A skeptic's perspective on LLMs argues that the real value isn't in generating content but in using them as functional building blocks for processing tasks that are hard to do with traditional programming. The author introduces the concept of 'function slop' to describe pointless LLM usage (like using an LLM for addition), and argues LLMs shine when used as engines behind functions for tasks like summarization and categorization. The piece concludes that knowing about LLMs doesn't mean you have to use them for everything.
Sort: