Raw LLM Responses

As someone who took ML/AI courses 4-5 years ago as these were all breaking towards the mainstream, people need to appreciate (but don’t) that under the hood all these models are is very, very large training datasets being governed by complex multivariate statistics with calculus layered on. In brief, it’s a complex set of weights and biases used to create and modify association between your input and its output. I won’t pretend to remember the nuance of these LLMs (like I remember learning about how they maintain correct grammatical sentence structure but fuck me if I remember the details) but at their core, they’re just numbers stored in RAM/Config files in a container/server.