What many people don't seem to want to hear is that most "ML" problems are simple ML problems. LLMs are a neat advancement and a great tool in a variety of appl

What many people don't seem to want to hear is that most "ML" problems are simple ML problems. LLMs are a neat advancement and a great tool in a variety of applications but for most applications they introduce a great deal of uncertainty and often at significant cost. They're a good complement to great design and infrastructure but you're fooling yourself if you think they're a replacement.