Image of candy hearts that say, "Be Mine" and "Hug Me" for Valentine's Day Image credit: https://www.flickr.com/photos/fdotguido/

From “Fancy My Hero” to “Tweet up Bat,” This Neural Network Wrote the Only Candy Heart Messages Fit for 2018

This article is over 6 years old and may contain outdated information

Recommended Videos

Researcher Janelle Shane, whose work with neural networks we’ve covered before, is at it again. This time, she tried to train a neural network to create its own candy heart messages. “I collected all the genuine heart messages I could find,” she explains at her website, “and then gave them to a learning algorithm called a neural network. Given a set of data, a neural network will learn the patterns that let it imitate the original data—although its imitation is sometimes imperfect. The candy heart messages it produced … well, you be the judge.”

The results were, perhaps unsurprisingly, mixed. Some of the bot’s suggestions, like “LOVE BUN,” seemed pretty accurate, while others like “YOU ARE BABE” and “FANCY MY HERO” only managed to capture the sentiment of a candy heart. From there, of course, things got weird. From “I HONKER” to “TWEET UP BAT” to “U HACK,” the network’s next tries were decidedly less human-sounding. Still, I have to love the sound of the verb “honker.”

Below are just a few of my favorite selections. You can see the full list over at Shane’s website. As always, tag yourself. (I’m HOGSYEA.)

  • AM GOOD
  • BEAR WIG
  • BOG LOVE
  • DEAR ME
  • FANCY MY HERO
  • HOW U HOT
  • I HONKER
  • LICK
  • LOVE 2000 HOGSYEA
  • LOVE BOT
  • LOVE BUN
  • MY BEAR
  • MY MY
  • YOU ARE BABE
  • PIN A FACE
  • STANK LOVE
  • TIME HUG
  • TIME STAR
  • TWEET UP BAT
  • U HACK
  • WIRLY OOT
  • YOU A GOO
  • YOU ARE BAG
  • YOU ARE BOA

Shane later increased the size of her dataset. For the first batch, she writes, “I could only find about 360 existing messages total. So, when I decided to generate some more messages, I decided to add my favorite neural network-generated messages to the original dataset, increasing the total to almost 500. This is still really small for a neural network, but a noticeable upgrade.”

What did the network do with the enhanced dataset? “The first thing I noticed,” Shane said, “is the definite upswing in the number of messages involving bears.”

  • BE MY BEAR
  • LOVE BEAR
  • LOOK BEAR
  • WINK BEAR
  • BEAR BEAR
  • TIME BEAR

The other new suggestions included the below nonsense phrases, and you can see more in the full second list at Shane’s website.

  • OOO
  • BEST MANE
  • IN A FAN
  • MAGE LOVE
  • WEAR ME
  • SWEET BOG
  • OY
  • MEAT LOVE
  • MY HAG
  • LOVE BAG
  • SWEET MEAT

Haven’t had enough yet? Shane writes that the neural network also generated some inappropriate suggestions, so you can request to view those via this form.

Shane’s previous neural network experiments include Dungeons & Dragons spells, guinea pig names, and cookbook recipes. If you have any questions about playing around with your own neural networks, she has a great FAQ at her site.

(via NPR and Nerdist; image: Flickr/Frank Guido under CC 2.0)

Want more stories like this? Become a subscriber and support the site!

—The Mary Sue has a strict comment policy that forbids, but is not limited to, personal insults toward anyone, hate speech, and trolling.—


The Mary Sue is supported by our audience. When you purchase through links on our site, we may earn a small affiliate commission. Learn more about our Affiliate Policy
Author