Dungeons and Dragons creatures, generated by Neural Networks
I'm not sure I can sufficiently explain this one, but basically over on Letting neural networks be weird they've trained a Neural Network to create D&D monsters, it's almost like random generation, but it's come up with some weird and wonderful creatures. I just wish it made the stats too.
The game Dungeons and Dragons has all sorts of lists of spells and abilities you can use for gameplay. I trained an algorithm called a recurrent neural network to generate more spells - like Barking Sphere, Hold Mouse, and Gland Growth. Then, with a larger set of spells in my training dataset, I trained a better neural net that generated even more - like Song of the Dave, Summon Ass, and Shield of Farts.
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. 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.
As I had hoped, the neural network generated creatures that would probably be pretty awesome.
Cloud of Chaos
It also generated some creatures that you should probably run from until you figure out what they are. (Though Dome Animal might simply be a cool turtle)
Horse (Spider, Brain
Undead Lake Man, Fire
Fish, Sun of
Pat, Great, Space
Spectral Woof Greepy
There were 118 dragons in the original dataset, so of course the neural net liked generating new dragons. Some perhaps better-conceived than others.
Dragon, Death Seep
Dragon, Purple Fang
Dragon, Curple Lard
And it generated new unicorns!
Unicorn, Black Willow
Unicorn, Sith Sheet
These might be possible misses, though.