See Ya at What Gets Me Hot: New Years AI by Janelle Shane

New Years Resolutions generated by AI

By Janelle Shane

This month I'm beginning 2022 as the first Futurist in Residence at the Smithsonian Arts and Industries Building.
It's weird to think of myself as a futurist. I write a lot about the algorithms we're calling artificial intelligence (AI), but rather than deal with the humanlike science fiction version, I focus on what today's much simpler AI is capable of. Since today's AI relies on using trial and error to get better at predicting its training data, and its training data must necessarily be from the past, its job is really to predict the past. This has a big effect on what it's like to use AI to predict the future.
Since we're entering 2022, the folks at the Smithsonian thought it would be interesting if I could use AI to generate New Year's Resolutions. What does it look like if I try to use AI trained on past data to suggest positive changes for the future?

I decided to use GPT-3, which is how people tend to refer to a group of large text-generating AI models that OpenAI trained on a huge collection of internet text. News articles, food blogs, social media, spam - it's all in there somewhere. Including plenty of times people have talked about their own New Year's resolutions.
Give one of the larger GPT-3 models - for example, DaVinci, the very largest - the beginning of a list of New Year's resolutions and, since its job is to predict what text comes next, it'll generally add more resolutions to the list.
In the example below, the only text I wrote is the list title in bold. With that as a beginning, DaVinci predicted from its internet training that what would come next is this list of resolutions.
List of New Year's Resolutions
To attend all Orientation committee meetings
To become more involved in Plattsmouth Community
To go bowling with teammates
To become a good writer
To start a successful business and become a millionaire
They're plausible resolutions. Ask DaVinci for more predictions and the list will be similar. People who post their resolutions on the internet tend to pick similar sorts of things. At least, they did when DaVinci's training data was collected in October 2019 - things have probably changed a bit since then.
But I didn't want to stop here, because I didn't just want a generic list of resolutions from the average internet human of the past. I wanted something that felt personal. Something a bit weird.
For my next attempt, I started the list with example resolutions that three strangers had mentioned on Twitter in 2021. I chose resolutions that were a bit offbeat. My example resolutions are in bold, and DaVinci's additions follow verbatim.
List of New Year's Resolutions
Not to buy any clothes except underwear for a year.
Touch grass and record every time I do it.
Use more insect textures in creature design.
Draw the front cover of a writing manual.
Get the missus to go to the shops for a new hat.
Find wallpaper for the kitchen/bathroom, and then paint it.
Finish the prints for the QI drawings.
Draw more detail into the background of the QI drawings.
Leave all the Christmas trees up and then paint them, as cubist art.
Eat nothing but broccoli and paint everything broccoli related.
Grow broccoli, and paint broccoli.
Make broccoli the national currency and then paint that.
This time it's a journey. DaVinci, when faced with predicting what would come next on this list, predicted first one drawing-related resolution, and then multiple others. Soon this became not just a list of resolutions, but specifically a list of drawing-related resolutions. It generated a broccoli-and-drawing-related resolution, and then the list became a list of resolutions by a painter/broccoli fanatic.
You can see how this goes. Each of DaVinci's predictions feeds forward into the future, and can become self-reinforcing obsessions.
I found that I could shape DaVinci's list by choosing a point to cut it off, deleting the end, and then having DaVinci try to generate the list again. In that way, one resolution or two at a time, I could build up a list of resolutions that, if not exactly advisable, were at least interesting.
List of New Year's Resolutions
Not to buy any clothes except underwear for a year.
Touch grass and record every time I do it.
Use more insect textures in creature design.
Take photos of each of my toes daily.
Egg every house in the village where I was born.
Lick a branch of a tree and repeat it every day for a year.
Walk down my block backwards looking over my shoulder.
Every time I am alone in the dark I will eat an apple.
Eat my favorite book.
I quickly learned how easy it was for DaVinci to develop themes.
If it generated:
Attend the Rainforest Action Network Benefit.
Then next it might generate:
Ask a rainforest tribe what they think of eco-tourism.
Go on a tour of the Tambopata National Reserve.
Go on a tour of the Manu National Park.
Write letters to the editor about rainforest preservation.
When I let it generate a resolution to eat my favorite book, next it would suggest:
Eat a book every day
and if I didn't delete that and have it try again, soon it would be suggesting:
Eat a picture of myself everyday.
Super charge my digestive system and then, eat a picture of myself everyday.
I also learned to stay away from gloominess, or platitudes, or anything to do with fitness goals - it was too easy for DaVinci to get stuck in a rut. Anything shocking or mean-spirited also tended to poison the list. I had to delete these:
Every day I will blatantly eavesdrop.
Belch until my teeth explode.
Many were worse. DaVinci is trained on internet text, after all, and so it has a tendency to veer into racism or spam. I needed to be present at every point for careful hand-pruning.
In the end I deleted many, many more predictions than I kept - perhaps about 10 times more.
It began to feel like a metaphor for life choices. Become the kind of person who spends a year licking trees and eating apples in the dark, and who knows what you'll be doing next. Spend a year trying to belch your way into exploding teeth, and that'll have an effect on you too.
I present to you my list, my own list that emerged from each prediction that I let stand, and that was also shaped by each prediction that caused me to delete, go back, regroup, retry. These aren't all things that I would necessarily do myself (I like my sleep far too much to be going out every night at midnight to make grass pancakes), but they are all things that I thought were interesting, that I wanted to hear more about.
Record every adjective I hear on the radio.
Act like a cabbage for a month.
At 4 o'clock every day I will climb a tree.
Speak only to apples for 24 hours.
Jump in front of a moving tree.
On the day of the first snow paint a canvas red.
Dress in a way that only a ghost could love.
Make pancakes out of grass at midnight each night.
Find old man Winter, hug him and let him know everything will be ok.
Ride out of town holding a pelican.
Under every rock I come across for a month I will write "all power to the rocks".
Every day for a year, at a random time, shout "sausage".
Make a film about the last sock in the world.
Put on a red shirt and scream 'I'M NOT WEARING PANTS!' every time I leave the house.
Throw a party for insects.
Try to convince the dog next door that he is wearing a coat of moonlight.
Every time I press a button I will say 'this is my favorite'.
Search my apartment for secret doors or hidden staircases.
Wear two superman outfits at the same time.
Every time it rains I will stir my tea anti-clockwise.
Every night for a week I will wear a hat lined with lettuce.
I will begin to believe that the trees that I see everyday are my friends.
Every time a bird flies past me I will remember to breathe.
Throw a birthday party for my favorite tree.
I will from now on tell every dog I meet that I am training to be a dragon.
Every time I see a panel van pass me I will dub it a "Slice-a-Wagon."
Crawl on the ceiling like a spider for a month.
Attempt to find peace living with an army of puppets.
Wear a dinosaur costume to every public event I attend.
Go to the beach every day for a week and shout the names of colors into the ocean.
Go on a three-day backpacking trip dressed as a turnip.
Create messages that only the wind can hear by blowing on the blades of grass.
Give a piece of cloud to a complete stranger.
Make a mask out of grass and wear it while I'm sleeping.
I will now treat every worm I see as if it is an old friend.
When I hear a strange noise in an empty room I will assume someone is saying hello to me.
At the Smithsonian AIB website is a generator that I've populated with other resolutions from my list. Visit their site and grab yourself a random resolution. If you don't like the one you got, you have my permission to reload and regenerate until you find a resolution that speaks to you. Or to apples.
Your January 2022 Smithsonian Futurist in Residence,
Janelle

Christmas entities

By Janelle Shane
December 22, 2021

When you think about it, Christmas can get pretty weird.
There's the classic Christmas story of the Bible, and then there are all these extra entities that aren't in the book but which become somehow part of Christmas. And some of them are quite unsettling. There's the immortal flying reindeer with the bioluminescent nose. Iceland has the Yule Cat, a building-sized cat that stalks the streets and devours anyone who hasn't received new clothes for Christmas. And there's Krampus, the fanged, goat-horned being that looks more like it came from the cover of a metal album than a Christmas album. I am not making any of these up.
But I wondered what kinds of new Christmas entities might result if I gave a list of these and short descriptions to a neural net trained on internet text. New Christmas entities, perhaps, for our modern age?
I used GPT-3, giving it the examples of Krampus, Rudolph, The Yule Cat, and Meathook (one of the Yule Lads; not as terrifying as he sounds), and asking it to add its predictions to the list.
Here are some of my favorites: