Start training your expert chat like an employee
If you're treating GenAI as a chat assistant, it will be nothing more than just that... a chat assistant (ok, maybe it will be a bit smarter than those rule-based chat assistants that has missing paths and dead ends resulting in weird error messages...but we will likely not see any transformative outcomes that are really impacting our ways of working).
Yes, I'm perhaps stating the obvious here, just like the famous quote:
‘Insanity is doing the same thing over and over again and expecting different results’
I'm anyways sharing this experience and reflection as it was a bit of an "Aha Moment" for me last week when I joined the prompting course hosted by Everyday AI and Jordan Wilson. I've been trying to give Copilot, ChatGPT or any of the technologies I've used as much context as I can, but I've always done it in one go, ending with the prompt. This has resulted in somewhat more or less valuable outputs, many times leaving me frustrated having to do "too much" of the work myself and wondering why I read about all great productivity cases and scenarios where people get a whole lot done by interacting with GenAI.
Now I at least have a methodology to follow with some key principles that help me evaluate the cases for engaging with GenAI and for my prompts to be more effective. The notion of training your expert chats like you train your employees is something that is now stuck to my brain as well as a key takeaway for me: "you are not looking for an output, you are training a skillset". The course presents the method of PPP:
领英推荐
I'm not getting anything in return for recommending the course but I simply just want to do a shout out and say: if you want to improve your prompting skills and have an hour to spare to take this free course provided by Jordan - it's time well invested!
I'd also be happy to hear more experiences on what works well when prompting, your methods and what value you've experienced when interacting with GenAI chats. Let's learn together!
Flank
9 个月I really like the framing here of "giving the AI context" - so often I've made the mistake of essentially writing list of "dos" and "dont's" as a prompt. This tends to become messy. I would add consistency of style as being something I've learnt in prompting. When interacting with an AI, my experience has been that the human sets the tone, not only by instructing the AI how to reply (e.g. "reply succinctly"), but also in your style of writing. So it's worth taking the time to ensure your prompt reflects how you want the conversation to feel!