We've seen neural networks create some amazing--and amazingly weird--things, from bizarre candy heart messages to designing dinosaurs out of flowers. You might remember research scientist Janelle Shane from her algorithm-generated Dungeons & Dragons spell names last year. She's at it again, this time using a neural network to create creatures you might find in the weirdest game of D&D you've ever played.
(Although with creatures with names like Walfablang and Giant Dwarf, it might also be the best game of D&D you've ever played.)
On her blog post, Janelle goes into a little more detail about both the neural network and the monsters it generated. "Colin Fredricks, who created the RPG Sufficiently Advanced, was kind enough to send me the names of 2,205 creatures from the 2nd edition monster manual," she wrote. She then trained her algorithm with the dataset, and it made magic.
The neural network created a long list of magical creatures, from dragons to unicorns to slugs, that we're already dying to battle. Some of them include:
- Spectral Slug
- Goblin Dog
- Purple Fang Dragon
- Bat Unicorn
And there's even more, if you sign up for email updates. You'll get neural network humor delivered straight to your inbox, including even more D&D creature names--like the Giant Cobra Elf or Winded Genie.
Janelle is already hard at work at her next neural network dataset to generate D&D character names, and she needs your help! Enter your character's info to help train the algorithm, and who knows what kinds of heroes you might meet in your next adventure.
Which of these crazy creatures do you want to fight? (Or keep as a pet?) Tell us in the comments.
Featured Image: Wizards of the Coast
More fantasy news you might use:
- Nicolas Cage meets Dragon Ball.
- The most popular D&D race and class combos.
- Russia's fire-breathing dragon statue.