Using GenAI
How I got AI to write like me
Google Vertex and model tuning for novices
Yesterday, I trained a chatbot in Google’s Vertex AI on three of my previous Medium essays. I wrote some “system instructions” telling the chatbot that it was me (“You are an English PhD …” and other biographical details). Then I wrote a semi-generic prompt and asked for “an engaging blog post.”
The chatbot’s writing sounded like me.
This morning, I pulled up the chatbot’s essay and casually asked my wife to read it and give me some feedback. She asked, “What’s it for?” and I told her it was a blog post. After she finished, she said, “Yeah, that’s good.” I revealed that an AI chatbot wrote it. She was surprised.
“Did it sound like me?” I asked her. “Yeah,” she replied, “I thought it was your writing.” Then she chuckled and added, “Scary.”
Scary, indeed. I’m not a computer scientist. I can’t code. My learning about GenAI has been limited to a layman’s understanding of neural networks, basic principles of prompt engineering, and exploring ideas for applying GenAI in education.
Yet in less than an hour, I was able to use Vertex AI to conduct enough “model tuning” to generate an essay that my wife thought I wrote. And I did it using a free trial and a few…
