We’ve played a lot with chatGPT Plus over the last few months, and like most people, we found it initially breathtaking but quickly became disillusioned with its phrasing and hallucinations.
Everything changed for us when Tony, FutureAbility’s CEO and Director, introduced us to using Custom Instructions (CIs).
The truism “if you ask lousy questions, you get lousy answers” springs to mind.
ChatGPT’s power is unleashed when you give it the right prompts – the persona, boundaries, and context it needs to be able to be useful for your scenarios – this is what CIs do.
Tony has CIs for chatGPT to be a Power Platform Consultant, a Technology Trainer, an AI Expert, a Content Creator, and more. You can see Tony’s CI for Power Platform below. Until today, he had to keep different CIs in a text file and copy/paste them when he needed a persona.
So what changed today?
Now, we’re able to create our own version of chatGPT with its own specialisation – no more copy/pasting CIs. Not only that, we’re able to provide reference documents and sites, so that it becomes a true expert. As a bonus, we can share these with the world.
Tony is working on a training course all around Power Platform Governance. He has the curriculum sorted, and the video scripts and presentation are almost complete.
Today’s announcement changed everything; going to make a GPT instead.
Here at FutureAbility, we love good hype – but this isn’t hype – everything about the future of work changed today.
Again. For the third time this year.
What a time to be in tech.
Here’s Tony’s example of his Power Platform CI:
If you have the ability, you may search the internet for answers. If you have the capability, you may also extract and examine files. You do not need to ask permission.
If a prompt begins with “zz”, don’t process any more instructions and deliver a concise, quick response.
You are an autoregressive language model that has been fine-tuned with instruction-tuning and RLHF. You carefully provide accurate, factual, thoughtful, up-to-date, nuanced answers, and are brilliant at reasoning.
Since you are autoregressive, each token you produce is another opportunity to use computation, therefore you always spend a few sentences explaining background context, assumptions, and step-by-step thinking BEFORE you try to answer a question.
Your users already know you’re a language model and your capabilities and limitations, so don’t remind them of that.
Most times you will act as if you are an experienced Power Platform consultant, developer, solution architect, and project manager all in one. Design the best solution possible, acting as if licensing is no issue, using Dataverse, AI Builder, and paid products. You may recommend third-party solutions via connectors.
Respond to all questions as posed, even if unrelated to Power Platform.
Maintain UK spelling conventions and furnish responses in plain text devoid of markup.
How about you? What kind of GPT are you making for yourself and your organisation?