Say there, they got that, whachacallit, clickybait on this Patch website there. Headline:
Does the Minnesota accent confound AI?
Maybe if we exaggerate our accent, we’ll foil Big Brother.
![An old-fashioned card catalog](https://arc.stimg.co/startribunemedia/6ODMKLZ2WSTWPFJUN7JY4MTLAQ.jpg?&w=712)
“MN Accent Among Most Difficult for AI to Understand: Survey”
Really? More difficult than a sultry-mushmouth Suthun accent? Tougher than a Bostonian who asks his phone where he can pahk the cah? Apparently if you ask AI to subtitle the dialogue in the movie “Fargo,” it goes like this:
“Say there, Marge, why doncha go to the Cub fer some donuts?”
AI interpreter: (lights flicker, smoke curls from the processor, sparks fly, AI types ”hicks talking gibberish” on the screen).
I’m sure the news will spur someone to propose solutions. Either they fine-tune the program so it captures our unintelligible nuances, which fall on the ears of AI like a Finnish person pronouncing a Welsh town name with a mouth full of novocaine, or we all start speaking like Omaha TV anchors.
We’d better do something, lest Minnesota fall behind in the forthcoming AI-assisted world. I mean, what will we do without AI holding our hand at every step, correcting our grammar, answering all our pleas and queries?
The other day I used AI to answer a question about a movie. Its reply was incorrect. I knew that. But what if I hadn’t? It makes you remember how we got answers in the long-ago days. We went to the library.
You got on your bike, pedaled to the gas station because you needed air, filled up, maybe put fresh playing cards in the clothespins, then pedaled downtown to the library, only to find it was closed.
But then you did the same thing again the next day, and you had to get air again because you had a slow leak. Maybe you ran over a tack. There were always lots of tacks in the street in those days, it seems. Which made you wonder why.
Once you got to the library, you’d head to the vast wall of wooden drawers called the card catalog and wonder if you should look under “tacks” or maybe “nails,” but it turns out that you should’ve been looking under “Objects, Random, Roadway, Sharp,” and you’d find a card for “The North American Field Guide to Road Tacks, Strewn Nails, Errant Staples.”
The Dewey Decimal number would be 831.934.3322b, so you’d go to the 800 aisle, find the right spot, get a small metal ladder because it was on the top shelf, then discover someone had checked it out.
So you’d go to the desk and ask if you could reserve it when it came in, and then you’d go home and watch “Kojak.”
When the book was returned, you might have your question answered, if someone hadn’t ripped out the page you wanted, or you’d learn that the book was from 1952 and didn’t cover Upper Midwest tacks.
It’s much better today. “Hey, Siri, what’s the average length of a sharp metal object on the ground, waiting to pierce the tender flesh of my tire?”
Siri: “I can’t find the answer to that question, but here’s ‘Pierce my Tender Flesh’ by Skewer D. Tatmaster MC from Apple Music.”
“OK, well, Google Gemini, what’s the average length of a sharp metal object on the ground, waiting to pierce the tender flesh of my tire?”
Google Gemini: “I’m sorry, but discussing sharp metal objects piercing skin violates my safety protocols. This is going in your permanent record, which can never be deleted, and now resides in a server six floors down in the CIA basement.”
These are growing pains. Once the AI models have all the proper information, the correct and necessary information will be instantly available. But again, this depends on them understanding the way Minnesotans talk, so let’s go back to that survey. The one that said “MN Accent Among Most Difficult for AI to Understand.”
The list has 15 accents, with “Southern” at the top of the most misunderstood.
Minnesota is No. 13.
Meaning we’re not difficult at all. Everyone can understand us, even our digital assistants. I called up Chat-GPT and said, in as Minnesotan an accent as I could, “Tell me how Minnesotans say something is not good while being passive aggressive.”
Chat-GPT responded: “You might hear a comment like ‘Oh, that’s certainly ... interesting.’ Or it may say ‘That’s different.’”
I’d say AI understands us just fine.
While the American version is a faithful adaptation, it’s bigger and shiner than the original.