In a previous experiment, I let a neural network trained on the complete works of H. P. Lovecraft finish phrases from cookbook recipes. Now, I tried it the other way around, in which I gave phrases from Lovecraftian horror to an innocent neural network trained on 30MB of cookbook recipes.
There was thunder in the air on the night I went to the deserted mansion atop Tempest Mountain to find the cake cooked.
I was not alone, for foolhardiness was not then mixed with the ham slices.
Now and then, beneath the brown pall of leaves that rotted and festered in the antediluvian forest darkness, I could trace the sinister outlines of some of the cooking pancakes.
For I, and I only, know what manner of fear lurked on a cookie cutter.
The pitiful throngs of natives shrieked and whined of the unnamable powder served with the flour and red pepper.
Everything seemed to me tainted with a loathsome contagion, and inspired by a noxious alliance with the steamed chicken.
All was in vain; the death that had come had left no trace save the steamed red peppers and chicken broth.
Sometimes, in the throes of a nightmare when unseen powers whirl one over the roofs of strange dead cities toward the grinning chasm of Nis, it is a relief and even a delight to make the soup.
There are a lot of strange courses that make it into a college course catalog. What would artificial intelligence make of them?
I train machine learning programs called neural networks to try to imitate human things - human things they are absolutely are not prepared to understand. I’ve trained them to generate paint colors (Shy Bather or Stanky Bean, anyone?) and cat names (Mr. Tinkles is very affectionate) and even pie (please have a slice of Cromberry Yas). Could it have similar “success” at inventing new college courses?
UC San Diego’s Triton alumni magazine gave me UCSD’s entire course catalog, from “A Glimpse into Acting” to “Zionism and Post Zionism”, a few of which I recognized from when I was a grad student at UCSD. (Apparently I totally missed my opportunity to take “What the *#!?: An uncensored introduction to language”) I gave the course catalog to a neural network framework called textgenrnn which took a look at all the existing courses and tried its best to figure out how to make more like them.
It did come up with some intriguing courses. I’m not sure what these are, but I would at least read the course description.
Strange and Modern Biology Marine Writing General Almosts of Anthropology Werestory Deathchip Study Advanced Smiling Equations Genies and Engineering Language of Circus Processing Practicum Geology-Love Electronics of Faces Marine Structures Devilogy Psychology of Pictures in Archaeology Melodic Studies in Collegine Mathematics
These next ones definitely sound as if they were written by a computer. Since this algorithm learns by example, any phrase, word, or even part of word that it sees repeatedly is likely to become one of its favorites. It knows that “istics” and “ing” both go at the end of words. But it doesn’t know which words, since it doesn’t know what words actually mean. It’s hard to tell if it’s trying to invent new college courses, or trying to make fun of them.
Advanced Computational Collegy The Papering II The Special Research Introduction to Oceanies Biologrative Studies Professional Professional Pattering II Every Methods Introduction study to the Advanced Practices Computer Programmic Mathematics of Paths Paperistics Media I Full Sciences Chemistry of Chemistry Internship to the Great The Sciences of Prettyniss Secrets Health Survivery Introduction to Economic Projects and Advanced Care and Station Amazies Geophing and Braining Marine Computational Secretites
It’s anyone’s guess what these next courses are, though, or what their prerequisites could possibly be. At least when you’re out looking for a job, you’ll be the only one with experience in programpineerstance.
Ancient Anthlographychology Design and Equilitistry The Boplecters Numbling Hiss I Advanced Indeptics and Techniques Introduction in the Nano Care Practice of Planetical Stories Ethemishing Health Analysis in Several Special Computer Plantinary III Field Complexity in Computational Electrical Marketineering and Biology Applechology: Media The Conseminacy The Sun Programpineerstance and Development Egglish Computational Human Analysis Advanced A World Globbilian Applications Ethrography in Topics in the Chin Seminar Seminar and Contemporary & Archase Acoa-Bloop African Computational for Project Laboration and Market for Plun: Oceanography
Remember, artificial intelligence is the future! And without a strong background in Globbilian Applications, you’ll be left totally behind.
Just to see what would happen, I also did an experiment where I trained the neural net both on UCSD courses and on Dungeons and Dragons spells. The result was indeed quite weird. To read that set of courses (as well as optionally to get bonus material every time I post), enter your email here.
These are some of the most amazing generated images I’ve ever seen. Introducing BigGAN, a neural network that generates high-resolution, sometimes photorealistic, imitations of photos it’s seen. None of the images below are real - they’re all generated by BigGAN.
Preprints of the BigGAN paper are here and here. It’s been causing a buzz in the machine learning community. For generated images, their 512x512 pixel resolution is high, and they scored impressively well on a standard benchmark known as Inception. They were able to scale up to huge processing power (512 TPUv3′s), and they’ve also introduced some strategies that help them achieve both photorealism and variety. (They also told us what *didn’t* work, which was nice of them.) Some of the images are so good that the researchers had to check the original ImageNet dataset to make sure it hadn’t simply copied one of its training images - it hadn’t.
Now, the images above were selected for the paper because they’re especially impressive. BigGAN does well on common objects like dogs and simple landscapes where the pose is pretty consistent, and less well on rarer, more-varied things like crowds. But the researchers also posted a huge set of example BigGAN images and some of the less photorealistic ones are the most interesting.
I’m pretty sure this is how clocks look in my dreams. BigGAN’s writing generally looks like this, maybe an attempt to reconcile the variety of alphabets and characters in its dataset. And Generative Adversarial Networks (and BigGAN is no exception) have trouble counting things. So clocks end up with too many hands, spiders and frogs end up with too many eyes and legs, and the occasional train has two ends.
And its humans… the problem is that we’re really attuned to look for things that are slightly “off” in the faces and bodies of other humans. Even though BigGAN did a comparatively “good job” with these, we are so deep in the uncanny valley that the effect is utterly distressing.
So let’s quickly scroll past BigGAN’s humans and look at some of its other generated images, many of which I find strangely, gloriously beautiful.
Its landscapes and cityscapes, for example, often follow rules of composition and lighting that it learned from the dataset, and the result is both familiar and deeply weird.
Its attempts to reproduce human devices (washing machines? furnaces?) often result in an aesthetic I find very compelling. I would totally watch a movie that looked like this.
It even manages to imitate macro-like soft focus. I don’t know what these tiny objects are, and they’re possibly haunted, but I want them.
Even the most ordinary of objects become interesting and otherworldly. These are a shopping cart, a spiderweb, and socks.
Some of these pictures are definitely beautiful, or haunting, or weirdly appealing. Is this art? BigGAN isn’t creating these with any sort of intent - it’s just imitating the data it sees. And although some artists curate their own datasets so that they can produce GANs with carefully designed artistic results, BigGAN’s training dataset was simply ImageNet, a huge all-purpose utilitarian dataset used to train all kinds of image-handling algorithms.
But the human endeavor of going through BigGAN’s output and looking for compelling images, or collecting them to tell a story or send a message - like I’ve done here - that’s definitely an artistic act. You could illustrate a story this way, or make a hauntingly beautiful movie set. It all depends on the dataset you collect, and the outputs you choose. And that, I think, is where algorithms like BigGAN are going to change human art - not by replacing human artists, but by becoming a powerful new collaborative tool.
The BigGAN authors have posted over 1GB of these images, and it’s so fun to go through them. I’ve collected a few more of my favorites - you can read them (and optionally get bonus material every time I post) by entering your email here.
At Burlington City Arts, Crystal Wagner’s first-ever single work existing in both the interior and exterior of a space comes with “Traverse.” See more on HiFructose.com.
Machine learning algorithms can uncover complex patterns in the data they see, making them useful for image recognition, predicting customer service questions, or recommending movies. They can even do a decent job at naming craftbeers, kittens, or guinea pigs. But one thing it turns out they’re bad at? Understanding what humans find sexy.
I had my first sign that this was a problem when I trained a neural network to generate new Halloween costumes and saw its attempts at the “sexy” category of names - it came up with ideas like Sexy Gargles, Pretty zombie Space Suit, and Sexy the Spock. So when Scarlett O'Hairdye contacted me saying they were putting together an AI-themed burlesque show (yes you read that right), and asked me to train a neural network to generate possible names for the show… I knew the neural network was going to be in for a confusing time.
Now first let me talk about burlesque. If you’re not familiar with it, think feather boas, ruffled skirts, and fishnet stockings. These days, themed burlesque shows are all the rage, with names like “That Ass, Poor Yorick” “Star Trek: The Sexed Generation” and “Burl-X Files”. Scarlett provided me with 450 examples of existing shows and yes, the neural network proceeded to get very confused.
One thing it tried was making up words that sounded to it like sexytimes. It made no sense, but it was strangely adorable.
Booky Ampitions - A Stravaganza Starstox! A Burlesque 2 Booms A Shagack! SPOW! Holiday Fishing Glasties off! Moosters, A burlesque tribute Homper Gurder Burlesque Show Show! Thag Ag After Dark Woncerless! Boodnass Tronpboons if Mongerland Bonshows of thong Yes of Nevering Eightthows! MACTAON! A Nighty Boosh Burlesque! Deeptert! “Thawls of Vinderland II - A Burrrrs?! Burlesque Revue” BUR! The Sexed Garks of Burlesque Adventure
Sometimes the sexy-sounding words it generated were already other words. For some reason it was trying hard to make vases sexy. It has an even harder task ahead of it with its other favorite words, “warts” and “fart”.
My 2017 Farty Burlesque Adventure Vase Burlesque Revue The Warts of Burlesque! The Wonderland - March of Farty Fundraiser Teaks of Fame Legends Tree! A Burlesque Revue Vase Show Gourdraiser! Sex-Pone Cabaret Sticker Burlesque Burlesque Show The Pans of The Panners Burlesque Revie The Adomic Eso Space Scream Show The Hare and the Rare and the Mar Chas Burlesque Revue Farty Fasties: A Burlesque Show The Rank Show Vase & Show Vase Farts The Stripper Stripper Dave Burlesque Show Adventure To Burlesque Came Farts of Burlesque Revue Seattle Burlesque Show & Tangy Future and Warty For the Blue Door
Here I think it was trying to spell “boobs” or maybe “bombshells” but had a bit of an issue.
Burlesque Bonbs and Constray Burlesque Borbshells Burlesque Borbs and Monstrous Burlesque Show Bolbshells!
Other times it got the words right, but used them rather… unskillfully.
Sex Your Eye Out! The Parts and Burlesque Revue The Sexed Show The Pank: A Burlesque Revue Peepsing Tarts Burlesque Show A Hot Care Show! Well New Cheapless! The Sexies of Burlesque Revue The Hand Show Burlesque Show About Your Peek Show Derrierer: A Burlesque Show Cone With 9s Cabaret Derriere The Pants of Fame Burlesque Adventure
And the name that was chosen? May I present to you the first-ever AI-themed burlesque show:
If you are lucky enough to be in Seattle, WA on July 21, 2018 (and are over 21), you can experience some of the strangest sexytimes that technology has to offer. Tickets!
This one is loosely based on walking around my hometown at night; but not a specific time, more of drawing from the mood and setting and reasons from many times in the past
.
.
.
#art #artwork #painting #inkdrawing #mixedmedia #ink #artist
Another loose, kinda experimentation-focused piece. I’ve been drawing from some sketches and doodles, and just building layers without a set plan. This one is a little more energetic than the last one
.
.
.
#painting #artist #artwork #art
I’ve been adding layers to this for the past week or so and I’m not sure where I’m going with it but it’s been fun getting back into any kind of painting again
.
.
.
#painting #mixedmedia #art #artwork #artist
Happy Pride Month, everyone! I’ll be spending the weekend in the Utah Pride Festival’s 1 to 5 Club booth, sheltering from the sun and hopefully working on some art! I’ll also be marching in the parade: I’ll be the one trying to avoid poking other people with my gothy parasol. I hope everyone has a wonderful Pride!
Just a short piece of writing from last night that was mostly just stream-of-conciousness and not that good (I’m a visual artist not a writer) but there were some interesting pieces in there that I liked:
I feel like I’m still rediscovering my voice
As an artist
As a person
I didn’t have one for so long
Not an authentic one
Not mine
Not Me
Maybe there isn'tone out there
Even recently I was too far gone
Too close to silence
Even now I know I don’t have it in me to be strong
I’ll try
Sure
But it’s a long journey
It’s only the journey, nothing else
After all these years I realized the real me was silent
Fake to everyone
Just what they expected
And to myself
Just what I thought I deserved
Never truly dreaming
But now maybe awake
Even if this is a nightmare
A nightmare is still a dream
Not a void, something of substance
Not desirable, something to fear
But better than nothing
