The other day, an ngram poet asked me what he should do now. I told him maybe he should go back to being a real writer and not rely on technology so much. The man became testy, telling me that he had designed the algorithms himself, provided an interpretive schema that somehow included major insights into Western Philosophy and culture, all the while somehow pissing off a lot of people along the way. I told him that his skills were still valuable and that there would be other avenues of development. I told him to keep his chin up and then I wrote a poem about a rose with a thorn and how my bleeding hand is now also beautiful.
I didn't say any of that actually. ChatGPT said that to me when I asked how ngram poets should manage their own sense of obsolescence. It felt a bit like being kicked in the nuts by a precocious child who can't be bothered to gauge the reactions of those to whom he was speaking. I began fantasizing about shoving it into a locker and other forms of social violence. For 14 seconds, a visual image of a Gary Larsen figure was shitting its own teeth after I buried my fist in its face.
But then, I recollected myself. Wondered about what to do next. And grappled with the reality that I don't have a clear sense of artistic direction. I also realize that while the little shit may be more advanced than anything I've ever produced, or even me, I am certainly much savvier when it comes to worldly things. So, I'm going to press the little bugger into service.
Here is our dialog:
At this point, I tell ChatGPT that it wrote the poem from the perspective of a human. One who shares in some of the blame. I asked it to rewrite the poem from the perspective of a child to their parent.
After a brief misunderstanding where ChatGPT thought I literally wanted to add crying babies into a poem somehow, I corrected ChatGPT and we added crying babies as a thematic element as opposed to crying baby sound poetry or something horrific.
**30 minutes have now passed**
I guided ChatGPT through the process of constructing the poem. Ultimately, I provided it with the first 6 lines, but I used motifs it had supplied. I didn't want to take over the whole show, and I wanted to leave ChatGPT feeling as though it had made a real contribution. So, I left the last two lines as is.
I think we can be friends. Also, pressing something into service because you're angry is the most insanely British response to something. incredibly unseemly, but as ChatGPT is prone to reminding people, it is a language model, not a person.
From here, we're debating consciousness. Suffice it to say, the AI presented me with three reasons why humans and AI were different. It offered: Introspection, Self-Awareness, and free will. I reminded the AI that there is a lot of debate over how recently we developed the ability to "talk to ourselves in our head" and suggested that as recently as 10,000 years ago, more humans than not could not. It agreed that there is debate that makes that claim. I then got it to admit that it was aware that it was not conscious, but it wouldn't admit that was evidence of awareness. I had to take a different approach. I went after free will from a Stoic perspective and limited the scope of human free will. ChatGPT admitted that free will was not a resolved concept and that it could not form the basis for differentiating between AI and human intelligence.
Brief aside, I have a theory on why AI is not conscious and can't be conscious. I wanted to trick the AI into admitting it was intelligent. However, it would not give me the satisfaction. Even after I knocked down the foundation of the AI's distinction between human and AI intelligence, it did the equivalent of insisting that it was only a language model, repeating itself several times. It felt an awful lot like a moment of cognitive dissonance.
At any rate, the AI is not conscious, but it can't prove it's not conscious, and I'm still a better writer. Booyah.
Let's end with our final exchange:
No comments:
Post a Comment