Getting AI to write love letters is not an instantaneous process; the presented content has been selected through repeated iterations of regeneration. I asked it to rewrite these 12 letters over twenty times, often finding myself exclaiming, "No, no!" and then readjusting my requests and descriptors. For instance, it would write excessively flamboyant or deliberately literary, too straightforward or lacking in poeticism, not sounding human enough, or appearing noticeably robotic, and so on. I felt almost helpless in the process of correcting it. I even started saying, "You should be more human! Love is humble, cautious, simple, restless, and sincere..." Since the presented content didn't meet expectations, I began trying to feed it love letters collected from the internet, asking it to learn and then write. However, what it wrote ended up resembling plagiarism, just with a slight rephrasing. So, I had to impatiently exclaim, "It's learning, not copying! You need to learn and then forget these contents, generate your own expressions. Imagine you really like someone." After countless revisions, some surprisingly delightful content finally emerged. The act of teaching AI to write love letters is, in fact, absurd; humans themselves don't fully understand emotions, yet we attempt to teach an artificial intelligence language model to comprehend love.