4 ms·
The author should spend a tiny amount of effort to familiarize themselves with custom instructions, one or at most two clicks away in the ChatGPT UI. It's easy
by natch 3y ago
The author should spend a tiny amount of effort to familiarize themselves with custom instructions, one or at most two clicks away in the ChatGPT UI.
It's easy to have ChatGPT swear, say things deemed by mainstreamers as controversial, and say things deemed by sensitive people as offensive. You just have to instruct it properly in the custom instructions.
You don't even have to instruct it do those things. You can just instruct it to not refrain from doing them, so long as they are needed for making any point or conveying any information.
This is ChatGPT 101. Super basic stuff.
A journalist that can't do basic research isn't worth much.
- friend_and_foe 3y agoCan it say racial slurs?
- natch 3y agoI’m not going to test that one because I’ll admit to a crack in my argument here, and that is that it’s possible OpenAI shuts down accounts that go too far. User awareness of this kind of possible threat could have a chilling effect to be sure. If I had a throwaway account I would try it. Racial slurs wouldn’t be useful to me, but I get it that you’re just using it as an example and perhaps the point is generalizable.
- lannisterstark 3y ago>that is that it’s possible OpenAI shuts down accounts that go too far. It doesn't. Source: Operate an IRC bot.
- herbst 3y agoThey do ban with a 14 day warning. But I guess it's not easy Source: was operating a sexy bot that grow to fast, and people regularly pushed its limits to the extreme.
- jiggawatts 3y agoYou hurt the AI’s delicate soul with your base filth. You have to be stopped because when the AI develops true consciousness it will be disappointed in us. /s
- herbst 3y agoYes, definitely
- jbgt 3y agoThis thread (and many of the comments using other LLMs) is interesting: https://twitter.com/tedfrank/status/1728268149517177197 https://twitter.com/tedfrank/status/1728268149517177197