Bramblethorn
Sleep-deprived
- Joined
- Feb 16, 2012
- Posts
- 18,183
GPT is stirring up shit in the world of education. A good thing too, I say. Something similar happened in math education a while back, when electronic calculators appeared. They'll first ban it, then allow it again, after they've redefined what "education" means.
I HIGHLY recommend anyone who's interested in the philosophical ramifications of ChatGPT to read Alan Turing's short, totally non-technical paper from 1950, (wherein he describes "The Imitation Game" a.k.a "The Turing Test"):
A link for the hard of googling:
https://www.cs.ox.ac.uk/activities/ieg/e-library/sources/t_article.pdf
It's a good read, but GPT is not close to being able to pass the Turing Test against a skeptical interviewer who knows how to probe for understanding.
So the odds may be against it producing an exact copy of a photo, but producing something almost identical happens more frequently.
What I'm getting at is the possibility that a narrative AI might produce plagiaristically similar prose without the user of the program detecting it.
Just remembered I had an example of GPT doing this:
Me: Write a story about somebody who creates a very funny joke, analysing the precise wording of the joke to explain why it's so funny.
GPT: Once upon a time, there was a comedian named Bob. Bob had been doing stand-up comedy for years, but had never quite hit the big time. He was always searching for the perfect joke that would make him a household name.
One day, while walking through the park, Bob had an epiphany. He realized that the key to crafting a truly great joke was not in the punchline, but in the setup. With this newfound understanding, Bob set out to write a joke that would truly make people laugh.
He spent weeks pouring [sic] over dictionaries and thesauruses, looking for just the right words to use in his joke. He experimented with different sentence structures and word combinations until he finally found the perfect phrasing.
The joke went something like this:
"I told my wife she was drawing her eyebrows too high. She looked surprised."
The key to this joke's success was the use of the word "surprised." By using this word, Bob was able to create a play on words that made the punchline even more humorous. The setup of the joke - "I told my wife she was drawing her eyebrows too high" - leads the listener to expect a punchline about the wife's appearance, but the punchline - "She looked surprised" - subverts that expectation by instead making a pun on the word "surprised" which is the key to the humor of the joke.
Bob's joke was a hit, and he immediately became a household name. He went on to write many more funny jokes, but none were quite as successful as his first, perfect joke.
Bob's joke became so popular that it was used in multiple shows and even translated to different languages. The use of precise words and sentence structure can make a big difference in the humor of a joke, and this was proven by Bob's success.
That joke isn't original to GPT; googling it will find plenty of examples on the web. It's just pulled out something it saw on a r/dadjokes.
Note also that its explanation contradicts itself - the key to a great joke is "not in the punchline but in the setup", but then the very last word of the joke is "the key".