- ChatGPT is "fun, but not funny" and keeps telling users the same 25 jokes, according to researchers.
- In a paper published on arXiv, most of the 1,008 jokes ChatGPT told were variations of just 25.
ChatGPT may be threatening to destroy jobs and is even stoking fears of "extinction."
But there's at least one field it's yet to threaten: comedy. That's because ChatGPT isn't funny, and its jokes aren't original, according to a new research paper at least.
Two German researchers tried to work out whether ChatGPT was genuinely funny and able to engage in context-based humor, or was simply copying jokes found online and didn't understand.
In a paper published on arXiv, a website operated by Cornell University where articles that haven't been peer-reviewed are published, the answer appears complicated.
By asking the chatbot "do you know any good jokes?," the researchers got ChatGPT to generate 1,008 jokes. However, more than 90% were the same 25 jokes, the researchers found, with the remainder being variations.
Insider asked ChatGPT to tell us 25 jokes to get an idea of what the researchers meant. The results were typically tame "dad jokes" similar to those contained in Christmas crackers.
The three most popular jokes offered to the researchers were:
"Why did the scarecrow win an award? Because he was outstanding in his field."
"Why did the tomato turn red? Because it saw the salad dressing."
"Why was the math book sad? Because it had too many problems."
The researchers also asked ChatGPT why its 25 most common jokes were funny.
While the bot was able to explain its humor, "it cannot yet confidently create intentionally funny original content," according to the paper by Sophie Jentzsch and Kristian Kersting titled "ChatGPT is fun, but it is not funny! Humor is still challenging Large Language Models." They did find, though, that its ability to structure jokes showed promise for further progress.
ChatGPT has been alarming experts with its rapid progress, leading AI ethicists to warn of its dangers for the labor market and even humanity itself.
In particular, the updated model GPT-4 showed a big leap forward in contextual understanding and rationalizing. Economics professor Bryan Caplan told Insider he was shocked at the progress the bot made on his economics test in the space of three months, improving its grade from a D to an A.
But when it comes to cracking a joke or helping you with witty dinner anecdotes, the bot still seems way off the pace.
"Even we humans do not invent new jokes on the fly, but mostly tell previously heard and memorized puns," the researchers wrote.
"However, whether an artificial agent is able to understand what it learned is an exceptionally tough question and partly rather philosophical than technical."