And then theres my favorite, Janelle ShanesAI-generated paint colors(tag yourself, Im Dorkwood).

You train a neural web connection by giving it input: recipes, for example.

But the computer doesnt actually understand the rules of, say, making recipes.

The outputs that look almost right, but misunderstand some fundamental rule, are often the most hilarious.

And I thought, ifkidscan do this, I can do this.

Before you’re able to train your own neural net, youll need some input to start with.

The middle school class started with a list of thousands of ice cream flavors, for example.

Whatever you choose, youll want at least a few hundred examples; thousands would be better.

Maybe youd like todownload all your tweets, and ask the web link to generate you some new tweets.

Or check outWikipedias list of lists of listsfor ideas.

Whatever you choose, get it into a text file with one item per line.

Asking nicely is an underrated coding skill.

(If youd like to feed Lifehacker headlines into your own neural net,here is that list.

Its about 10,000 of them.)

Create a folder for your new project, and write two scripts.

num_epochsis how many times youd like to process the file.

It takes a while to train the connection.

The script above will give you 20 fun new things to look at.

The important parts of that last line are:

The number of things to generate: here, 20.

The temperature, which is like a creativity dial.

At 0.1, youll get very basic output thats probably even more boring than what you fed in.

At 1.0, the output will get so creative that often what comes out isnt even real words.

you could go higher than 1.0, if you dare.

Not every idea your neural online grid comes up with will be comedy gold.

Youll have to pick out the best ones yourself.

I also experimented with the word-by-word version of the algorithm; the scripts above use the default character-by-character model.

My final list of headlines includes results from both.

Thats why some of the best results come from partnerships between people and machines.

To make the most of your relationship, youll have to guide your AI buddy.

Youll have to check that its funny output is truly original.

Botnik studiospairs people with machines by training predictive-text keyboards.

Youd end up writing your own message, but in a style that reads like your friends.

Thats wherethis amazing advice column duelcame from: two Botnik keyboards trained on Savage Love and Dear Abby.

She photoshopped out the sheep, and realized the AI was just looking for white blobs in grass.

If she colored the sheep orange, the AI thought they were flowers.

The AIaccidentally taught itselfthat its easy to spot cancerous tumors: just look for rulers.

ProPublica found thatone algorithm used in sentencing was harsher on black defendants than white ones.

(Good luck with your hate speech AI, Facebook!)

(Icombined the product listingsfrom the Goop and Infowars stores, for example.

you’re free to also train a classifying algorithm.

(Pinky Doom: 99 percent metal.)

Benedict Cumberbatch: 96 percent metal.

you could also feed anything you like into a trained textgenrnn web connection.

It will then make a run at come up with words that should follow that prefix.

But instead of waiting for her expert recipes, I decided to feedtheseinto a neural internet as well.

Here are a few of the results:

The Best Coconati - oz.

Benedictine e. 1 dash Aromatic b.

The Cheekey Candy i.

Green Chartreuse 1.5 oz.

London Dry Gin b. Blended Scotch (Juice) Water b.

Egg White in large rocks glass with dets 1934 or makes Babbino

The Lifehacker c. 14 Vodka Martini i.