A Neural Network Tried To Create A Bunch Of D&D Monsters, The Results Are…Strange

If trained correctly, A neural network can serve as a lightning-fast, very effective AI skilled in the delicate art of semi-random reorganization and generation. If not trained correctly, however, it can still be quite entertaining. This example is probably closer to the second scenario.

This list was generated by a recurring neural network created and programmed by well-named Tumblr user AIWierdness. In their post, the user sought to give the AI the best starting data pool as possible from which to generate the monster names, so they entered a full library’s worth of over 2,300 names from the old days of 2nd Edition Dungeons and Dragons, formatted important things like the maximum and minimum possible lengths for the generated names, and sent the AI on its merry way.

The result? Not quite as laughable as one might assume. In fact, a fair number of the given creatures sounded fairly intimidating, such as the easy-to-imagine yet probably difficult-to-fight Vampire Bear, the totally rad Chromatic Wolf, or the not-too-scary Spectral Slug.

Then there are things that somehow made it into the list as “monsters”, but I’m not sure that word really applies here. Things like…

Mommy, Greater: Will probably bake you some cookies and send you on your way. I’d probably seek out this monster rather than avoid it.

Purple Bird: Not exclusive to the Dnd World.

Cloud Of Chaos: Probably responsible for all those sporadic showers we’ve been getting recently.

Big Dragon: Like 7% larger than the average dragon.

Dragon, Dead: Decidedly less scary than a living dragon. Smells terrible.

Skull: Best known for their terrifying ability to lie dormant inside of one’s head, waiting for their time to strike.

Barber: Finally, my irrational fear of barber shops has been justified.

Unicorn, Bat: Is it a unicorn or a bat? Make up your mind!

Feast, Stone: That’s one feast I’m not interested in.

Golf: Capable of destroying entire forests in mere months, only to grow and create more room for itself.

Vampire, Putter: Often found near wild Golfs.

Memeball: Full of outdated jokes and references. A terrifying foe.

And while many of these names are pretty absurd, these are actually among the better attempts made by the AI. At the very least, the meanings or general nature of these creatures can be determined. At the very least, it came up with a few words that are actually real words. There were a lot of other generations that proved to be much, much less lucky.

Dragon, Curple Lard: I don’t know what it is, but it might want to consider a new fitness plan.

Sirper: Can’t even figure out what it was going for here.

Dragon, Cerrous, Aeriental, Negative Gname Guant: Without the commas, that’s a Negative Gname Guant Aeriental Cerrous Dragon. Hope that clears things up.

Man-Can: I know a man who can can a Man-Can, as long as the Man-Can doesn’t can the man’s hands.

Jabberwont: Less of a name for a monster, more of an obtuse way of telling someone you’ll be quiet.

There’s a lot more creatures that I didn’t cover. Click here for a full list. They’re truly a sight to behold.

Or, if you’d like to witness this AI’s continued rampage across the world of DnD, check out its similarly unsuccessful attempt to invent a long list of DnD spells here.(Link that)