On Human Authorship: What it Means to Write Something Real

Some Thoughts on Human Authorship

ChatGPT suffers from eternal writer’s block. Functional language is not inspired language. There’s something there, that it can’t access.

How do I know this?

I guess I began thinking about it within the context of my own creative work. I love a good Wham-Bam-Done, ChatGPT-drafted email or letter to a client. Hit all the key points, share relevant info, how easy was that! But when I ask it to help me capture something subtle. When I ask it to deliver some fulsome beauty or truth in prose, I rejoice, because it can’t.  

It’s an inspiration problem that dooms AI to permanent writer’s block. Compare it to the human writer’s faculty. We also regurgitate. We also gravitate to probabilistic construction of our prose. But we also read over the stuff we just wrote and see it as bad, crap, and not right. This is something AI (with no successive training) literally can’t do. What’s it missing?

‍

When I write a sentence, paragraph, or word, I either strike truth or I don’t. The chance to say something true is what drives me to write. Inspiration, for me, is the sense that truth really is there, and that I have a fighting chance to get my words around it.

Now, let’s assume ChatGPT were to hire me to defend its viability as a creative writer. I might argue that the abundance of data on which the AI is trained dwarfs the accumulated knowledge of any human, this includes its knowledge of the most prevalent formulations of prose. Therefore, the written material it generates, if not inspired in the classic sense, may at least prove supremely accurate, more so than any human output.

‍

I’m sure you’ve come across many highly accurate AI-generated formulations of the written word. If that’s all there is to the craft of writing, then sure, count me out of a job. But if there’s a little, or maybe a lot more to be desired from what we write and read, then I say, keep fighting.