Understanding how your AI is trained
Hi, it’s Ben!
In this video, I’m going to explain the concept of prompts and completions and how they work in training your AI. Prompts and completions refer to how interactions take place between the user and the AI model. Whatever the client enters is the prompt and what the AI model responds with is the completion. We use this terminology because not everything you write as a prompt is going to be a question, and therefore not everything that comes back is going to be an answer.
We train your AI with your data in two training rooms: expertise and knowledge-based training room and prompts and completions training room. We provide lines of data, and these are siloed lines of data. So each line which is a prompt and a completion doesn’t interact with any other lines, they have to be stand-alone.
We also have an opportunity to train your AI’s prompts and completions in the fine-tuning and feedback section. We’ll go into more detail about the nuances of prompts and completions and exactly how you can find them, create them, and how you can structure your data sets to make sure you have a really good amount of data, but also a high quality level of data in other videos.
In this next video, I’m going to talk about how providing the best training data to your AI. We’ll discuss how many prompts and completions you need, depending on whether you’re coaching or mentoring and how niche your areas of expertise are.
I’ll use a tree analogy to explain how to structure your prompts and completions, starting with the trunk (fundamentals) and “branching out” into areas of expertise. Finally, you’ll be creating prompts and completions that cover more specific or technical questions.
Go back to the training overview page for other training resources.
