Mathematically Speaking

Another try with Pi, this time when robots ask and humans answer.

If you caught my last article, you know that I work with Conversational AI and have for a long time. I tested heypi.com with a specific query in mind and followed it all the way through with my test criteria to evaluate how good it was at, well, conversation.

This time, I was looking for something different — an actual illustration about the way these types of bots can and should talk, but how that gives the user certain set of expectations. You see, I have something I want to say and I need help saying it better in a more visual way to say it.

Here’s how it went.

A good start because it committed to engage in the conversation.

I appreciated this engagement to my attempt to set a baseline test to understand if Pi could perceive the question itself (since I learned from last time that my ability to ask the right question is a major ingredient to a successful conversation with AI).

I decide to validate its understanding with an implicit confirmation of my own.