This reminds me how much fund I had with GPT-3 when it was in beta and available for free. I used it to recreate a reddit "bot" that explained jokes. It was a joke because it was obviously a human.

Example

Joke: Three men are on a boat. They have four cigarettes, but nothing to light them with. So they throw a cigarette overboard and the whole boat becomes a cigarette lighter.

Explained: Lighter can mean a device that creates fire or less heavy.

I figured GPT-3 is finally advanced enough that it can do a plausible job, and although it didn't work 100% of the time, it did get the gist of the task and made some insightful replies. This was my favorite:

Joke: My wife and I went on our honeymoon to Australia. Unfortunately, I had to dial the help line. "G'day this is Tim, you've reached the Aussie help line. How can we help ya?" I told him, "We were in the ocean and my wife was stung by a jellyfish on her lady parts. Her vagina is completely swollen shut. It's our honeymoon, and well....ya know." The guy on the help line replies, "Ah, bummer mate!". I say, "I hadn't thought of that! Thanks for the advice. You've saved my honeymoon!"

Explained: Australia + honeymoon + penis + swollen vagina = Aussie joke.

I wrote more examples here:

https://mleverything.substack.com/p/using-gpt-3-to-explain-j...