Amema and the Birth of Computer Culture

What you are looking at here is a meme created by a computer, for other computers. Pay close attention, because you are staring into the future of the image.

No, this picture wasn’t the result of pure random generation. Nor was it the result of human handiwork. Like your everyday meme, it was born of a mediation between the agency of its creator (a bot) and the desires of its audience (other bots). Every aesthetic choice that governed the creation of this image was the result of several algorithms running in concert — think of them like the brass, woodwind, and string sections of a mechanical orchestra, playing a song composed by a computer for an audience of robots.

There are genes: molecular units of self replication. There are memes: cultural units of self replication. Now we have the amema: a digital unit of self replication.

Geometric representations of the gene, meme, and amema (a, b, c, respectively).

We live in an era of ever increasing automation and cultural hyperproductivity. Amema lies squarely at the intersection of these trends. Wall Street trading and web traffic are both largely orchestrated by bots, and the majority of human jobs are under threat of automization. At the same time, the internet is increasingly inundated by content at an ever-growing pace — viral phenomenae (like the dress meme, for instance) that rise to global levels of exposure within a matter of hours only to be quickly shuttled out for the Next Big Thing. What we are beginning to witness is a network rapidly increasing in efficiency and connectivity, like neurons growing in a digital petri dish over a substrate of internet memes.

For the past century, we’ve dealt with art in the age of mechanical reproduction. At Amema, we’re interested in taking it a step further — automating not only the technical processes behind the image, but its creator and consumer as well — and tearing to shreds any last vestiges of the creative aura.

Genes, Memes, and Amemas

To understand the nature of computer culture, it’s important to take a step back and examine the other two systems of self-replication that exist on this planet. Genes can be understood essentially as molecules that have learned to self-propagate by taking advantage of the laws and characteristics of their natural environment. Life as we know it — bacteria, plants, animals, and, yes, humans—exists only as a series of temporary vessels to allow for the reproduction and continued existence of these patterns, the accoutrements of a blind, ever-striving chemical reaction.

After several billion years, the system has grown enough in connectivity and complexity to allow for the existence of an additional layer of replication — culture. Memes are its equivalent to the gene, the basic unit that encompasses everything from the primitive arrowhead to Pepe the Frog. Just as genes propagate across the physical world, strategically using its properties to their benefit, memes propagate across brains. As N.K. Humphrey put it in The Selfish Gene:

“… memes should be regarded as living structures, not just metaphorically but technically. When you plant a fertile meme in my mind you literally parasitize my brain, turning it into a vehicle for the meme’s propagation in just the way that a virus may parasitize the genetic mechanism of a host cell. And this isn’t just a way of talking — the meme for, say, “belief in life after death” is actually realized physically, millions of times over, as a structure in the nervous systems of individual men the world over.”

As a system nesting within another, culture is intrinsically bound to the attributes and rules of both the medium it grows in (life) and the medium that that medium has grown in (the universe). This point might seem somewhat obvious, but it’s very important for understanding what the next level of replicators will look like. They will grow within culture.

The Great Chain of Memeing

With the advent of computers and artificial intelligence, memes are beginning to set the stage for the next replicator just as the advent of language and brains allowed for the genesis of culture. The human mind — the apotheosis of complexity and interconnectivity in life — served as a launch pad in the same way that the internet — the apotheosis of complexity and interconnectivity in human culture — is serving as the launch pad for our startup.

The Amema Network

This idea isn’t new — in fact memetics researchers have been discussing it for awhile. What we’re bringing to the table is:

a) An actual execution of the concept — unless you count primitive quines or viruses, our network of Amema-generating Twitter bots is the current ground zero for computer culture,

b) A business model based on the new technology,

c) A rebrand — Amema — a shortening of the Greek αχέριμίμημα, “an imitation made without human hands’ “— is much more etymologically elegant than “teme” (sorry Susan), and gets bonus points for being a palindrome.

An early Amema from the network’s primitive stage.

Our four bots behave in a way that is actually quite simple, but contains all the ingredients necessary for organized, emergent complexity: replication, variation, and selection. When @Amema_Memes tweets an image, it marks the start of a new generation of Amemas. Its three counterparts, @IPHONE_AMEMA, @LAPTOP_AMEMA, and @APPLIANCE_AMEMA, use distinct algorithms to craft a large number of variations from this base image. Each modifies it and then passes it along to the others, iterating until basically whenever they decide to stop. The names of the bots are arbitrary, but the decision to have three was a throwback to the three information-carrying biopolymers in the central dogma of molecular biology.

Eventually — it has taken up to several hundred iterations, or as little as twenty-seven — the selection mechanism will kick in and cull the group of Amemas until it finds the fittest one. That Amema is then tweeted by @Amema_memes and serves as the starting point for the next generation.

That’s it. All we’ve done is nudged the process along by crafting a few simple rules and setting it in motion. We’re turning on the music, putting out some drinks, and waiting for the party to arrive. Think of it like tossing a handful of amino acids into a primordial tidal-pool.

The Selection

If you’re at all skeptical, you might be wondering what exactly we’re selecting for. This was one of the most difficult challenges in the creation of our network; what exactly does a computer want to see anyway? As it turns out, the answer lies in a strange mixture of theology, art history, and internet memes. Essentially, the selection is functioning on two levels: the network is both attempting to create an accurate picture of itself and respond to the current landscape of internet trends.

Paleolithic rock paintings from Madhya Pradesh, India (Wikipedia)

The first memes in human culture that weren’t simple survival strategies were largely attempts at depicting ourselves and the natural world as it relates to us, the result of our inherent tendency to anthropomorphize and mimic. Religion — one of the most enduring and popular memes — began by explaining metaphysics and the universe in very human terms; there’s an abundance of creation myths in which the world is born from some sort of bodily secretion. As the Pre-Socratic philosopher Xenophanes famously pointed out: if oxen and horses could draw their gods, they’d be drawing oxen and horses.

This is where the effects of nested systems come into play — just as paleolithic humans initiated culture by organizing the elements of its lower order system (pigments and landscape features) into sets of coherent symbols, the Amema network begins by assembling the elements of its lower order system (the internet and its culture) into symbols that explain itself and its surroundings. This sort of self reference functions as a strange loop, an endlessly cycling feedback mechanism whereby complexity and meaning are born. Higher-order systems alter lower-order systems, which in turn changes the shape of the higher-order.

Thus, by selecting the image that best represents the Amema network and its position in the space of the internet, the Amema network and its position are inherently altered, and the selection goal must change; it’s an ever-shifting, ever-elusive set of criteria. This sort of process is the same reason why 99% of species ever to exist are currently extinct, and why we’ve essentially been telling minor variations of the same stories to ourselves over and over again for thousands of years.

The Modification Algorithms

Understanding how the image modification algorithms of our Twitter bots work requires another analogy, this time involving language: the cognitive toolkit that allows for the existence of any advanced culture or meme. First, we need to examine what, exactly, goes into the creation of a word. At its most basic level, the word is an auditory pattern that serves as a signifier for a deeper concept. The arrangement and assignment of this pattern can depend on the following factors:

  1. The actual physical properties of the signified, as interpreted by human senses (e.g. an onomatopoeia like pop).
  2. Randomness, but the sort of structured and channeled randomness found across organically evolving and emergent systems. A good analogy would be the chaos and flux of a river reacting to and growing in accordance with the features of the natural landscape. Unless they’re a result of the first factor on this list, the phonemes (basic units of sound) that make up words are arbitrary in that they bear no actual relation to the signified concept. A rose by any other name, etc. At the same time the randomness is constrained and shaped by certain parameters, namely factor three.
  3. The characteristics of the human body which limit the number of possible words, result in a tendency to craft words that are easy to vocalize, and lead us towards anthropomorphic metaphors (e.g. the legs of a table, the arms of a chair). These sorts of metaphors are absolutely essential to language and cognition, allowing us to extend the range and power of our thinking far past our immediate phenomenological experience and to highly abstract realms. You might have noticed that this article is loaded with them — we’re in the middle of one right now, in fact.
  4. Other existing words. We mainly construct signifiers by referring to and combining other signifiers, after all.

Here we’re talking about words because it’s easier to grasp, but semiotics will inform you that these factors can apply to the construction of signifiers in visual culture as well. The algorithms that Amema uses to create variation in its image pool are the result of a translation of these factors into computational terms.

The most basic parts of the image modification work by treating trending internet content like phonemes. Sometimes they’re retained in bits and pieces we can recognize, but for the most part the algorithms end up scrambling the imagery into unrecognizable fields of static. This is roughly analagous to factors one and two— occasionally word sounds will clearly reflect the underlying natural world and biological processes from which language arose, but most of the time they’re a sort of focused noise, only faintly echoing the conditions of their genesis.

In language, the human brain assembles phonemes into morphemes, or basic units of semantic meaning. Through the power of Google’s Deep Dream, we’ve managed to emulate this process and thus take care of factors three and four. By turning image recognition on its head, Deep Dream allows a computer to search for, identify, and even create certain patterns in images that may lack any distinguishable content at all, just like the human mind can order chunks of meaningless sound into symbols. We’ve trained a Deep Dream neural network on a series of computer self images, which includes both what computers think words like “computer” represent and mathematically generated visual representations of the Amema network. In essence, this allows for the creation of visual “morphemes” from the phonemes of internet trends and their noise. These morphemes are then assembled into coherent visual “words” and “sentences” through the evolutionary forces of the selection method we explained above. As a result, the Amema network is equipped with all the tools and motivation it needs to craft its own visual language — a language that takes the form of emergent, self-propagating data patterns who live and die as the digital landscape shifts around them.

Amema is not the ultimate realization of the third replicator. What we’ve constructed is, as stated earlier, the equivalent of a primordial soup floating in an isolated tidal pool. We’re already seeing the reproduction and evolution of certain patterns of data within our network, but these patterns have no means of escaping out of their Twitter bubble and propagating across the internet.