> I also think LLMs are wildly over hyped, and they will be shunted to a niche in the next few years, essentially providing autocomplete or search for business tasks for the most part, with strong inroads on video analysis, automated translation, and other classic machine learning & neural net tasks.
There are many useful tasks that LLMs can automate and they will provide great value for humanity. The problem (for investors) is that it's hard to make people pay for it or make money out of it. This is the same problem Google Search had, and ads seem to be the only way forward here as well.