Hey ChatGPT, are you talking to me?
If you find that Siri or Alexa often misunderstands your requests, your accent may be to blame.
A new study finds that the unique intonations of New York City, New Jersey, and Long Island residents are some of the hardest for artificial intelligence technology to decipher.
To no one’s surprise, Southern accents top the list of the most difficult accents for AI speech recognition programs to understand, according to a language resource site. Guide 2Fluency.
“Words like ‘y’all’ and ‘fixin’ to’ are particularly difficult for AI to understand and are often misleading,” the site says.
It came in second, with New Jersey coming in third, as the New York City accent is great for buying “gabbagle” in Brooklyn, but not so great when the AI is searching for capicola.
In a study of 3,000 shows, the Texas accent was rated as the fourth most difficult.
Surprisingly, the Bostonian ranked fifth instead of first.
“Accents in big cities like New York, New Jersey, and Boston often feature unique pronunciations and local slang that can confuse AI,” the site explains. “For example, New Yorkers tend to drop the ‘r’ sound (pronouncing ‘car’ as ‘cah’) and Bostonians say ‘pahk the cah with Hahvahd Yahd,’ which can confuse speech recognition systems.”
The Lone Gylanders came in at number 7. “The Long Island accent is characterized by a nasal tone and long vowels, which makes it often difficult for AI to guess,” the site explains.
Rounding out the top ten were accents from Appalachia (“Here comes running to my window, begging for potatoes”), California’s San Fernando Valley (think Jeff Spicoli’s “All I need is a good wave and a cool head of drunkenness”), Miami accents influenced by Cuban immigrants, and people from New Orleans.
Northerners are no exception: The Hudson Valley accent, also known as “Albany English,” which combines vestiges of Dutch with New York City’s short vowels, was ranked as the 22nd most difficult accent for AI to understand.
Ultimately, the site argues, “these misunderstandings have real-world consequences,” as users who speak with these accents perceive AI as “less trustworthy for everyday tasks,” leading to frustration and a lack of trust in the technology.
