On September 16th 2019 I published an article that had been written by artificial intelligence — all I did was provide a 40 word prompt. As I said in the article, I created two versions:

I was reminded of this because there’s been a bit of a buzz around the fact that The Guardian recently used AI (GPT-3) to generate an article in a similar way. I still think my income is safe for a while. It’s repetitive, lacks any flair and is unauthorative. What do I mean by that? There is nothing in or about the article that makes me feel in safe hands. It cites Gandhi, and my first thought was “I wonder if that quote has been made up?” As it happens, it wasn’t. I searched for the quote, and discovered it immediately from a website called Brainy Quotes — which is probably what the AI did.

I think automated writing is probably fine for generating readable reports from tables of statistics. I think it will not be long before it can generate articles that are more rich than the GPT-3 one cited. But — and this is how I console myself — I think the day when AI-generated articles are indistinguishable from those of real writers (by which I mean those people who love their raft and take it seriously) is a long way off.

See also:

Automatic Writing (Updated)