Wednesday, July 8, 2015

Hallucination Engines

At this point I am fairly certain that most people who read my blog have seen the Google Deep Dream images. Those trippy reinterpretations of reality that seem to be covered with eyeballs no matter what else is going on. If you are not familiar with this, do an internet search for the words"deep dream google", then come back and finish this article. I feel this opens many doors for the attentive and proactive Mutant Mage. There are several aspects that I personally find intriguing, recursive feedback loops, mechanical neural networks, and projective imagery. The crossover between technology (the study of craft) and magic (the craft of creating) is not only natural but wholly interdependent. This has given rise to what I have come to think of as Modern Folk Magic. I will outline exactly what I think distinguishes Modern Folk Magic from other paths in another essay. For now I want to discuss what technology can teach us about magic.


Technomancy, quite simply is the use of technology for magical divination. I have come to use the term to describe a wide variety of magic that uses technology, such as, audio/visual cut up divinations, or recordings of mantra recitations, to magic that imitates technology like Gordon White's Robo Fish. Many Mutant Mages including Phil Hine, use Byron Gysin's cut up method, which was popularized by William S. Burroughs. All of these methods share a commonality in that they work under the premise of universe as information, by altering the information you alter the reality represented by that information. This is just a modern revision of magical theory stated in accordance with modern parlance.

My approach is simply that Folk Magic springs from the environment surrounding it, and is informed by the dominant paradigm. We are surrounded by technology and it should be part of our magic as well. Hoo doo root doctors used horseshoes, keys, locks, etc. because that was what they had. I am quite sure that the first person to use a hurricane lamp for conjure was frowned upon by the community at large, then again maybe not so much. The quest for "purity" and the resultant de facto Orthodoxy that springs from it is a modern affliction for our whole culture, not magical circles alone. It is what feeds most Ism's and I suspect it is born out of fear and special snowflake syndrome. As science expands our knowledge of the world around us, we the mutant mages should learn to use it to our advantage. This leads us back to Deep Dream.


The process Deep dream uses to produce images is basically a large recursive loop. The machine is fed millions of images of an object in order to "learn " what the object "is". From Popular Science article, These Are What the Google Artificial Intelligence’s Dreams Look Like, by Dave Gregshorn:
Google’s artificial neural networks (ANNs) are stacked layers of artificial neurons (run on computers) used to process Google Images. To understand how computers dream, we first need to understand how they learn. In basic terms, Google's programmers teach an ANN what a fork is by showing it millions of pictures of forks, and designating that each one is what a fork looks like. Each of network's 10-30 layers extracts progressively more complex information from the picture, from edges to shapes to finally the idea of a fork. Eventually, the neural network understands a fork has a handle and two to four tines, and if there are any errors, the team corrects what the computer is misreading and tries again.
Interestingly enough the AI did not find a "perfect Platonic form of the object" even after being fed millions of images of the same object, an example of this:

This showed that even when shown millions of photos, the computer couldn’t come up with a perfect Platonic form of an object. For instance, when asked to create a dumbbell, the computer depicted long, stringy arm-things stretching from the dumbbell shapes. Arms were often found in pictures of dumbbells, so the computer thought that sometimes dumbbells had arms.
Google's artificial neural network's take on a what dumbbells looks like.
This should immediately point to new ways to think about the assumptions we make when interpreting data. Even when we have  millions of data points we can get it wrong, perhaps the spirits we work with are like the technicians trying to correct the machines errors, and we are the machine. How do we break free of this? Perhaps we should imitate Deep Dream and replicate the image over and over again until its absurdity becomes clear. Perhaps the reason your prosperity spell isn't working is because you don't have a fundamental understanding of finances. If we are putting arms on dumb bells we have not fully understood what a dumb bell is.


Much like Deep Dream is trying to interpret layers of data with millions of images, and failing to find a quintessential essence to the objects, so too the human brain and especially the collective dream. Perhaps we need to see that dumb bells have arms, to develop a more fluid view of Universe. The collective dream is nothing more than an attempt by monolithic cultural structures to codify experience into a manageable commodity. Universe is the ultimate open source platform, our brains are lazy and project learned patterns onto the canvas of life. We fall into patterns, perform rituals, follow scripts, practice daily; none of which is bad as long as we don't let it cloud our vision. 

Rote and repetition are great tools for memorization, but a bad way to conduct a relationship with the divine. Daily practice is important, but it must be performed with sincerity, simple repetition will not suffice. In a world ruled by clock and dollar, it is easy to slip into a routine that can turn into a rut. There is hope, if Google can teach a machine to dream, imagine what you could do with your brain. There are programs that have been running for many millennia that need to be dismantled and reassembled. Folk Catholicism is far more vibrant than the Church in Rome, and probably better for the soul. Most of Europe runs on Democracy 2.0, the United States is still running the beta version, we need an upgrade. Unlike social revolutions of the past, which were collective and dismissive of the past, we must build upon the past but in a manner that advances the individual as much as the mass if not more.

Upgrade your brain and your magic. Quantum theory has much to support the aspiring Mutant Mage. AI can teach us much about how our brains process information about the environment around us, and thus lead us to greater understanding. Magic seems to work best as a hedged bet, if your working is able to bring about two separate but equally beneficial outcomes, you will be surprised with a third alternative that you never conceived. It will be better than what you thought you wanted. Rote and repetition work to our advantage when we want to create new patterns. When we create two or more new patterns in reality, the overlaps create new possibilities. We can anticipate some of the overlaps, but it is the unintended consequences that will always teach us the greatest lessons.


Why did Google develop an AI to produce trippy images? Truth is it didn't, Deep Dream is a side effect. The project that the engineers were working on was called "Deep Stereo". It is an imaging recognition system that would be able to take two to five images of a space and based on its library of images make a 3D rendering filling in empty spaces. In other words, they were creating a machine that could take incomplete data and transform it into an accurate 3D model based simply on what the machine has been taught should fill the gaps. Sound familiar? One day someone got the idea to ask the machine to make renderings based on what it had learned. Thus they were presented with the following:
The Google team realized that the same process used to discern images could be used to generate images as well. The logic holds: if you know what a fork looks like, you can ostensibly draw a fork.
 This is what Google's neural network thinks animals and objects look like.
Unintended consequence that lead to new discoveries. Look at all the fun people are having with this accident. Always record ALL results from your workings, further review may reveal something you had missed. The great thing about these images for me on a personal level is that they illustrate better than anything else my impression of the divine. Textures, lighting, vague impressions combined with distinct details, yes, this is an apt metaphor for life itself. Looking at the above images makes me realize that I have not noticed things that a machine considered intrinsic. The hartebeest for instance, is the grassland and the community of the herd. Look closely, every one of the renderings includes the surrounding environment, one does not exist without the other. Think about that, a hallucination engine is able to see the fluid, dynamic, non dualistic nature of Universe. 

Deep Dream teaches that unintended consequences lead to new discoveries. It shows the world through the eyes of an intelligence that views objects in a relational rather than isolated way. This machine, this hallucination engine, by following rigid programs after being fed millions of data points has come to have a fluid view of its world. There is deep wisdom there, and hope. Revisit your failed workings and find the gold buried there, nothing is as static as it seems. Perhaps, instead of showing the world what we believe it wants to see we can learn to show the world our Deep Dreams.

This is a great lecture from 1976, that has many techniques and ideas that still need further research.

No comments:

Post a Comment