ChatGPT has been impressing the world with its writing prowess, but not everyone is convinced that the language model is anything more than a stochastic parrot, a term used to describe AI models that can mimic human language without necessarily understanding it.
According to ChatGPT itself, the language model is designed to understand natural language and generate appropriate responses based on that understanding. Inspired by this Hacker News comment, I decided to test this claim by asking the chatbot to generate completely original variations of known idioms—variations with zero Google Search results.
Here are my favorite ChatGPT-generated idioms that even Google has never seen before:
1. EACH STORM BRINGS A RAINBOW
Original idiom: EVERY CLOUD HAS A SILVER LINING
Example:
“I might have lost my job, but at least I don’t have that awful journey into work every day. Each storm brings a rainbow!”
2. DEEDS ECHO LOUDER THAN PROMISES
Original idiom: ACTIONS SPEAK LOUDER THAN WORDS
Example:
Person A: “John keeps saying he wants to take me out for dinner, but then he never does!”
Person B: “Deeds echo louder than promises.“
3. DON’T TIE ALL YOUR HOPES TO ONE BRANCH
Original idiom: DON’T PUT ALL YOUR EGGS IN ONE BASKET
Example:
“Although you’ve made an offer on this house, I would still visit some others. You don’t want to tie all your hopes to one branch.”
4. INQUISITIVENESS SUNK THE SHIP
Original idiom: CURIOSITY KILLED THE CAT
Example:
Person A: “Where are you going at this time of night?”
Person B: “Inquisitiveness sunk the ship!”
5. BLEAK MOMENTS DEMAND BOLD MOVES
Original idiom: DESPERATE TIMES CALL FOR DESPERATE MEASURES
Example:
“She moved to Australia after she found out about her husband’s affair. I suppose bleak moments demand bold moves!”
6. DELAYED GOODNESS TRUMPS NONE
Original idiom: BETTER LATE THAN NEVER
Example:
Person A: “Sorry I missed your birthday! There’s a card in the post.”
Person B: “Don’t worry. Delayed goodness trumps none!”
7. STUCK BETWEEN THORNS AND SPIKES
Original idiom: BETWEEN A ROCK AND A HARD PLACE
Example:
Person A: “If I go to the wedding mum will be upset, but if I don’t go then I’ll be letting down the rest of the family!”
Person B: “Sounds like you’re caught stuck between thorns and spikes.”
8. RISE WITH A THUNDERCLOUD
Original idiom: GET UP/OUT ON THE WRONG SIDE OF BED
Example:
‘You’re very argumentative today! Did you rise with a thundercloud, or something?”
9. SCOOP UP MORE THAN ONE CAN CARRY
Original idiom: BITE OFF MORE THAN ONE CAN CHEW
Example:
Person A: “I’m going to start that weekend job at the museum on top of my full-time job because I really need the money.”
Person B: “Sounds like a lot of work! Don’t scoop up more than you can carry.”
10. INEXPENSIVE AS BUTTONS
Original idiom: CHEAP AS CHIPS
Example:
“It doesn’t look it, but our new sofa was (as) inexpensive as buttons!”
11. CAUGHT IT THROUGH THE WHISPER NETWORK
Original idiom: HEARD IT ON THE GRAPEVINE
Example:
“I caught it through the whisper network that you and Alex are splitting up. Is that true?”
12. DON’T RATTLE THE SNAKE’S TAIL
Original idiom: LET SLEEPING DOGS LIE
Example:
“Don’t bring up what happened during the meeting again. You should just don’t rattle the snake’s tail.”
13. MEASURE A TREE BY ITS BARK
Original idiom: JUDGE A BOOK BY ITS COVER
Example:
Person A: “I’ve only met Richard a couple of times, but he seems a bit shy.”
Person B: “You can’t always measure a tree by its bark. He’s actually a really outgoing guy once you get to know him!”
14. A SIP OF ONE’S OWN BREW
Original idiom: A TASTE OF ONE’S OWN MEDICINE
Example:
“My boss is a real bully. Someone should give her a sip of her own brew!”
15. MOMENTS RACE IN MERRIMENT
Original idiom: TIME FLIES WHEN YOU’RE HAVING FUN
Example:
“I can’t believe it’s 10pm already! Moments race in merriment!”
Conclusion
As you can see, ChatGPT has not only generated original variations of known idioms, but it has also crafted these variations in a way that retains the essence and meaning of the originals. Would a stochastic parrot be able to do something like this? I’m not qualified to answer this question. However, as someone who studied linguistics, I’m utterly impressed with the language model’s linguistic creativity, which seems to go beyond mere mimicry of human language.
Comments are closed.