When an easier interface makes machine learning useful
A lot of business problems are classification or regression problems. Classical ML can solve many of them, but you have to label data, train a model, and host it somewhere. That's a lot more work than calling a text-based API, and companies need to know those methods exist in the first place.
I've used LLMs to help build classical ML models and clean data that would have taken me much longer without them. Jev got me wondering whether an easier interface could make classification and regression accessible to more people.
I still want to understand the tradeoff. If using a model like Jev well takes substantial know-how, at what point would you be better off using an LLM to help you build a classical model? I'm interested in trying it, but I don't have a settled answer.
Originally posted on X. Lightly edited for reading here.