What happens if you let a neural network loose on inventing names for monsters in Dungeons and Dragons? Well, it turns out it comes up with some rather ridiculous suggestions.
Research scientist Janelle Shane from Boulder, Colorado previously used a recurrent neural network to come up with some odd spell names for D&D, but this time around she turned her powers of hilarity towards creating new names for monsters.
“It turns out that in addition to spellbooks, Dungeons and Dragons also has monster manuals – books full of the names and descriptions of creatures that adventurers can encounter,” she wrote on her blog AI Weirdness.
So she fed the names of 2,205 creatures from the second edition monster manual into her neural network. This led to a variety of rather wonderful names: Spectral Slug, Vampire Bear, Spare Ogre, and Cloud of Chaos among them.
We loved Goblin Dog, too.Then there was a range of somewhat fearsome-sounding creatures, including Fire Brain, Marraganralleraith (?), and Fire Undead Lake Man. Oh and perhaps the more timid sounding Desert Beeple.
And there were a few dragons too, including the imaginatively named Big Dragon, and also Purple Fang Dragon, and Dead Dragon. Finally, a few unicorns – Fumble Unicorn and Black Willow Unicorn among them.
Shane said she was now asking people to enter their own character’s name, race, and class from D&D, so she could try and generate character names in a future project. And she’s looking for some character backstories too, presumably to try and get the neural network to write a few of these.
This isn’t the first time we’ve seen AI have a stab at coming up with ideas. Shane has previously shared her AI attempts at convincing-sounding disease names, "sexy" Halloween costume ideas, and guinea pig names.
It's even more difficult for AI to get creative. Last year, there was a rather amusing attempt to write the next Game of Thrones book, while we also saw a hilarious attempt at writing Harry Potter fan fiction. And don't forget those inspirational posters.
It’s not all fun and games though. Earlier this year, AI was successfully used to finally crack the mysterious Voynich manuscript, and a deep neural network was used to recreate images from inside the human brain.