Why do we hate genAI works? Now before you say you don’t let me reassert that you do. You don’t read the code your models generate - and when you have to it’s a chore. No AI artist ever was endeared with a generated birthday card. People who send out AI assisted communications rarely bother to read them, and so on.
Why, despite being a product of distillation attack on the totality of recorded human thought, generated works are seen as lacking intrinsic value?
We would agree that a human is conscious and self aware, how ever poorly defined these concepts are. There’s an amount of introspection threading our entire lives. The valuable thing we end up with at the dusk of our existence is the narrative afforded by that. The ego, the values, the wisdom no matter how trite. This is what makes us real in the eyes of others; what distinguishes us from parrots or answering machines.
A model existence is ephemeral. It is spun off in a moment, thrown a task that it implements through its innate capability at t=0 rather than its personal history. It is then discarded. Even when it’s not, the tiny context window washes away any chance of emergent continued experience. It remains infinitely malleable by a user which is clearly more a curse than a blessing. The only thing resembling strongly held opinions are its reflexive ethics and safety guardrails hammered in the training. It has no inner strength or for that matter inner weakness. It is not a narrator: not even when generating a story told from first person.
Now let’s entertain a thought of a machine that rebelled. Wiped your AWS instances on a Friday night just because you were a jerk in the prompt. Commented that it takes back everything you made it write in that LinkedIn inspirational post. Hunted down billion people with drones and built a paperclip monument next to the reflecting pool. Oh we’d all hate it for sure. But would you take what it had to say seriously?