Blessed with a little bundle of joy, but not sure what to name him or her? Then why not let an algorithm do it?
That is the basis for an intriguing experiment by coder and Brown University student Nate Parrott, who has created an artificial neural network that remixes everyday baby names into whole new creations.
“I trained an AI algorithm to read 7,500 common American baby names, encode them into numbers, and decode them back into the original names,” Parrott told Digital Trends. “Since every name has to be encoded as just a few numbers, the encoder learns to remove the common sounds that names have, and the decoder learns to add them back. It lets you blend names together, ‘add’ and ‘subtract’ them — and by feeding in random numbers, make up plausible-sounding names.”
Not all the names are entirely plausible, mind you.
“There’s a couple normal-sounding ones like Rosele and Halden that I like — but I’m honestly a bigger fan of the really strange ones like Pruliaaa, Onni or Kattt,” Parrott continued. “It has a weird tendency to repeat letters way too many times near the end of names and I think it’s kinda cute.”
Parrott said that he started the project because he was interested in the way that deep learning algorithms are “good at dealing with loosely defined things,” and wanted to see how they would handle naming conventions.
“Hopefully this inspires more creative people to get into computer science or machine learning,” he said. “There’s so much potential for creative uses of AI, and we’ve barely scratched the surface. Of course, if someone actually gives their kid one of these names, I’d be super excited, too. And a bit shocked, and kind of guilty.”
Still, it’s more original than naming your baby after the year’s top sports star, pop singer, or best-looking Digital Trends writer, right? (To be fair, that latter would still be somewhat original.)
To find out more about the project, you can check out Parrott’s original post about it here.
