AI writing?

Everyone reading this is familiar with predictive text when typing a text message, and most would agree that the relevance of the predicted text seems to be more and more relevant over time. The basis for this capability is the use of large language models, which applies machine learning techniques to enormous databases of the trillions of words spewed by humans online. Increasingly, applications of these techniques are producing longer and longer, more and more coherent text. New York Times columnist Farhad Manjoo took on this topic four years ago in How Do  You Know a Human Wrote this? Techniques have evolved over the last for years. In 2020, a large language model known as GPT-3 became publicly available: 

Generative Pre-trained Transformer 3 (GPT-3) is an autoregressivelanguage model that uses deep learning to produce human-like text. It is the third-generation language prediction model in the GPT-n series (and the successor to GPT-2) created by OpenAI, a San Francisco-based artificial intelligence research laboratory.[2] GPT-3's full version has a capacity of 175 billion machine learning parameters. GPT-3, which was introduced in May 2020, and was in beta testing as of July 2020,[3] is part of a trend in natural language processing (NLP) systems of pre-trained language representations.

As often happens these days, this new technological capability has collided with power of commercial online attention-farming. A certain variety of blog known as a "content farm" hires writers to generate large quantities of articles that are generally not of high intellectual merit; the purpose of the flow of not very good texts about a variety of topics is to create a medium for serving ads. Crafting these crappy but numerous articles generates income through search engine optimization techniques that nudge the articles and accompanying ads into the results returned for searches made by individuals. Now an app offers the capability of using GPT-3 to quickly generate and optimize blog post outlines for content farmers. Drafts created by hired writers can be analyzed and optimized. Seven bucks for three brainstormed outlines.

I decided to try it. When I started writing professionally, I was staff writer for the Institute of Noetic Sciences, founded by Apollo astronaut Edgar mitchell, who had a transcendent experience on a spacewalk. He told me that every astronaut who had space-walked experienced something profound. At an airshow in London, the press thought it would be a good idea to photograph Mitchell in a cockpit with a Soviet cosmonaut. Mitchell told me that as soon as they were in the cockpit with the microphone turned off, the cosmonaut asked Mitchell: "Did you see God out there, too?"

So I entered "transcendent experiences in space" into the blog idea generator, which returned the following outline:

Outline

H2 - Overview: Space, Religion, And Transcendent Experiences

H2 - Person 1: Astronaut Terry Virts

H2 - Person 2: Cosmonaut Alexey Leonov

H2 - Person 3: Astronaut Richard Mastracchio

H2 - Person 4: Cosmonaut Sergei Volkov

H2 - Person 5: Astronaut Charlie Precourt

H2 - Person 6: Cosmonaut Alexander Kaleri

H2 - What Do Astronauts And Cosmonauts Have In Common?

H2 - 1. Astronaut Michael Lopez

Pretty cool! I didn't know about any of those spacefarers, but a few minutes searching could give me material to fill in the text under the headings.

What if the text corpus could include all MY writing?  I'm not thinking of GPT-3 capabilities so much as robots that could replace writers, but something that could do for idea generation what word processing does for writing.