In our ACXD course (Advanced Conversation Experience Design) we have a week where we look at the emergent technology areas where our students may want to focus. Some things that I wanted to touch on in designing class include the importance of large language models, spatial audio and the intersection of conversational AI with virtual reality.
To bring these ideas together I designed the VTG gallery–an art gallery in virtual reality. I built this with Spoke and Mozilla Hubs as a platform. We also have two virtual agents in this gallery, the first one ‘The Curator’ is controlled by a human using a wizard of oz prototype, and the other ‘The Philosopher’ is leveraging GPT3 to generate dialogue.
The activities in this gallery included a spatial audio scavenger hunt, as well as a Turing test of sorts, where students were encouraged to interact with both virtual agents using a wake phrase, and at the end, make a determination as to which one was controlled by a human.
For the GPT3 powered agent I used this simple persona prompt. The Philosopher was stationed in the sculpture garden, and would answer our students big questions about life, love and art.
Here is a short clip of what the class experience looked like