BUT I'M NOT ALLOWED TO NAME THINGS

They gave me a number to shoot for.  It was about twice than what I had been making at the time.

So I went online, did some research, found some figures backing up my claim, put it all together and went to my boss.

I got what I asked for.

If it hadn’t been for that person in accounting telling me I was way underpaid, I’d have never known.  I went from barely scraping by to being able to have a savings account and getting all my debts paid thanks to them.

1. magneto was right (accepted all mutants to his cause instead of just the pretty ones, taught them to love and respect themselves, actively fought against anti-mutant rhetoric in order to prevent another holocaust)

It’s hard to come up with ideas for Halloween costumes, especially when it seems like all the good ones are taken. And don’t you hate showing up at a party only to discover that there’s *another* pajama cardinalfish?

I train neural networks, a type of machine learning algorithm, to write humor by giving them datasets that they have to teach themselves to mimic. They can sometimes do a surprisingly good job, coming up with a metal band called Chaosrug, a craft beer called Yamquak and another called The Fine Stranger (which now exists!), and a My Little Pony called Blue Cuss.

So, I wanted to find out if a neural network could help invent Halloween costumes. I couldn’t find a big enough dataset, so I crowdsourced it by asking readers to list awesome Halloween costumes. I got over 4,500 submissions.

The most popular submitted costumes are the classics (42 witches, 32 ghosts, 30 pirates, 22 Batmans, 21 cats (30 incl sexy cats), 19 vampires, and 17 each of pumpkins and sexy nurses). There are about 300 costumes with “sexy” in their names; some of the most eyebrow-raising include sexy anglerfish, sexy Dumbledore, sexy golden pheasant, sexy eyeball, sexy Mothra, Sexy poop emoji, Sexy Darth Vader, Sexy Ben Franklin, Sexy TARDIS, Sexy Cookie Monster, and Sexy DVORAK keyboard. In the “technical challenge” department, we have costumes like Invisible Pink Unicorn, Whale-frog, Glow Cloud, Lake Michigan, Toaster Oven, and Garnet.

All this is to say that humans are very creative, and this task was going to be tricky for a neural network. The sensible approach would be to try to use a neural network that actually knows what the words mean - there are such things, trained by reading, for example, all of Google News and figuring out which words are used in similar ways. There’s a fun demo of this here. It doesn’t have an entry for “Sexy_Gandalf” but for “sexy” it suggests “saucy” and “sassy”, and for “Gandalf” it suggests “Frodo”, “Gollum”, and “Voldemort”, so you could use this approach to go from “Sexy Gandalf” to “Sassy Voldemort”. 

I wanted something a bit weirder. So, I used a neural network that learns words from scratch, letter by letter, with no knowledge of their meaning, an open-source char-rnn neural network written in Torch. I simply dumped the 4500 Halloween costumes on it, and told the neural network to figure it out.

Early in the training process, I decided to check in to see how it was doing.

Sexy sexy Dombie Sexy Cat Sexy A stare Rowan Sexy RoR A the Rog Sexy Cot Sexy Purbie Lampire Poth Rat Sexy Por Man The Wombue Pombie Con A A Cat The Ran Spean Sexy Sexy Pon Sexy Dander Sexy Cat The Gull Wot Sexy Pot Hot

In retrospect, I should have expected this. With a dataset this varied, the words the neural network learns first are the most common ones.

I checked in a little later, and things had improved somewhat. (Omitted: numerous repetitions of “sexy nurse”). Still the only thing that makes sense is the word Sexy.

Sexy The Carding Ging Farbat of the Cower Sexy The Hirler A costume Sexy Menus Sexy Sure Frankenstein’s Denter A cardian of the Pirate Ging butter Sexy the Girl Pirate

By the time I checked on the neural network again, it was not only better, but astoundingly good. I hadn’t expected this. But the neural network had found its niche: costume mashups. These are actually comprehensible, if a bit hard to explain:

Punk Tree Disco Monster Spartan Gandalf Starfleet Shark A masked box Martian Devil Panda Clam Potato man Shark Cow Space Batman The shark knight Snape Scarecrow Gandalf the Good Witch Professor Panda Strawberry shark Vampire big bird Samurai Angel lady Garbage Pirate firefighter Fairy Batman

Other costumes were still a bit more random.

Aldonald the Goddess of the Chicken Celery Blue Frankenstein Dancing Bellyfish Dragon of Liberty A shark princess Statue of Witch Cupcake pants Bird Scientist Giant Two butter The Twin Spider Mermaid The Game of Nightmare Lightbare Share Bat The Rocky Monster Mario lander Spork Sand Statue of pizza The Spiding hood A card Convention Sailor Potter Shower Witch The Little Pond Spice of pokeman Bill of Liberty A spock Count Drunk Doll of Princess Petty fairy Pumpkin picard Statue of the Spice of the underworker

It still was fond of using made-up words, though. You’d be the only one at the party dressed as whatever these are.

Sparra A masked scorby-babbersy Scormboor Magic an of the foand tood-computer A barban The Gumbkin Scorbs Monster A cat loory Duck The Barboon Flatue doctor Sparrow Plapper Grankenstein The Spongebog Minional marty clown Count Vorror Rairol Mencoon A neaving hold Sexy Avical Ster of a balana Aly Huntle starber pirate

And it ended up producing a few like this.

Sports costume Sexy scare costume General Scare construct

The reason? Apparently someone decided to help out by entering an entire costume store’s inventory. (”What are you supposed to be?” “Oh, I’m Mens Deluxe IT Costume - Size Standard.”) 

There were also some like this:

Rink Rater Ginsburg A winged boxer Ginsburg Bed ridingh in a box Buther Ginsburg Skeleton Ginsburg Zombie Fire Cith Bader Ginsburg

Because someone had entered about 50 variations on Ruth Bader Ginsberg puns (Ruth Tater Ginsberg, Sleuth Bader Ginsber, Rock Paper Ginsberg).

It invented some awesome new superheroes/supervillains.

Glow Wonder Woman The Bunnizer Ladybog Light man Bearley Quinn Glad woman robot Werewolf super Pun Super of a bog Space Pants Barfer buster pirate Skull Skywolk lady Skynation the Goddess Fred of Lizard

And oh, the sexy costumes. Hundreds of sexy costumes, yet it never quite got the hang of it.

Sexy Scare Sexy the Pumpkin Saxy Pumpkins Sexy the Pirate Sexy Pumpkin Pirate Sexy Gumb Man Sexy barber Sexy Gargles Sexy humblebee Sexy The Gate Sexy Lamp Sexy Ducty monster Sexy conchpaper Sexy the Bumble Sexy the Super bass Pretty zombie Space Suit sexy Drangers Sexy the Spock

You bet there are bonus names - and oh please go read them because they are so good and it was so hard to decide which ones to fit into the main article. Includes the poop jokes. You’re welcome.

I’ve posted the entire dataset as open-source on GitHub.

“Meadowalker.”

“…”

Or, what happens if you train a neural network on the titles and plot summaries of over 100,000 works of Harry Potter fan fiction.

In the decades since the Harry Potter books were published, fans have written literally hundreds of thousands of Harry Potter stories of their own, and shared them online. Can a neural network join in on the fun?

In a way, everything a recurrent neural network writes is fan fiction. A recurrent neural network looks at an example dataset (such as the complete Sherlock Holmes stories) and teaches itself the patterns and conventions that it sees. So, if it’s given Sherlock Holmes stories, it will become obsessed with Holmes and Watson, and if it’s given knock-knock jokes, it will spend all day telling awful knock-knock jokes of its own.

Thanks to an idea by a couple of readers, some heroic work by @b8horpet in scraping (with permission) hundreds of thousands of Harry Potter fan fiction titles and summaries from AO3, and a flexible new recurrent neural network implementation by Chen Liang, the neural network’s latest obsession is Harry Potter.

The Perfect Party by iamisaac Draco has been left alone, and Ginny confused must learn and who has his best friend. They were breathed by a love that didn’t become his grounds and the flowers begin.

This is a typical example of the neural network’s fan fiction - romantic pairings of two or more Harry Potter characters (called “ships” in fan fiction-speak). In this case, it even has chosen a plausible author: iamisaac is a real and fairly prolific fan fiction author whose works do tend to be of the “romantic” variety. 

The Garden by perverse_idyll for lexigilite Ron and Hermione move after a man party. What did her best things go and has to deal with people she loves? How many imperfect love really belonges them and needs to be a person? Or will they learn and more than the war? Mirror Thing by Queen_Elexhan “Are you there for a relationship? I was a sad future for your love.”  Harry and Ginny find out the meaning is.

Shatters by Kis [archived by TheHexFiles_archivist ] Based on the Spot Are It Falls Into A Heir by NextrangeOnTheThree Draco and Hermione share a whole indescribbening.

Again, “perverse_idyll” and “TheHexFiles_archivist” are fairly active authors. (Hi, if you’re reading! The neural network seems to like your writing, and is writing fan fiction of your fan fiction!) Those familiar with Harry Potter fan fiction will not be surprised to learn that the neural network really likes to generate ships; pretty much every combination of characters is represented (some of the more unusual combinations being “The Snow/Voldemort”, “The Ministry/Draco Malfoy”, and “Voldemort/Random Quidditch Child”).

By turning down the neural network’s creativity setting to near-zero, we arrive at its vision of what the quintessential Harry Potter fan fiction would be like - and we also learn its favorite ship:

Persuading by theladyblack Harry and Draco are still a second chance at the end of the war.  Will they be able to do with the fairy tale of the first time they were a strange stranger to the street of the war and the war is over?

It turns out the neural network is obsessed with Harry/Draco, although in a pinch, Sirius/Remus will also do.

The neural network also seems to really like stories about Professor Snape trying to do rather ordinary things:

New Moon Boys by Dungoonke for Loki_Kukaka Severus Snape comes back to a night’s politics.

In the Reason Is Blinders by LittleRoma Severus has been through his lost remote.

In The Alteri Silence by Forest_of_Holly for roscreens41 Snape receives life after plants to do by work over whether they get into. Just Hell.

A Second Chance by DarkCorgi Snape had a second thing, and that is better than anything for for the rest of his life.

Mirror by orphan_account Severus Snape tries to get a lot of dragons and that was to be more than he didn’t expect to continue. He has always been a bit of an old and a baby to stay the way he’d been the brother at Hogwarts and he keeps the chance of meeting… Deception by FlyingEyes Snape is a British Robes of interesting things and worrys like a little fun and sees the pretty battle for a while.

Another thing that happened, which is pretty much my favorite thing ever, is that the neural network apparently encountered some fan fiction stories that were not in English. As a result, it learned to do this from time to time:

The Secretary Of the World  Challenge inspired by GoF and la mating resigns de la mill colors per mereple beruit carteur la pelete el wert rardo completing and herillo intus den una a des rush sentines kelta an transoles… 

Between by Cheyangel13 A series of fivers are unexpectedly depressed and controlled by the bed, with least more from una perfemale erpensa de the maesse akai suidadium dela vida call de la los se terriuus do form en sou dies de fasurard il resisted de for dogs la sementu sein prong colors itu dee adte se sige natard…

The neural network has also learned to employ capital letters:

Les finds love by violet_quill for starstruck1986 Severus Snape wanted him to be more and she likes Draco.  The person he wants an energy to him.  WHALIDE NO GEATIRE SOURR INSPE AHARMANABLISH ALL SOME TO VERY THE RERIDE!!!!!!!

secret Quidditch by snapsleert Collapse and find the second worst and very different. See Gain and Descent motivate surprising death. Unbusing one of the months: should make more bumo.choooshots. HUGULATED

And the neural network occasionally uses content warnings, although it seems to have a rather fuzzy idea about what to warn its readers about:

Better With The Broom Complicate by Margyn_Black Tonks gets more than the best girl of creation. (Rated Maturisle, mark, a violence, contract) (slash] part of themes) ferret.

Art for the Sun a Scary by disillusionist9 A collection of warnings: characters and situations of silence.

Some of the neural network’s stories, though, are just plain weird.

Harry Potter and the Painful Eyes by dark_pook A Birthday drabble about the problems and a woman who shows up a lot less than she checks at Hogwarts in the destiny to the infamous adventure of control of the Art of The Good Boy Kings With Hermione. Harry and the Blue Special Delicious by apolavia_scg An unexpected potions messaged in the world their lives are to find friendship following the day of different pagers. James and Lily come to the summer before the war.

The Perfect Cow by alafaye Severus and Hermione start a horcruxes

Art: Let Draco roll the light of the moon, and means. by Dangelanne What happens after the war. Not drawn to Draco Malfoy jumpers. Originally written in 2008.

Birds of a Saturday by SasuNarufan13 Harry Potter is drunk and discovers he is an alternate universe.

Holly theody by yesIpxdishoftlyGrinli What would be dangerous! Side Voldemort Jones does all lord off the sunshine show.

Lily Evans and the Ravenclaw of a Christmas Surprise by ci Severus angst the truth of a frighten situation for the wink.

Persuasion by Samanthian The Sorting Hat is fighting in one of the houses.

lily’s family by sharkle Harry woke up in searching after a werewolf Sherlock’s picnic. He is furious.

As a bonus, I leave you with some fairly-plausible screennames the neural network invented, which appear not to be taken (yet):