We often use words purely as a means to an end without remembering that they are often ends in themselves. “Language,” William Burroughs once said, “is a virus.” And the terminology and words we use to define and describe generative AI reveal our expectations and assumptions about it. One popular example of that is the term “stochastic parrot.” It treats AI purely as a probability-based imitator, copying and pasting from its training data without adding any intrinsic understanding or innovation. And while it is technically accurate, it’s also a description that carries an assumption of fundamental ephemerality, a suggestion that the output of AI is always a transient trick, and never provides genuine insight or inspiration.
I’d propose an alternative description that I think may more suitably encompass AI’s potential: “Mimetic Chameleon”. We are, after all, simulating human creativity. We are executing the creative act in a fraction of the time it would take a human to perform the same task. The simulation here is a controlled re-creation of a process or system.
Generative AI is not merely copying or parroting. The model is simulating a human approach to a task based on an estimation of all the times we’ve done it before. Whether that’s writing a poem, painting a picture, or analyzing complex data, the AI is simulating the creative or analytical tendency and then crafting it’s output in the context of that simulation.
Consider a photograph. What we see isn’t an exact replication of a moment but an interpretation, a view through a lens, a capture of light and shadow, a chemical reaction. And yet, we don’t discount the authenticity of a photograph. Instead, we value it for the new perspectives it offers. Similarly, every AI-creation is packed with artifacts from the process of its creation. Our ability to identify these artifacts not only lets us to separate AI-created content from human-made, it adds another layer to our understanding of its utility.
We often use the term ‘artificial’ in other contexts, typically assuming the result won’t perfectly replicate the original. Instead, a phrase like “artificial flavor” evokes only a semblance of the genuine article. It should enough to satisfy some agreed upon understanding of what “banana” or “green apple” might taste like even as we talk describe how no banana has ever really tasted like a banana jelly bean.
And yet, when tasting wine, we do the opposite. We talk about “notes” of flavor that are often nothing more than the artifacts of our individual sense of taste and perception because they mostly aren’t inherent flavors in the wine itself. We may find ourself agreeing that we taste “vanilla” or “leather”even though we might not have ever come up with the same flavors ourselves.
As the latest innovations in simulation have burst onto the scene, the reaction of media seems to have been to lean into the “artificial” and not the exploratory. These new words and images should not move us. It is suspect. It is “fake”. Not a flavor we are tasting but a chemical experience forced upon us.
But as AI evolves, our definitions and understanding of authenticity are bound to change. As AI-generated content becomes more commonplace, it’s likely we’ll adjust our perceptions of what constitutes ‘real’ or ‘genuine’ art, writing, and other creative work. We still see the artifacts in AI-generated content, the telltale signs of their non-human origin, and embrace them as a part of this new form of artistic expression.
At the core of human experiences are the choices that we make. As humans we strive to create what we consider to be pleasing and beautiful — or at least impactful. We use our art to adorn and enhance our environments — real and virtual. Through our craft we give a touch of connection to our experience, whether that is occurring inside of a machine or in our actual environment.
To think that the point of AI is to fully replace reality right out of the gate is, I believe, to miss the opportunities these tools are bringing us. It is to perceive us as utterly vulnerable to AI as a tool for deception. Worse yet it considers that the fundamental”success” of AI lies in its ability to replace the real.
A useful tool offers us an array of choices and perspectives that we might never have accessed otherwise. At its core, the tools we use are not about forging a perfect simulation of human experience but about providing us with the means to explore, experiment, and push the boundaries of what we can create.
In the end, the successful integration of AI into our creative workflow doesn’t hinge on its ability to replace the ‘real’. It lies in its capacity to assist us in our endeavors, to enrich our experiences, and to propel us into a future where our ability to create and understand is augmented, not diminished, by artificial intelligence.
No matter how innovative or apocalyptic the implications of the technology, the tools that we will ultimately create with AI won’t be all that different from the tools that have come before. They will not make choices for us, but give us more to choose from. And while the things we choose to create and the ways we choose to use them may sometimes be terrifying, that’s exactly what we build our technology for and what we want it to do.