But what about the artistic side of A.I.?
Some researchers want to discover what happens when they give their neural networks a chance to show their creative side -- and the results are surreal to say the least.
Neural networks go naked
A.I. researcher Robbie Barrat fed a generative adversarial network (GAN) thousands of nude portraits from a dataset and then trained it to create its own version of the artwork it processed. The resulting A.I. art of surreal, swirly naked female forms would impress even the likes of Salvador Dali.
Generative adversarial networks are artificial intelligence algorithms that are used in unsupervised machine learning. A GAN uses two different neural networks -- a generator and a discriminator.
The generator tries to come up with images that fool the discriminator, and the discriminator tries to learn how to tell the difference between real images from the dataset and fake images the generator feeds it. Over time, the more realistic the output will be -- in this case nude paintings.
However, sometimes the generator and discriminator will keep trying to fool each other without actually getting better at the task at hand. And in the case of Barrat's neural network, the artwork looks like a surrealist nightmare.
Unhappy little trees
The MIT-trained roboticist and artist Alexander Reben uses his art to explore what goes wrong when machine learning filters a seemingly innocent episode of the PBS retro TV show The Joy of Painting through the neural net.
In the video Deeply Artificial Trees by Reben's artist alter ego artBoffin, we see the PBS show host, artist Bob Ross, looking less like the '70s art icon who loved to paint calming landscapes and more like a character featured in a bad acid trip.
The video shows the "happy, little trees" Ross is famous for painting reimagined as horrific-looking Lovecraftian insects summoned by the fictional grimoire The Necronomicon.
"This artwork represents what it would be like for an A.I. to watch Bob Ross on LSD (once someone invents digital drugs)," artBoffin writes in the video description. "It shows some of the unreasonable effectiveness and strange inner workings of deep learning systems. The unique characteristics of the human voice are learned and generated, as well as hallucinations of a system trying to find images which are not there."
A.I. explores dragons and romance
Sometimes dragons and candy hearts are the perfect muse for a neural network. Janelle Shane, a research engineer for an optics company who also likes to experiment with neural-network programming, trained a machine-learning system to create new monsters for the fantasy tabletop game Dungeons & Dragons.
Shane gathered 2,205 creature names from the second-edition Dungeons & Dragons monster manual, and her neural network transformed those names into new imaginary creatures like a Curple Lard Dragon, Wolfworm, Spectral Slug, Jabberwont, Burglestar, and Marraganralleraith.
She then created a recurrent neural-network algorithm to generate D&D spells like Barking Sphere, Hold Mouse, and Gland Growth.
Shane also trained her machine-learning system to generate new romantic phrases for Valentine's Day candy hearts. Instead of the typical "Be Mine" message, the A.I. came up with "Hot Give," "Team Bear," and "Time Hug."
The surrealist art movement seems boringly normal compared with this new wave of A.I. artists. Here's hoping one day, art museums everywhere will have a wing dedicated to the unusual art of neural networks, and the humans who trained them.