GVU Brown Bag: Is it Friendship if One of Us is Computed and Rendered?

About 10 years ago, I had the opportunity to build an early prototype of a virtual neuroscience instructor and apply it in a classroom setting; students were able to interact with prerecorded video of a professor and ask a set of questions about the human visual processing system. Based on the questions, the system offered video responses, suggested related follow-up questions, and pointed the students to supplemental information.

A few years after that, I began looking at applying virtual characters for customer service applications. At the time, the artificial intelligence components behind these systems was pretty limited. Even some of the products rolling out onto the web used natural language processing and case-based reasoning systems that required a lot of handcrafted' hocus-pocus' to make it match up with the domain. All and all, I felt that the ratio of A-to-I work (artificial work to intelligent work) was a bit too high in these systems -- too much human involvement to make it even remotely believable or useful.

Fast forward 10 years, and Larry Hodges, formerly of Georgia Tech, now of UNC Charlotte, brings us up to speed on the field of virtual characters and sheds some light on what is really possible, and more importantly, what is really effective when trying to place these characters into "serious" situations. This talk is from from the Georgia Tech GVU brown bag series earlier this year. Larry Hodges at GVU Brown Bag 2/7/2008: Is it Friendship if One of Us is Computed and Rendered?