LOVE 2000 HOGS, YEA, and other candy heart messages fresh from the neural networks

Valentine’s Day is almost here, and with it, all the Hallmark-mandated signifiers that accompany it: pink and red everywhere, ubiquitous flower delivery ads on your favorite podcasts, and, of course, candy hearts, the flavorless, disease-ridden carrier pigeons of the romantic missive world. Let’s not mince words here, folks: Candy hearts suck, and it’s up to technology to save us from them, once and for all.

Hence the latest work from the ever-ambitious dorks who love to feed various pieces of mashed-up information into their home-grown neural networks and see what gets crapped out. Janelle Shane—whose work on computer-designed ponies and paint colors we’ve covered here before—has now produced modern A.I.’s latest attempt to crack the candy heart code, and it’s honestly not too bad. Sure, there are some generically romantic duds, like “CUTE KISS” and “LOVE BUN,” but it doesn’t take long to create a new paradigm. Welcome, digital romantics, to the ALL HOVER, TIME HUG world.

Shane fed all the candy hearts she could find into a learning algorithm, and categorized the results based on how close they got to the original intent. Some—like the poignant “LOVE BOT”—were close, while others, like “STANK LOVE,” were not. Honestly, we kind of prefer the latter; romantic language has really stultified over the years, and it needs a “YOU ARE BAG” or “LOVE 2000 HOGSYEA” to freshen things up.

Normally, when we read about these instances of The Machines taking over some new aspect of human creativity, it’s with a certain measure of trepidation. But we’re perfectly happy to let the computers have this one; we’ll be more worried when scientists finally design a device capable of enjoying these sugary little nuggets of chalk, because that will mean humanity has finally, officially been surpassed.