Over the past few years there has been a lot of attention devoted to the idea that AI — artificial intelligence — will eventually be writing fiction. There’s tons of example in the Twitter and YouTube communities, mostly due to the efforts of comedy writer Keaton Patti, who uses predictive text to compose random, hilarious bits of writing based on infomercials, Hallmark movies, and the like. They’re entertaining, but often attributed to a bot forced to watched movies or TV commercials. Nothing could be further than the truth. Predictive text as exists on a cell phone and a neural network trained to read, and generate, text are two different things. Neural network text tends to meander off topic, according to Janelle Shane, and shows a lack of meaning and context. In other words, it doesn’t quite all fit together. At the present point, anything it generates still needs human tweaking to become readable, and amusing, and not a random word salad.
Before InferKit went to pay-only ($20 a month to use the engine, which I can’t afford) I generated a ton of random nonsense. Every once in a while the neural network came through out something unexpected. Like this bit of a top-secret memo.
 |
To: RvB-Scot
Just remember the following; His Majesty’s Air Force bombed CERBERUS to bring her up to Standard Operating Procedures (SOP). And being a Black Uncle Tom, it worked brilliantly. |
 |
… or this piece of literary fiction (tweaked by me.)
|
All over, the redwood forests are burning. Woods are stripped and harvests burned; I saw the sun fall like a dragon’s breath. The sound of the fire makes me paralyzed and the whup whup whup of burning wood, some flames with faces like babies. | |
I was able to string a group of thematically connected phrases into song lyrics that could have been written by the Canadian rock group Rush.
|
CONNECTION AND SPEED
Hopelessly lost on the moon, |
|
Mostly, though, the random pieces — paragraphs and sentences standing out in a list — were parts of reviews for movies, albums, books, or DVDs.
|
On the Slopes also features songs such as “A Lighter Shade of Blue,” “Sweet Simple Kettle,” “Moon to Go Blue” and “Let Me Take Your Picture.” During this trilogy of 5 discs, the Twilight movies will only appear on one side of the disc. However, each (side) of the 1 disc is an animated horror movie. | |
|
In a scene worthy of a filmmaker’s wildest imagination, the cousins marvel as the hypnotic beings merge into a grinning formation: Three dwarfs, a gothic crossdresser, a recently deceased mermaid and a magician. | |
|
PULSIVE ARTIST PORTRAITS, by Olafur Eliasson. Scenes from the author’s experiences of living with and, increasingly, alongside the disabled. |
|
The review below was as long as it got.
|
Teach Me to Read It comes from Scottish author Ewan Morrison (my usual fave.) It’s a historical tale of swordfighter Ewan Marnin, who goes on a treasure hunt around the world. This limited edition features scenes from all the books, though here I only saw the action from Journey to the Sunset in particular (maybe it was because it was the one I read first?) | |
Predictive text generates more coherent pieces. Using the Harry Potter keyboard from the robotnik.com site, I generated this short Harry Potter chapter, which sort of makes sense if you squint at it sideways.
|
Harry is not quiet with his magic
|
|
