Janelle Shane @janellecshane@wandering.shop

So happy to share with you: the first scifi story I've ever sold, appearing today in Strange Horizons!

Contains:
- robots
- cryogenic exoplanets
- linguistics
- beeping
http://strangehorizons.com/fiction/68hazardcold/

When an AI is trained on words, weird things can happen to the physical domain.

I asked the OpenAI API about horses.
https://aiweirdness.com/post/621186154843324416/all-your-questions-answered

I've been previewing the new @OpenAI API, and honestly despite myself I was a bit nervous when I typed "go ahead and try it." It's that good.
https://aiweirdness.com/post/620645957819875328/this-is-the-openai-api-it-makes-spookily-good

What happens when a neural net tries to write rhyming lyrics?

Apparently, an ode to defunct Swiss chemical company Alusuisse.

The fine rhyming work of
Mark Riedl's Weird A.I. program

https://aiweirdness.com/post/619190785060028416/rhyming-is-hard

I trained a neural net on 1100 escape room names, and now we have some interesting new ways to be stuck in a room.

names and descriptions via Max Woolf's GPT-2-simple
creepy images via artbreeder.com
https://aiweirdness.com/post/618298096035741696/escape-rooms

I had fun experimenting with OpenAI's #jukebox , a neural net that can generate not just musical notation but the entire sound file, including lyrics and shrieks and applause.

of COURSE i used it to produce several unlikely covers of Baby Shark

https://aiweirdness.com/post/617558958271561728/jukebox-of-weirdness

Someone asked me what would happen if I took a neural net trained to produce images of cats and then trained it on #BritishBakeoff images instead.

If you've ever wished Bakeoff was more like a periodic function I have great news for you

https://aiweirdness.com/post/617199435268734976/from-cat-to-kitchen-a-cautionary-tale

This is delightful: a transcript of a Murderbot/ART AMA interview

Not a question, but I love you MURDERBOT & ART is awesome!
ART: Thank you. You get to live.
MB: It’s kidding.
ART: Okay.

https://www.tor.com/2020/04/24/feelings-redacted-what-happened-when-murderbot-and-art-talked-to-instagram/

You know there's something strange going on when you ask your neural net for "a horse on a bicycle" and you get this

https://aiweirdness.com/post/616292784700276736/a-weirdly-broken-ai

Trained StyleGAN2 on pictures of my cat.
StyleGAN2 had been originally trained on human faces.
This is an intermediate training stage.

Not sure whether to laugh or scream
https://aiweirdness.com/post/615654447163621376/trained-a-neural-net-on-my-cat-and-regret

It's interesting to see what the neural net got right about the eggs. They're convincingly 3D. But it completely failed to pick up on the speckled eggs and instead did these weird layered pearlescent eggs.

Trained StyleGAN2 on easter eggs

results strangely wobbly

occasionally furry

it's probably fine
https://aiweirdness.com/post/614935849124102144/nonexistent-easter-eggs

I used the neural net GPT-2 to generate some April Fools pranks.

conveniently they're all pranks you play on yourself. or possibly solo performance art.
https://aiweirdness.com/post/614209593176899584/an-ais-idea-of-a-prank

Trained a neural net on the Great British Bakeoff

results were less than cozy

it tried, though

StyleGAN2 via runwayml
https://aiweirdness.com/post/613754259190808576/this-is-not-cozy-ai-attempts-the-great-british

Did you know there's a paper where they spent 3 weeks training a neural net to transform giraffes into birds?

new blog post
https://aiweirdness.com/post/613123668121419776/how-to-change-a-giraffe-into-a-bird

Would love to see these neural net generated court cases rendered by a real court artist.

There's a whole class of weird US court cases against inanimate objects:

United States v. Approximately 64,695 Pounds of Shark Fins
United States v One Solid Gold Object In The Form Of A Rooster

I decided to see what a neural net would make of them.

https://aiweirdness.com/post/612669075940900864/court-cases-that-sound-like-the-weirdest-fights

These candles don't (and probably shouldn't) exist.

I trained the neural net GPT-2 on just 72 descriptions of scented candles collected by Amy Pollien The neural net relied on its general internet training for the rest.

https://aiweirdness.com/post/611853103688499200/this-neural-net-knows-what-smells-good

Scented candles from a neural net trained on 1000 existing candles

The neural net is based on math, so it is infallible. This is the future.

https://aiweirdness.com/post/611219307217797120/smells-like-the-future