fbpx
Magai
Sign Up
Log In

Understanding Variations in Magai’s Responses

Question: Why does sometimes Magai responds differently than the default OpenAI’s ChatGPT even with the same prompt?

Since AI models are engineered to avoid writing the same exact message twice, you’ll always get a different result—whether that’s in native ChatGPT or in Magai.

Then there’s the question of Personas or custom instructions.

  1. Did you have a Persona selected in Magai?
  2. Did you have any custom instructions in ChatGPT?

Magai also puts a max tokens limit on outputs to ensure the highest quality. The reason for this is that quality decreases with higher output. Asking for the AI to write 1,200 words at a time will typically yield lower quality results than if you were to request 300 words 4 times. We also talk about this in our article: ChatGPT Writing Process Steps for 10x Better Results

In short, here are a few key reasons why Magai may sometimes respond differently than ChatGPT even with the same prompt:

  1. Personas and custom instructions: Magai allows users to select Personas which can influence the responses. ChatGPT does not have predefined Personas. Custom instructions in ChatGPT could also lead to differences.
  2. Maximum token limits: Magai sets a max token limit on responses to help ensure quality, whereas ChatGPT responses are not limited in the same way. Longer ChatGPT responses could diverge more from a given prompt.
  3. Iterative training: Magai and ChatGPT are trained separately on different data over time, so their underlying models will evolve somewhat differently even when starting from the same base (e.g. GPT-3). This impacts how they respond to prompts.
  4. Context & memory: While both preserve conversation context, subtle differences in how context is stored and retrieved could influence responses to the same prompt later in a conversation.
  5. Non-determinism: Like all AI systems, neither ChatGPT nor Magai are perfectly deterministic – small internal variations mean the same prompt may not always receive identical responses.

In summary, while similarities exist between the two platforms, factors like Personas, length limits, separate training, and non-determinism can all contribute to some variation in responses to the same prompt between Magai and ChatGPT. The emphasis for both is on coherent and helpful responses rather than precise duplication.