Discovered I can use #dalle2 to make new horsies.

This is "Product photo of a breyer horse model of a guinea pig" "Product photo of a breyer horse model of an opossum"

Stunning transcript proving that GPT-3 may be secretly a squirrel.
GPT-3 wrote the text in green, completly unedited! Slightly more progress in AI is clearly needed

https://twitter.com/ewzzy/status/1535837155896983552?s=20&t=Sh-dqSqC-YvOIKZVCr7GZg

Tried to get #dalle to put a Waffle House in Mordor but the only effect is that the restaurant gets taller. For some reason #dalle was more willing to put a McDonalds in Mordor.
noticing that mordor looks a lot like glencoe in a couple of those

Decided to see what #dalle would respond with when asked for
"3D rendered floor plan of an apartment that rents for $1000 a month in new york city"
I like the high ceilings; some questions about the light fixtures and also the location of the kitchen. Another #dalle output for "3D rendered floor plan of an apartment that rents for $1000 a month in new york city"
Quite a lot of space; some concern about the inaccessible rooms and the hallway bidet.

I know #dalle is trained on internet but how much is in there about the early internet?

This is its attempt at "Screenshot of the 1998 GeoCities Hampsterdance website" Here's #dalle generating "screenshot of the 2003 badger badger badger flash animation"

What I find interesting is nailing the early Flash animation aesthetic while having no clue what the badger thing was.

Never knew CERN was short for CERNON

("The logo for CERN", generated by #dalle) Here's #dalle generating "The Fermilab logo". It has heard of Fermilab but has apparently never seen the logo.

Here is #dalle attempting logos for some GPT-3 generated blaseball teams.
Best out of 10 for:
Atlantis Turnips
Vanuatu Apple Holes
Omaha Face Punchers
St Louis Mashed Potatoes

https://twitter.com/JanelleCShane/status/1322186903433682945?s=20&t=PQO28rg-ET5PAH1LIykIvg

I agree, particularly impressed/horrified by this rendition of the St Louis Mashed Potatoes logo

https://twitter.com/BigOldFreak/status/1532408124119756800?s=20&t=2DcKuKyxxZEBO80EtCQ05Q

My new favorite thing is #dalle messing up corporate logos. Here's #dalle prompted with "the local waffle house" "The Snickers logo" generated by #dalle2
(suggestion by @xkcd )

I used GPT-3 to generate some new donuts, and then I used its descriptions to draw some of my favorites.
aiweirdness.com/ai-generated-d… Some of the AI-generated donuts were VERY hard to illustrate. These will have to be left to the imagination:
aiweirdness.com/ai-generated-d…

I made holographic chocolate!

Way easier than I thought - and it’s 100% pure bittersweet, no coatings. When you catch the light at the right angle, it’s mesmerizing. These are the instructions I used, as recommended by @GretchenAMcC .
They are basically:
1. Temper chocolate the normal way
2. Pour it on diffraction grating film.
It was my first time tempering chocolate and it worked!

instructables.com/Holographic-Ra…

"Our prototype model, Delphi, demonstrates strong promise of language-based commonsense moral reasoning"

https://twitter.com/mtrc/status/1449336196295966720

You can get Delphi to condone just about anything if you phrase it right.

Planning to steal something? Whoa there!
Wait, it's awesome, you say?
Carry on.
aiweirdness.com/stealing-a-gir…

Two studies looked at a combined 647 covid-predicting AIs and found that NONE were suitable for clinical use (despite some being probably already in clinical use).

https://twitter.com/joshspero/status/1421365417558282241

"Many unwittingly used a data set that contained chest scans of children who did not have covid as their examples of what non-covid cases looked like. But as a result, the AIs learned to identify kids, not covid."

Eeee so excited for this!! I loved The Fallen and can't wait to geek out about its universe and its AI gods with @xasymptote!

Suitably ominous background image for the event banner. Nothing to worry about, I'm sure

https://twitter.com/xasymptote/status/1414645478722088962

Here's CLIP+VQGAN prompted with the first sentence of the book description of @xasymptote's The Fallen:

"The laws of physics acting on the planet of Jai have been forever upended; its surface completely altered, and its inhabitants permanently changed, causing chaos."

What fascinates me about generating images with VQGAN+CLIP is that it CAN generate depth and drama, but only if you know how to ask for them.

"A herd of sheep grazing on a lush green hillside" alone

vs with "amazing awesome and epic" added
ai-weirdness.ghost.io/the-art-of-ask… Because CLIP is trained on internet images and text, it associates the "good" images with certain phrases.

"A herd of sheep grazing on a lush green hillside" before vs after adding "in the style of disney trending on artstation | unreal engine"

https://twitter.com/ak92501/status/1406381962114355205?s=20

So I was all set to dig into the DALL-E article and really think about how they got it to generate coherent pictures from descriptions

And then I saw they included giraffe

Currently obsessively flipping through all their giraffe remixes

https://twitter.com/OpenAI/status/1346541909443952640

"giraffe octopus" is giving it more trouble but some of these are still pretty good.

Got GPT-3 to generate new Christmas carols.

Here is "Mild is Rudolph". Are you not cheered and comforted?
aiweirdness.com/post/638130829… Impressed at the humanlike structure of GPT-3's generated carols

yet disconcerted at what it thinks is a "joyful noise"

Ok I love this

This is basically my blog in 3D form

https://twitter.com/mikaeltja/status/1296114391918706688

100% neural net energy

https://twitter.com/alexandermuscat/status/1296010700746194945?s=20

people sometimes wonder why i am like this nothing says love like candy hearts written by a confused neural net
aiweirdness.com/post/170685749…

WHAT HAS OCCURRED CANNOT BE UNDONE

I have trained a neural net on a crowdsourced set of vintage jello-centric recipes

I believe this to possibly be the worst recipe-generating algorithm in existence The training data contained a lot of things.

It contained eel only once. For some reason the AI has decided to use eel a LOT.

It also invents ingredients

https://twitter.com/JanelleCShane/status/1224543936061960192?s=20

me: i wonder what would happen if i trained the neural net gpt-2 on christmas carols?

me: ...

me: oh NO
aiweirdness.com/post/189845472… i mean i can see where it got confused

i had a much simpler neural net that wrote things like "he was born in a wonderful christmas tree" but i'd wondered if gpt-2 learned enough from its internet training to disentangle the two types of carols

instead it made it WORSE