We rack our brains making prompts for #StableDiffusion and Language Models. But a lot of prompt engineering can be done *automatically* using simple gradient-based optimization. And the cold calculating efficiency of the machine crushes human creativity.
