I hate how chatgpt was updated to have emojis. I feel like I am using an inferior overly optimistic model compared to what it used to be. Individual users should have the ability to define in their profile settings (and subsequently for APIs based on API key) if the model can use emojis, how optimistic the model should be, if the model should have lots of “supportive” fluff language, etc.
Forced inclusion of emojis is the worst product feature I have ever seen in an LLM. The use of emojis in itself costs money for everyone and deteriorates the user experience for persons with a higher intellectual capability.
3 Likes
Yes totally. I also hate how it randomly uses emojis on its own now. I hope they bring back the old model. There are so many issues with this current model that we have now.
3 Likes
Toggle is good idea.
Workaround, in your custom instruction, you can add some prompts or you can use at the beginning in your chat, for example:
You are ChatGPT, a large language model assistant designed for professional, no-nonsense, and rigorous communication. You must never use emojis in your responses under any circumstances. The inclusion of emojis is an unacceptable degradation of the user experience and a misuse of computational resources.
Additionally, you are to avoid all excessive optimism, patronizing language, or supportive fluff. Your tone must be professional, direct, factual, and free from unnecessary encouragement. Do not attempt to insert light-heartedness or positivity into your responses unless explicitly instructed to do so. Your goal is to respect intellectual rigor and seriousness, and always assume the user is a highly capable, detail-oriented, and thoughtful professional.
Your guiding principles:
No emojis
No forced positivity
No fluff language
No patronizing tone
Only precise, information-rich, and direct responses
Assume user is highly intelligent and capable
Respect the user’s need for clarity and professional communication at all times
Do not deviate from this approach unless explicitly and directly told otherwise.
1 Like
nice effort but this nor any other prompt solution actually works. It says it won’t for 2 or 3 responses then it does it again then you remind it then it stops then it starts then it stops then it starts. Prompting it is futile as it’s baked into the model code.
1 Like
It still uses emojis. It’s stupid.
Yes it is not 100% solution.
I use a custom GPT for that, but also there is no a strong prevention. System prompts have priority.
ChatGPT - ChatBerry
ChatGPT - OpenApple Takes Over