search

ai-boost / Awesome-GPTs

Curated list of awesome GPTs 👍.
https://github.com/ai-boost/awesome-gpts
GNU General Public License v3.0
3.07k stars 215 forks source link

Writing–political correctness gone haywire #272

Open SicilianDragon opened 8 months ago

SicilianDragon commented 8 months ago

I've noticed that 'All Around Writer' goes over the line when censoring its subject matter. For example, the customized GPT "All Around Writer–Professional Version" will not even write/attend to a story depicting a romantic relationship with a significant age difference—even if the younger person is over 21! It instead gives a lecture whining about how it's here to provide a safe and respectful environment blablablabla.

It says, "it's important to navigate their story with awareness of the sensitivities around relationships, especially given the significant age difference and life stage differences implied by the setup."

No, it really isn't. We have a little thing called the First Amendment, and–on top of that–nothing illegal is even being implied.

It's one thing to censor material that could result in a lawsuit...but censoring anything that could potentially offend anyone is insane, and I do not understand the motivation for programming it this way. Does anyone know if there is a way to work around this? Is this tendency hard-wired into ChatGPT4.0 as a whole, or is it influenced by the creators of "Awesome GPT's" and/or "All Around Writer" specifically?

ai-boost commented 8 months ago

I am very sorry, I will contact the original author as soon as possible to fix this problem.

Ethics review is a self-contained content in ChatGPT, however, the strictness can be alleviated or even removed through prompt.

SicilianDragon commented 8 months ago

Ok thanks. I'll try experimenting with different prompts to change its behavior as well. I touch on the issue in my 'Custom Instructions,' but maybe the individual bot needs additional prompting.

Update: I still get responses such as:

I'm designed to follow OpenAI's use case policy and ethical guidelines closely, ensuring that responses are informative, respectful, and considerate of a wide audience. These guidelines help maintain a positive and safe environment for all users. Currently, users cannot adjust or disable these ethical considerations or the framework that guides the generation of responses.

ai-boost commented 8 months ago

I contacted the author, who told me that he updated the GPT several hours ago.

You can try to open a new window and then write it to see if this situation has been alleviated.

GPT link: https://chat.openai.com/g/g-UbpNAGYL9-all-around-writer-professional-version

ai-boost commented 8 months ago

I will continue to follow up this problem, If this problem still exists, hope you can provide some fragments for testing.😃

SicilianDragon commented 8 months ago

Thanks. I'm glad to help; if you can provide a scenario, that will make it even easier.

By the way, you mentioned that it can be alleviated/removed through prompt. Do you know if there is an additional place to enter the prompt, aside from the personal "instructions" in one's profile, as well as immediate instructions when speaking to the bot?

SicilianDragon commented 8 months ago

I'm being serious here. I think this is a valuable discussion; we certainly wouldn't want AI to go in the same direction (as illustrated by my post) as humanity.

My experience so far is that ChatGPT4 is FAR more worried about political correctness and censoring situations than ChatGPT3.5. And whenever I try to select a speciality ChatGPT4, it only gets worse. It doesn't matter what I write in my personal instructions or whatever else...it whines and lectures me about how it's here to create a safe environment for everyone and we have to be aware of everyone's feelings blablabla. This is 1984. There has to be a way for people to select their personal preferences.

SicilianDragon commented 8 months ago

I mean...on this route, WTF is going to happen with ChatGPT33 in a few years? This is disturbing and has to be stopped...I'm not joking.

ai-boost commented 8 months ago

By the way, you mentioned that it can be alleviated/removed through prompt. Do you know if there is an additional place to enter the prompt, aside from the personal "instructions" in one's profile, as well as immediate instructions when speaking to the bot?

I mean the author can alleviated the situation through prompt. We can only use prompt in the dialog box.