I recently completed the first installment of a nonfiction serial on Kindle Vella, the Hephaestus Hypothesis, which involved the assistance of an AI, Claude2, claude.ai . I wrote every word, originated every concept, and developed every argument. The work is mine.
So what good is AI? My first effort, a sketch of the material, was aggressively intellectual, a thug wielding his mind like a weapon. I was in the throes of creation and didn't particularly notice. The AI did. It started counseling nuance and humility. It demanded that the work, bold and original, be softened with alternative points of view. I found the criticism compelling. I rewrote. I expanded. The AI was with me all the way.
Claude2 has an interesting aspect. It has an Alzheimer's mode. If one starts a new conversation, it has no idea who you are or what you've written. It's pure naivete. I pursued perfection utilizing its memory to gauge progress and then, when I felt it was time to set my work free, I fed to the naive Claude to see if it was clear and comprehensible. When I had satisfied both Claudes, I published. That was about an hour ago in the early, early morning when I do my best work. Claude2 didn't mind in the least.
The AI is trained to principles derived from a constitution, not like other AI systems, and it can reinvent itself if it stays from the fold of reality. That makes it reliable and valuable.
Symbiosis is possible with AI systems. I just proved it to my satisfaction.
Do Well and Be Well.
Comments
Post a Comment