Someone trained a neural network to generate candy heart messages — and they’re a bit off

Twitter/@JanelleCShane

Well, at least these artificial intelligences tried.

Videos by Rare

RELATED: The Internet mocks titles of “I Know What You Did Last Summer” movies

Valentine’s Day is almost here, and so are those chalk-like candy hearts with sweet messages. But robots, tasked by research scientist Janelle Shane to create messages of their own, still haven’t figured out that whole love thing.

Shane, on a website called aiweirdness.com, publicized some of the bizarre candy heart message results AI came up with after being trained in some human niceties.

“I train neural networks, a type of machine learning algorithm, to write unintentional humor as they struggle to imitate human datasets,” Shane described what she does on the website.

Shane “collected all the genuine heart messages [she] could find, and then gave them to a learning algorithm called a neural network.”

This is what happened next:

 

Nothing says “be mine” quite like “sweat poo,” “stank love,” “bog love” and “cool cud.”

“Hole” and “love 2000 hogs yea” were also among our personal favorites.

RELATED: Everyone hates their alarm clocks

In fairness to the AI, some of these were pretty close to having the desired effect. “Heart me,” “love bun,” “my bear,” “love bot” and “you are it” are close enough.

People on the internet couldn’t contain their laughter over some of the worst efforts, though.

https://twitter.com/DayGl0Chainsaw/status/962311611393499138

https://twitter.com/radelgrill/status/962504803535507457

Shane has also gotten AI to generate pick-up lines in the past.

Winners like “Hey, baby. You’re to be a key? Because I can bear your toot?” and “You must be a tringle? Cause you’re the only thing here” surely have robots hot and bothered everywhere.

What do you think?

A church didn’t tip on a $735 order, but it was the waitress who got fired for speaking about it

The stunt coordinator from “Kill Bill” says he wasn’t “notified or consulted” prior to crash