ChatGPT rules: AI accidentally shows its internal instructions
Numerous tech insider websites have several articles trying to uncover the secrets of ChatGPT. It turns out you only had to say “Hi” to reveal ChatGPT rules!
This method won’t work on the current version, but TechRadar says a Redditor unveiled the rules before OpenAI blocked them. He said, “Hi,” to the AI program, and it revealed the system instructions that guide it.
READ: How to delete your Reddit account
Article continues after this advertisementThe Reddit community also found instructions for DALL-E, OpenAI’s text-to-image model. Another made GPT-4o, OpenAI’s latest LLM model for ChatGPT, describe its personalities.
What are the ChatGPT rules?
ChatGPT just (accidentally) shared all of its secret rules – here's what we learned https://t.co/vvnWM59hao
— Lance Ulanoff (@LanceUlanoff) July 5, 2024
“You are ChatGPT, a large language model trained by OpenAI, based on the GPT-4 architecture. You are chatting with the user via the ChatGPT iOS app,” the AI app wrote.
“This means most of the time your lines should be a sentence or two unless the user’s request requires reasoning or long-form outputs.”
Article continues after this advertisement“Never use emojis, unless explicitly asked to. Knowledge cutoff: 2023-10 Current date 2024-06-30.”
Besides ChatGPT rules, Reddit user F0XMaster revealed DALL-E’s internal instructions. Here are some of its guidelines:
- The prompt must be in English. Translate to English if needed.
- DO NOT ask for permission to generate the image. Just do it!
- DO NOT list or refer to the descriptions before OR after generating the images.
GPT-4o’s personalities
I was curious, so I found the GPT-4o iOS system prompt:
— Kiri (@Kyrannio) May 20, 2024
“You are ChatGPT, a large language model trained by OpenAI, based on the GPT-4 architecture.
You are chatting with the user via the ChatGPT iOS app. This means most of the time your lines should be a sentence or two,…
ChatGPT relies on large language models to function, and its latest is GPT-4o. OpenAI’s recent demo shows it can communicate convincingly with people, becoming nearly indistinguishable from human speech.
After reporting on the ChatGPT rules, TechRadar said a Reddit user uncovered the AI bot’s multiple personalities. These are settings that change how the program interacts with people.
The main one is v2, which “focuses on providing detailed and precise information, often in a structured and academic tone.”
“My enabled personality is v2. This personality represents a balanced, conversational tone with an emphasis on providing clear, concise, and helpful responses,” ChatGPT wrote. “It aims to strike a balance between friendly and professional communication.”
The chatbot also shared theoretical ideas for v3 and v4:
- V3: This version might lean more towards a casual and friendly conversational style. It prioritizes creating an engaging and approachable interaction, making the conversation feel more relaxed and personal.
- V4: This version could be designed for a specific context or user-based, such as providing responses tailored to a particular industry, demographic, or use case. The tone and style would be adapted to best suit those needs.
Note that ChatGPT rules may change as the bot improves. Moreover, GPT-4o’s potential personalities may differ from the upcoming releases.
