Neural Networks for Divination

For the past few months I have been running an experiment on using an image generative neural network as a method of divination. Neural networks are computer programs which are intended to model biological neurons in a brain, and are capable of learning to create and categorize things that are not directly a part of their code. They work by creating multiple grids of numbers which are linked to each other via specific math equations. Data is fed into them, gets processed through those equations, and comes out changed on the other side. They “learn” by being fed a specific set of data that has specific intended results. The results are compared to the intended results, and then the grids of numbers are subtly nudged until the intended result (or something acceptably close to that result) is produced.

The specific network I have been using, Deep Daze, is set up to generate images from text that is input into it. It is pre-trained, and I have not done any additional training on the network. My method has been to input text which is not a direct description of an intended image, but instead is an obfuscation (in that the pre-trained data set the neural network was taught likely does not contain the details of what I’ve written) of a topic. This is similar to Remote Viewing, where the target is obfuscated by a number or symbol. The network has no “memory” of previous text input or results of previous tests.

The following are some of the more interesting results I got.

Many of the images generated are related to the Traveling Museum of the Paranormal and Occult by Greg and Dana Newkirk of Hellier fame. They recently conducted an experiment involving UFOS and alien abductions which included a target number and intentions. For privacy reasons I will not include the details of this experiment nor the target number here. I entered the target number into the neural network, set my personal intention to get an image related to the experiment, and let it run. This is the final image it generated.

We see a number of strange faces here, in various stages of formation. This particular image could likely have benefited from a longer run time, but I am limited in how long I can let any individual test run. In theory, the network should have no knowledge of any members of the Paranormal Museum, their technology, or anything about the experiment at all, especially since all it was told was a series of numbers. There appears to be a box, on which is written an overlapping series of letters that may be forming “Greg Newkirk” and “Franks Box“. The Franks Box is a type of frequency sweeping radio which some paranormal investigators use to communicate with spirits or ultraterrestrials.

The next image was generated with the phrase “The Traveling Museum of the Paranormal and Occult”


The results of this one are less surprising. We see images fit for a museum display, as well as paranormal imagery. Variations of the input phrase are visible. It is impressive that the network was able to parse the intention such that the image has a creepy vibe to it, but nothing particularly out of the ordinary.

This image was generated with the phrase “Strange Frequencies“, the title of a book I had recently read.


We see the input phrase followed by a set of numbers, as well as imagery corresponding to wave patterns common to frequency visualization. We also see a building and a number of people. One appears to be in a meditative pose. These are interesting results, as the book is about using technology to interact with spirits.

The next two images were given the phrases “Bigfoot Pheromones” and “Greg Newkirk tastes bigfoot pheromones while Dana films it”.


These phrases are references to a charity drive the Newkirks did, wherein Greg offered to taste bigfoot pheromones if a certain donation threshold was reached. While the results are humorous, the second image is interesting as it contains faces which look reasonably like Greg and Dana with unpleasant expressions.

The neural network has no reason to know what Greg or Dana looks like. It also shows a man’s head on a furry body. Greg surprised everyone by wearing a bigfoot costume when he tasted the pheromones, and it was not known that this would happen at the time of this image’s creation.

This image was given the phrase “Geg and Danar talk to ghosts”. There is also a short video showing the generation of the image over time.


“Geg and Danar” are silly alter-egos of Greg and Dana Newkirk. We see here people seated on what may be a couch in a spooky looking location. The letters GD are between them, as are three ghosts (that we can see). The people look, again, suspiciously like Greg and Dana. Also of note is that they appear to be wearing headphones and have blacked out eyes, which is common to the Estes Method (a method of spirit mediumship using a sweeper radio, headphones, and a blindfold, invented by Connor Randall and Karl Pfeiffer.)

The phrase for the following image and video was “Billy, the N’kisi of Dreams”.


Billy is an N’Kisi idol that is part of the Paranormal Museum. N’Kisi are objects used throughout the Congo Basin of Central Africa to house spirits. Here we see a face among a field of stars and moons, as well as other heads which may be carved out of wood. Could this be the real face of Billy?

The phrase for this image was “Filming a documentary about The Crone”.


The Crone is another haunted object (formerly) owned by the Paranormal Museum. Here we see a camera and hand pointing at two individuals that appear to be a man and woman with a statue between them. The neural network should have no information about the fact that The Crone took the form of a small wooden idol, or that the documentary is being made by Greg and Dana. When the documentary comes out, I’d love to see if it contains anything like the carpet they are standing on in this image.

The phrase for the next image was “Why can’t I hold all these ghouls?”


A joke based on an offhand comment made by Greg during one of the Paranormal Museum’s livestreams. The image contains what seem to be a mass of skeletal figures, a jack-o-lantern face, and a figure who looks uncomfortably like an out-of-focus Greg.

“The Worm Man of Waverly Hills”

This is a phrase based on an investigation that Dana Newkirk and Tyler Strand did of Waverly Hills Sanatorium, where the name “worm man” came up during an Estes session. It appears to have a man melding into the building, with a tour group or ghost hunting team in the background. The walls and windows are reminiscent of Waverly Hills itself.

The phrase for the final image we’ll examine today is “Buderhaus”.


The cryptid alter-ego of Dana Newkirk, the neural network should have no indication of what this made-up word is. We see an old woman in a dress or shawl, surrounded by faces of children and cats. She appears to have snot dripping from her nose. The Buderhaus was made up when Dana was seen wearing a dirty piece of clothing covered in cat boogers.

This is just a small selection of the images that I’ve created so far. Not all have been noteworthy, and my methods could be refined. However, I believe this is worthy of further exploration. My next steps are to develop a consistent method of intention setting, to change the technological setup such that I can run the image generation for an arbitrary period of time without interruption, and to create larger images and longer videos.