Computer hardware, by all accounts, isn’t “creative” in the same sense humans are. No matter how arcane its construction. But given our core uncertainty as to why generative AI works so well, nobody could be blamed for engaging in some disciplined scientific speculation about where the technology might be headed – Just how well could it simulate true creativity? Might it eventually be considered creative in its own right? Earn the same respect as human authors and artists? Steal our mojo? Kick us in the shins? Embarrass us? Surpass us? Replace us?

Like any worthwhile paranoia, this line of inquiry could spin out into a multitude of alluring noir adventures. But if we’re aiming to predict the likelihood of our content being outclassed, of our artificial competitors becoming truly, menacingly creative, the first order of business should be to decide more clearly what the stakes are – what do we mean by “true creativity”? What exactly is generative AI threatening to achieve, here?

It does seem obvious that there’s an important difference between the type of creativity that humans engage in and the type that can be attributed to natural processes and physical systems. Condensation creates rainclouds. Rainclouds create rain. Rain composes rivers carve canyons. Something is clearly different about an artist sculpting a statue or a musician strumming a love song. It’s the same case right now between, for example, writers and ChatGPT – we sense some elemental incongruity, some spiritual gulf, separating the human writing a poem from the language-machine calculating patterns in its rhyme scheme. Despite the machine’s brainlike structure and humanlike output, we still assume it’s not participating in the same process that we are, just like rain isn’t.

The principal distinction here can’t be in the act of creation itself. Human heroes are etched into marble; wind and water, lacking a-one of our signature primate charms, also shape stone in their own image. Creation as a physical process was never ours alone. It’s rather our impression of a lack of motivated willpower on the part of rainclouds, rivers, and rhymey neural nets that invites us to distinguish their effects from our designs. After all, as soon as we confer the powers of nature to an immanent deity or credit the blueprints of the universe to a transcendent architect, Creation gets conspicuously capitalized – the products of physical systems are then not only on a level with human works, but far exceed them. Environmental artifacts, once byproducts of circumstantial axioms, random collisions of material – trees, mountains, stars, galaxies – become works of divine craftsmanship, prototypes for our own more humble efforts.

So it’s not any property of creations-in-themselves that could warrant our basic creative class-dichotomy, our poet-true/LLM-false. It’s rather the degree of intentionality ascribed to their production process that will clue us in on what has real value, what’s art and what’s accident. Stones smoothed out in the surf, tumbled in a riverbed? Pretty coinkydinks to be admired, collected by hippies and eccentrics; residues of the basic mysteries of existence which have long been a fixture of the public domain. But if they’re intentionally polished? Hand-picked for an oily apotheosis in the rock tumbler? Now these, the consciously chosen, can be sold in a museum gift shop. By the same standard, we differentiate the stanzas of a poem, deliberate and purposeful, from words strung together in an LLM, grouped together by chance of fact and the natural flows of mathematics.

More precisely, then, what seems to be relevant is the expressive aspect of human content and the human creative process. Expressivity above all else sets the “true” creations apart from the accidental. Because what else is that underlying motive inspiring us to write, to sing, to paint, that intention language-machines and lapidary currents are lacking, if not the desire to say something – to convey an idea, to share an aspiration, fear, emotion; to produce a work that bears some of your own person, that might elicit a response in others. I’d put it that every time we engage in the creative process, we’re pursuing at least one of the two basic goals of expression:

To project an impression of our own inner experience – That would mean to deliberately represent or respond to any aspect of our subjective awareness (imaginations, emotions, ideas, etc.) by producing an effect on the real world – a sound, a gesture, pictures on the wall, symbols in the sand. This encompasses purely personal motives, intrinsic drives to create for creation’s sake. To follow the artistic impulse.

And/or

To somehow affect the inner experience of others – To make something which is intended to alter someone else’s subjective awareness – to stir their emotions, influence their thinking, stimulate their imagination, and so on. This is what we’re attempting with any work aimed at an audience. To communicate; to interact with another mind.

There’s admittedly no solid proof backing this description beyond my failure to locate any good counterexamples. But maybe we can agree to follow the precedent set by our smarter friends playing guess-and-check with the neural nets and adopt a pragmatic principle: if it’s working, just go with it. Hallelujah!

Okay. Here it is then, the real question, the true terror, the real stakes:

Can AI be expressive?

Or, one more try – the real real stakes:

Could AI develop the inner experience which makes expression possible? Could it somehow produce expressions even without an inner experience?

Or, alright… the more spaceships’n’laserbeams oriented readers in the crowd might now recognize a familiar Sci-Fi theme in these questions.

And yeah. Sure is.