2 ms·
If it's filtering by taking the output of the first model as a prompt (with some framing), then that is equally susceptible to prompt engineering. Indeed, you c
by int_19h 4y ago
If it's filtering by taking the output of the first model as a prompt (with some framing), then that is equally susceptible to prompt engineering. Indeed, you can already tell ChatGPT to write a prompt for itself to do such and such, and it will do so. You can even tell it to write a prompt to write a prompt.