When I asked a Large Language Model (LLM) where it was most likely to make a difference, it listed topics that are not immediately useful for a user. Sure, if we spend a lot of time trawling around the bowels of the API, model, or attaching vector databases, we can get it to perform investment research or streamline analysis. But right now, even without having to muck around in the depths, these models are already really useful. Here is a list of the areas where I think you can use a model like ChatGPT today.

  • Proofreading. I use it to proofread emails, short documents, or even my CV. Obviously, I ask it to fix spelling and grammar, but its real strength comes in if you ask it to change the tone of an article. Ask it to be more humorous, more professional, simpler, or almost any style depending on the intended audience. Don't forget to read it afterwards. LLMs hallucinate.
  • As an onramp to the internet. It is especially good at research. Imagine trying to come to terms with a new concept or to compare the product offerings from many companies. ChatGPT (or Bing enhanced with GPT-4) can really speed up the process so you don’t have to do all the initial grunt work. If you treat it like a graduate trainee and take its work with a pinch of salt (it is repeating the sentiment in its training set, not giving you unbiased analysis), you can get up to speed very quickly.
  • Content Generation. With just a few constraints or pointers, LLMs can create very creative content. The art here is to provide it with the correct constraints and to try many alternatives, don’t settle for the first version. With computer-generated content, you need to kiss a lot of frogs before you find your prince. The ‘content’ the model covers can be very broad. I have used it for dinner menus (with shopping lists), birthday party ideas, code, and thank-you messages. I am sure you have many other problems that could benefit from ChatGPT.

Having played with the models a bit, my view is that these models are the equivalent of computers made out of vacuum tubes - big, slow and expensive. But, crucially, they are the future. However, what strikes me about these models is the rapid speed of their evolution. It took decades to move from vacuum tubes and punch cards to carrying the world's content in your pocket. LLMs are going to change the world in a much shorter timeframe. Don’t blink.

6. Where do LLMs fit in?