cause is that large language models do not reliably understand spelling and do not treat sentences as structured units of words and letters.

"How many Ps are in Google? According to Google, there are two. There's also is also "exactly 1 'r' in the word 'poop'," Google's AI Overview says, as well as two 'd's in the word journalism, yet spelled it: j-o-u-r-n-a-d-i-s-m. Google did at least identify that there is one P in the last name of the U.S. president, but spelled it as t-r-p-u-m."

"You didn't need to be a prophet to predict that Google's AI-forward Search overhaul was going to go over poorly. We've done this before. The first time Google added AI Overviews to Search, the feature ended up citing satirical posts from The Onion and Reddit, advising people to eat rocks and put glue on their pizza. This time around, as Google doubles down on its commitment to make generative AI the centerpiece of its 29-year-old flagship product, it's not surprising to see it stumble."

"Google's AI overview woes reach beyond silly spelling mistakes though. Google already patched an issue from last week in which searching the word " disregard " would yield what looked like a dictionary definition of the word, only the definition was shown as, "Understood. Let me know whenever you have a new prompt or question!" But these spelling errors have remained amusing because they're so difficult to quash."

"LLMs, the kind of artificial intelligence that powers chatbots and other text-generators, are not built to understand spelling. It's been a running joke for years that whenever a company unveils a new AI model, you should ask it how many 'r's are in the word strawberry. These AI models - which can code an app in seconds, or solve problems that have stumped mathematicians for decades - are about as good as a kindergartener at spelling."

Read at TechCrunch