A follow up to my previous LLM explainer. In that post, I discussed the recent evolution of LLMs and why the increase to their context window length improved their usefulness significantly.

In this post I will discuss one of the limitations of LLMs, and that is that they are a simplified representation of … their training data.

The M in LLM stands for ‘Model’.

The OED definition of model:

There’s been quite a few articles that’s been written about AI all over the internet, with various technical depths and contexts. But I find that there’s a opportunity here to provide an explainer without specialist ML/AI/Statistics background knowledge and jargon.