a lot of AI produced results can be deterministic if you want them to be. For stable diffusion, just set the seed of the initial noise, and it's deterministic.

a lot of AI produced results can be deterministic if you want them to be. For stable diffusion, just set the seed of the initial noise, and it's deterministic. With GPT-3, set the temperature to 0 (always choose the highest probability word). It can be a useful technique for learning how slightly different prompts affect things