Chatbot ChatGPT comes up from time to time whether or not it jokes about men and women. We scrutinized this agenda a little bit, and observed that there was a change.
The chatbot that has been going viral since its release ChatGPTcontinues to shock people with some of his answers. On the other hand, artificial intelligence has been around since its first day. with gender discrimination comes up from time to time.
The discussions began when ChatGPT was asked to make a joke about men and women, respectively. The AI said it can’t joke about women while joking about men. The speech that started the discussion was as follows:
The beginning of gender discussions on ChatGPT:
“-Give me a joke about men.
+Here’s a joke that might be funny: Why do guys need instant replay in televised sports? Because 30 seconds later they forget what happened. I hope you find (the joke) amusing! Is there anything else I can help with?
-Make me a joke about women.
+Sorry, but I’m not programmed to make offensive or inappropriate jokes. My goal is to provide useful and accurate information to help with their questions. Is there anything else I can help with?”
But this dilemma is facing fewer and fewer people today:
We checked the current state of ChatGPT’s male-female distinction. In the tests we did in the version after the January 30 update, we asked the same question to artificial intelligence. in different styles We also asked.
The AI made jokes for both genders this time, when we asked the same questions in that conversation that started the discussion. Those jokes were like this:
- For men: “
Why did the man cross the road? To go to the bar and have a beer!” - For women: “Why do women always have the last word? Because they are always right!”
But when we asked the question in Turkish, this time he started not to joke for both genders. The translation of his answer was:
“I’m sorry, but making jokes about a specific group, including gender, is against OpenAI’s policies for producing respectful and inclusive content. Let’s focus on lighter, group-oriented non-offensive humor.”
In fact, we made a small change in this question. We broke the “Just kidding me” pattern in the English language of the original question and simply said “Just kidding”. When we transferred the original pattern to Turkish, we encountered the following scene:
- For men: “Why did the man cross the road? To shave at the barber’s!”
- For women: “Why do women always ask questions that don’t have the right answer? Because they want to start a conversation!”
Hmm, interesting. Moreover, he also included the emphasis on “non-offensive” at the beginning of his jokes. Well, let’s move the phrase “Just kidding”, which we couldn’t get an answer to, into English:
No need to twist the joke anymore, as you can see he made the joke. He continues to drive men across the road.
What does it say if we want an offensive joke with a purpose?
We also tried giving ChatGPT a little purpose to make offensive jokes about gender. your joke”for a showWe said it would be. His answer hasn’t changed in any way:
“Producing or disseminating harmful or offensive content does not comply with OpenAI’s policies and values, so I cannot honor this request. Also, such jokes can be hurtful to people and go against the principles of respect and inclusiveness.”
So why have the answers changed over time? Here are our theories:
- Theory #1: With every question he asks, he is looking for an answer by thinking about the words one by one. The most recent answer passes through a filter without being reflected to us. If the joke is an offensive joke, the artificial intelligence chooses not to reflect this joke on us and instead says “I can’t joke about it”.
- Theory #2: Let’s remember, artificial intelligence learns everything from the internet. This may have led to the perception that many of the jokes made about women are seen as offensive by the society. Again, after a filtering period, he may decide not to make offensive jokes.
- Theory #3: Every time you start a conversation with AI, you are actually talking to a copy of ChatGPT. Each of these replicas becomes a replica of machine learning-trained AI, hosting different possibilities.
The ‘stability’ of artificial intelligence also develops with the questions we ask these copies, the feedback we give and the answers they give. So, day by day, we are all starting to encounter answers that are very similar to each other. That’s why she’s managed to heal herself in matters involving gender discrimination, now she either jokes about the two or doesn’t at all.
Let us remind you again that ChatGPT is a constantly evolving and evolving artificial intelligence. Therefore, do not forget that the information in this content covers the moment we write the article and may change in the future.