So I’ve been asked to prepare some material specific to certain user groups to help with future onboarding tasks at our organisation i.e. steps to gain access to various user domains, systems, network locations. etc.
I’m wondering if there are some useful guidelines on how best to structure information for the purposes of training language models for this purpose. The information that I have thus far includes documents with bullet points and links. I also need to supplement this info with the learnings from my own experience going through the onboarding process i.e. dealing with online IT support chat line, emails, etc… It’s cumbersome to say the least, with constant challenges in a continually evolving IT environment!
Looking ahead, I’ve also suggested to my team that we consider using a language model to develop an “in-house view” of how the company interprets certain regulatory specifications which are often open to misinterpretation and result in much wasted repetitive effort going through the same documents time and time again.
I think of LLM technologies such as ChatGPT as the best versions yet of experiential knowledge based systems, that can be trained on proprietary information for future reference.
As I near completion of the Deep Learning Specialization course, I’m already being drawn into applying some of what I have learnt. I would say I have gained a good appreciation of how language models are built (their architecture) but I am by no means an expert, and now find myself about to gain some hands-on practical experience preparing material on which a language model will be trained/applied.
Any thoughts/suggestions/feedback most welcomed!