Janelle Shane is quickly becoming the internet's neural network queen. She's come up with machine learning algorithms that have generated recipes and all those fake Pokémon we heard about a few days ago, for example. The Pokémon were convincing because nonsense was what made them appealing, but it turns out that doesn't work so well when the same neural network framework tries to come up with pickup lines. Shane tried that out, and the results were extremely varied (via BoingBoing).
Let's break down the results (which she shared here) into two basic groups, beginning with the ones that might work if the target of your affection thought that English was your second language and that the effort was cute.
"I want to get my heart with you."
"If I were to ask you out?"
"You are so beautiful that you make me feel better to see you."
"Are you a candle? Because you’re so hot of the looks with you."
"I want to see you to my heart."
"I have a really falling for you."
"Your beauty have a fine to me."
"I had a come to got your heart."
"You look like a thing and I love you."
"I don’t know you."
The last one could work if you're playing really hard to get. Inversely, here are a few that we couldn't imagine working out favorably in any scenario because they don't make a lick of sense.
"Are you a 4loce? Because you’re so hot!"
"I have a cenver? Because I just stowe must your worms."
"Hey baby, I’m swirked to gave ever to say it for drive."
"Hey baby, you’re to be a key? Because I can bear your toot?"
"You’re so beautiful that you say a bat on me and baby."
"If I had a rose for every time I thought of you, I have a price tighting."
"I have to give you a book, because you’re the only thing in your eyes."
For the most part, you can still tell that there's romantic intent behind these statements, but the made-up words and bonkers sentence structure convolute the message. Honestly, though, knowing that a robot came up with these, it's still pretty adorable. We recently learned, however, that for a robot to find love, the best way is to say nothing at all:
It may seem like the neural network didn't do the greatest job at coming up with pickup lines, but it's actually kind of incredible it even got to that point. In her blog post about working on the recipe-generating neural network, Shane shows off some results from less educated versions of the network, which help us understand how little neural networks really start with. After citing a paragraph of nonsense, Shane says it's impressive that "it’s already figured out that lower case letters predominate, and that there are lots of line breaks." Once you realize that this algorithm started out with no understanding of language whatsoever, the fact that it's able to achieve what it has is pretty spectacular.
Shane's Pokémon were brought to life, so I guess now we need one of you to film yourselves trying out these robotic pickup lines on people? Is there a possibility that any of them could actually work? Give us a shout on Twitter and let us know what you think!
Featured Image: Disney/Pixar