[Opinion] Prompt Metadata as Signals
A lot of people don't seem to have realised how important the aggregated meta data around prompts is as a signal for identifying future change. It's not much different from 2006-2010 where a lot of people didn't seem to realise how important the aggregated meta data around consumption of cloud services was a signal for identifying future changes. In both cases, it's easy to use an ILC model.
I know this is not new, and we've had plenty of discussions on ILC in the LLM/GPT space. I'm still just amazed at how something so obvious gets missed by so many, especially since we've already been through this same situation around 15 years ago. I could sort of forgive boardrooms being clueless last time but how many identical situations do you need to go through before the penny drops?
I say 'not much different' because at least this time some of the players (e.g. AWS with bedrock) understand the game.
Originally published on LinkedIn.