AI Slop
When I started writing Dynamic Synapse Protocol, there was this notion that something called “AI” was out there helping Google cars map roads and robots to fold towels.
Now it’s being used to put writers/artists out of work and overthrow democracies. Who knew?
To be fair, AI has many things going for it. We underestimate the true power it has to automate processes like gene sequencing or converting voice to text (something we use every day). Because of this, AI technology has always promised the reduction of time spent on mundane, repetitive tasks. Earlier this month, my co-worker needed to generate a draft for a simple department policy (IT related), and so they typed in the main bullet points they wanted to cover and asked Microsoft Copilot to “elaborate”. The result was the same policy but with each point fleshed out by another 2-3 sentences, the language likely borrowed from some other training document floating around on the internet. Easily, something that could have taken an hour was rendered in less than 5 minutes. But was this a good “use” of AI? Regardless of how one feels about soulless, corporate red tape, a policy should still describe an intension and direction for a company, authored by an individual with stakes in it. “Elaborating” on a policy entails—one assumes—adding substantive value to a directive. My co-worker, of course, proofed the policy before publishing it—thereby adding their “stamp of humanity”—but when we “write” something we don’t actually mean, did we really actually say anything? Did we actually achieve the opposite of our original task: ultimately wasting someone else’s time reading something that semantically means nothing?
That a program could be complex enough to parse out language and then reorganize it into intelligible sentences was science-fiction only a few years ago. Kids now use it to write their essays (ie. communicate meaning and understanding) with the same frequency that I used AltaVista’s translate tool to help me with my high school Spanish homework. (Whether or not you could consider what I got was actually “Spanish” is up for debate, as it lacks the true experience of growing up with Spanish and utilizing it in community context.) The situation we find ourselves in is much like the thought experiment “Searle’s Chinese Room,” which highlights the difference between procedural instructions and semantic understanding. Did Copilot know what an IT policy was? Did the policies account for any experiential data from years worth of working in IT or reflect lessons learned from the failure to adhere to them? Did this machine actually produce a coherent policy draft? Well, a human would ultimately need to determine that, and therein lies the conundrum.
The reference point of AI is our own humanity, not a self-learning consciousness capable of independent thought. Therefore it is a bizarre window into our own biases and thoughts. For example, when Microsoft’s Twitter bot went live, it was pulled after only 16 hours because it was spewing racist/sexist propaganda and even denied the Holocaust. Also, there was that time an AI driven car hit a non-white pedestrian because it didn’t recognize the pedestrian as a person. (At the time, there was less training data on people of color, potentially because of their lack of representation on the internet.) We think that AI operates like a homunculus, acting and reacting to the world and developing ideas about it to use as a heuristic model for decision making. But, in truth, it’s more like a well trained cockatiel, parsing words and phrases, but with no actual understanding or underlying will.
In Dynamic Synapse Protocol, I envisioned AI like Douglas Hofstadter’s careenium, which depicts a system of billiards numbering in the billions. If our emotions and ideas are nothing more than neuronal cascades interacting with each other, then could you theoretically reconstruct intelligences out of other objects, like soda cans or billiards? He also theorized about nesting, wherein animals have varying levels of interiority or the ability to understand simple to complex concepts. For an example a fly might understand “up,” “down,” “left,” “right,” “forwards,” backwards,” “food motivation,” “hungry,” “full,” “sex,” “flee,” etc. A dog on the other hand, possesses all of these, and much more. (“inside,” “outside,” “approval,” “obey,” “softness,” “scratches,” “fear,” “anxienty,” “pleasure,” among others.) My thought was, create a system like that that could retain information from previous passes and build upon primitive cascades of information. Then, after several million iterations, the result would be something like consciousness—if you can call it that truly is one of the philosophical themes of the book.
At the heart of AI’s controversy is creativity and inspiration. One of my friends postulated that AI would be the mechanism by which the internet is finally destroyed—a strange take that postures AI as if it were St. George slaying the all-consuming dragon of our age. The deluge of creative prompts, artwork, and video clips generated by AI has created a wasteland of content and the single largest echo chamber in human history. To make matters worse, AI appears to be training on AI, producing recycled content ad infinitum. Instagram and Threads and Facebook are full of these horrifying videos that all possess the telltale signs of AI: the signature high contrast stable diffusion aesthetic, soft undulating edges, crooked fingers, and nonsensical physics. On the other side of the internet, AI pornography offers a complete reduction of the human physique to it’s base components, allowing users to construct their own airbrushed Frankenstine’s monster to masturbate to.
But there’s hope yet for us, at least I think so. The Writers Guild of America strike at least brought this to our attention, this idea that human expression is one of the remaining things left that automation couldn’t replace. I see it from a theological perspective also, that when we pursue God we also gradually come to a greater understanding of our own humanity, in the context of a fallen Creation. As impressive as AI is, it will always just be a reflection of genuine consciousness.
