To enhance the usefulness of GPTs, incorporating more automation options would be beneficial. This would allow for lengthy processes to be executed extensively without requiring new inputs.
I recognize that cost factors play a role and need to be considered. Therefore, in line with the existing restriction of 50 inputs every 3 hours (or a similar framework), I suggest introducing an automated response feature.
In this proposed system, the GPT creator would determine the subsequent inputs for each new interaction. For instance, after a user’s initial request (input 1) like “Create a Python code that checks whether a number is prime…”, the GPT’s creator would predefine the following inputs, such as:
input2 = “Analyze the previous code step-by-step for performance improvements. Then, develop an enhanced version.”
input3 = “Evaluate the previously proposed code for potential enhancements. Create an improved version.”
And so forth.
Take another example: If the ‘Instructions’ determine two continuation options (A and B) at the end of every response, a simple command like “You choose” could enable the system to operate indefinitely.
This approach would significantly broaden the scope of potential applications. The ‘Instructions’ provided could be tailored to anticipate each input, making the system more deterministic and goal-oriented.
In practice, as the GPT creator decides on the number and nature of automated interactions, users would be informed beforehand, for example: “This GPT executes 15 prompts in a single interaction.”
Currently, the only method to achieve this is through an external application requiring an API key. This not only creates a higher barrier to entry but also incurs additional costs. Allowing average users to explore GPT-4’s capabilities within the ChatGPT interface, utilizing the standard limits of the Plus plan, would be far more advantageous and user-friendly.