"Janelle shane"

The problem with artificial intelligence is that it will do exactly what we ask it to do

Janelle Shane is an AI researcher. In this TED talk she explains that we should not be afraid that AIs are going to rebel against us. We should be afraid of AIs because they are going to do exactly what we tell them to do. "It's really easy to accidentally give AI the wrong problem to solve," she says, "and often we don't realize that until something has actually gone wrong." Read the rest

Neural net-generated prompts for Inktober

It's Inkotober, when "artists all over the world take on the Inktober drawing challenge by doing one ink drawing a day the entire month." In a fun experiment, Janelle Shane trained a neural net with prior Inktober prompts and picked out some promising concepts like "ornery beach sheep" and "BUG IN HUMAN SHAPE."

If you'd like to participate in the fun, pick one of the prompts and post your illustration for a chance to win an advance copy of Shane's new book:

My US and UK publishers are letting me give away some copies of my book to people who draw the AInktober prompts - tag your drawings with AInktober and every week I’ll choose a few people based on *handwaves* criteria to get an advance copy of my book. (US, UK, and Canada only, sorry)

Here are a few entries to date:

Read the rest

Mutant fly strains invented by a neural net

Janelle Shane, harbinger of the generative apocalypse, presents mutant fruit flies invented by a neural network: "Nothing to worry about."

I knew that fruit flies are a mainstay of research labs, but I had never given them much thought until Prof. Greg Neely emailed me to point out how weird the names of mutant fruit flies are. There’s a strain of mutants called “Out Cold” where the fly loses motor coordination below a certain temperature, and another nicknamed “Moonwalker” that walks backwards. Could a neural network learn how to invent new names and mutations?

An unsettling reminder that science fiction isn't a prediction, but the present placed in uncanny focus.

P.S. AI Weirdness is proof that blogging is still great) Read the rest

AI-generated pokemon that should not be

Michael Friesen generated these abominable pokemon sprites. Be sure to see a similar set hand-drawn by iguanamouth. [via Janelle Shane]

Read the rest

Neural network cookies

"Aw yeah it's time for cookies," writes AI ringmaster Janelle Shane (previously at BB).

One neural network I use, called textgenrnn, tries its best to imitate any kind of text you give it. I’ve given them paint colors, band names, and even guinea pig names and in each case their results are somewhat… mixed. (Paint colors called Stanky Bean, Stargoon, and Turdly, for example) The problem is that it doesn’t know what any of these words mean - it’s just picking letter combinations that seem likely to it.

This is what happened when I gave it all the cookies from a list of American recipes. This is what human cookies sound like to a neural network.

Previously, previously. Read the rest

Check out these machine-learned Burning Man camp names

The theme of this year's Burning Man is I, Robot, which focuses "on the many forms of artificial intelligence that permeate our lives..." So, naturally, someone trained a neural network to come up with some camp names.

It spit out believable names like Spankles, Astro Sparkin, and Space Rock Screamin Camp, as well as weirder names like Corn Viral Hammers, Wiq Renames Spaghette, and Hellball Lounge. Then it went with some truly bizarre ones like Cohnie Stacefur Ass Chaos, Sir Liberains the Wreck Middle, and Awes Orpoop.

The woman behind the experiment, research scientist Janelle Shane, writes:

Thanks to an anonymous burner, I had a list of 1593 past Burning Man camps to feed to a neural network. A neural network is a kind of machine learning algorithm that learns to imitate the data it sees. My starting point was a textgen-rnn neural net that had been previously trained on metal bands and roller derby names, so it had a few ideas of its own to bring to the table. It did not disappoint.

There's a bunch more of these machine-learned camp names over at Shane's site.

Let's hope life imitates art and some Burners out there actually create one (or more) of these camps this year on the playa!

Image via simon of the playa

Thanks, Dan S.! Read the rest

An ice-cream maker tries to figure out what AI ice-cream flavors derived from metal band-names would taste like

Janelle Shane (previously) is a delightful AI researcher who likes to use machine learning systems to produce absurd, inhuman outputs, such as a list of AI-created notional ice-cream flavors generated by merging a list of real ice-cream flavors with a list of metal band names and pressing "go." Read the rest

"Skyknit": Knitting patterns produced by a neural net

Janelle Shane of AI Weirdness is awesome: She's trained neural nets to invent all sorts of hilarious material, from the names of new colors to odd new food recipes to original Dungeons and Dragons spells.

Recently she decided to train a neural net on knitting patterns, and it began spitting out new ones. They were predictably strange, and to get a sense of precisely how strange, the fine folks at Ravelry -- a discussion site for those who knit, crochet, weave, and the like -- offered to actually produce some of patterns IRL. They've dubbed it "Skyknit".

If you belong to Ravelry you can see the results in the thread, but in case you aren't, behold pix of some of the creations above and below. The one above was knitted by MeganAnn, and it looks vaguely... organic? You can see some repeating patterns in it, but mashed together in a pretty strange fashion.

Here's another one knitted by MeganAnn ...

That one looks a bit more like a human-crafted pattern. Here's a really strange one made by datasock ...

... and winding-serpent pattern crafted by michaela112358 ...

Here's a pretty one by BellaG:

A sort of ... undersea creature? Created by also by michaela112358 ...

... and some that are pretty normal-looking! Like this one by geckogirl ...

... or this one by booksprink ...

... or this last one, by Farah Colchester:

Janelle Shane did a tweetstorm where she talked about the experience of watching the Ravelry crowd bring these things to life. Read the rest

What happens when a neural network proposes legislation?

In her delightful blog AI Weirdness, Janelle Shane entered 18,458 unique bills introduced in Massachusetts into a neural network, which then created some rather hilarious bills, including: Read the rest

My Little Ponies generated by neural network

Janelle Shane‏ trained a neural network on the names and attributes of My Little Ponies, then shared "some of the worst ones."

I used a program called a character-level recurrent neural network (char-rnn), which looks at examples of text (Pokemon, or Harry Potter fan fiction, or even guinea pig names) and learns to imitate them. I gave the neural network more than 1,500 names from My Little Pony Friendship is Magic Wiki, and let it start learning.

Read the rest

Robot wisdom from a deep learning system trained on ancient proverbs

Janelle Shane trained a recurrent neural network with a data-set of more than 2000 ancient proverbs and asked it to think up its own: "A fox smells it better than a fool’s for a day." Read the rest

Dungeons & Dragons spells created by a neural network

Neural networks, it is said, cannot explain their decisions. Which is probably a good thing, at least when it comes to the machine mind's ideas for new Dungeons & Dragons spells, as guided by Janelle Shane. [via Patrick Ziselberger‏]

It’s a really small dataset, actually - so small that in almost no time at all, it learned to reproduce the original input data verbatim, in order. But by setting the “temperature” flag to a really high value (i.e. it has a higher chance of NOT going with its best guess for the next character in the phrase), I can at least induce spelling mistakes. Then the neural network has to try to recover from these, with often entertaining results.

Moss Healing Word Hold Mouse Barking Sphere Heat on Farm True Steake Finger of Enftebtemang Fomend’s Beating Sphere

For the best one you'll have to click through. Read the rest

A neural network generated these can't-fail pickup lines

Neural nets are starting to wake up. These pickup lines, generated by a neural net maintained by research scientist Janelle Shane are much more interesting than standard pickup lines.

Are you a 4loce? Because you’re so hot!

I want to get my heart with you.

You are so beautiful that you know what I mean.

I have a cenver? Because I just stowe must your worms.

Hey baby, I’m swirked to gave ever to say it for drive.

If I were to ask you out?

You must be a tringle? Cause you’re the only thing here.

I’m not on your wears, but I want to see your start.

You are so beautiful that you make me feel better to see you.

Hey baby, you’re to be a key? Because I can bear your toot?

I don’t know you.

I have to give you a book, because you’re the only thing in your eyes.

Are you a candle? Because you’re so hot of the looks with you.

I want to see you to my heart.

If I had a rose for every time I thought of you, I have a price tighting.

I have a really falling for you.

Your beauty have a fine to me.

Are you a camera? Because I want to see the most beautiful than you.

I had a come to got your heart.

You’re so beautiful that you say a bat on me and baby.

You look like a thing and I love you.

Hello.

Read the rest

Horspest and Shivercell: a neural net invents names for fruit varieties

Research scientist Janelle Shane writes: "I’ve been training a neural network (based on this open-source neural network framework from Andrej Karpathy) on datasets from recipes, to lists of Pokemon, to superhero names. I decided to see if it could invent names for new fruit varieties - I fed it a list of apple, peach, pear, plum, and cherry varieties, and asked it to generate more."

Sunbrown Stanker Pork Gala Horspest Shivercell Hencough Moregall Brown Soften Ruby Wally Ruck Nagtort Blee Red Redcells Zuby Glong Zeelcher Hacker Gala Soften Fuji Klunk 134 Horking Read the rest

Neural network comes up with crazy food recipes

In her spare time, University of California, San Diego engineer Janelle Shane trained a neural network to generate recipes for new dishes. Informed by its reading of existing recipes, the neural network did improve over time yet it's clearly not quite ready for Iron Chef. Here are two recipes from her Tumblr, Postcards from the Frontiers of Science:

Pears Or To Garnestmeam

meats

¼ lb bones or fresh bread; optional½ cup flour1 teaspoon vinegar¼ teaspoon lime juice2  eggs

Brown salmon in oil. Add creamed meat and another deep mixture.

Discard filets. Discard head and turn into a nonstick spice. Pour 4 eggs onto clean a thin fat to sink halves.

Brush each with roast and refrigerate.  Lay tart in deep baking dish in chipec sweet body; cut oof with crosswise and onions.  Remove peas and place in a 4-dgg serving. Cover lightly with plastic wrap.  Chill in refrigerator until casseroles are tender and ridges done.  Serve immediately in sugar may be added 2 handles overginger or with boiling water until very cracker pudding is hot.

Yield: 4 servings

This is from a network that’s been trained for a relatively long time - starting from a complete unawareness of whether it’s looking at prose or code, English or Spanish, etc, it’s already got a lot of the vocabulary and structure worked out. 

This is particularly impressive given that it has the memory of a goldfish - it can only analyze 65 characters at a time, so by the time it begins the instructions, the recipe title has already passed out of its memory, and it has to guess what it’s making.

Read the rest

Crimm Grunk Garlic Cleas - and other recipes created by a neural network

Research scientist Janelle Shane has been training a neural network to generate food recipes by giving it tens of thousands of cookbook recipes. The neural net's recipes are excellent:

Beef Soup With Swamp Peef And Cheese Chocolate Chops & Chocolate Chips Crimm Grunk Garlic Cleas Beasy Mist Export Bean Spoons In Pie-Shell, Top If Spoon and Whip The Mustard Chocolate Pickle Sauce Whole Chicken Cookies Salmon Beef Style Chicken Bottom Star * Cover Meats Out Of Meat Completely Meat Circle Completely Meat Chocolate Pie Cabbage Pot Cookies Artichoke Gelatin Dogs Crockpot Cold Water

Read the rest

:)