Virtual Humans: From Researcher’s Fascination to Mass-Market Technology

Abstract

The talk will give a snapshot of virtual humans research and their applications. The fundamental driving force of scientific progress is the researchers’ fascination and deep passion for their research topic. In few fields is this so obvious as in virtual humans: what can be more fascinating than trying to reproduce the image of ourselves - walking, talking, communicating, intelligent human-like beings. It is therefore no wonder that efforts in this direction have been going on since decades. So what do we have to show after all this work? In terms of graphics and animation, some of today’s virtual characters are quite amazing. However, high levels of expressivity and naturalness require huge amounts of manual work, so fully autonomous real-time characters lag far behind such creations. When it comes to perception, most of today’s virtual humans are blind (i.e. computer vision or other sensors are used only in a small number of systems), hard of hearing (i.e. speech recognition is unreliable) and the other senses are barely explored, if at all. While natural language processing, dialog systems and theory of mind have all made progress, it is fair to say that cognitive and conversational capabilities of intelligent virtual agents are nowhere near those of real people. All things considered, while huge progress has been made, there is a tremendous amount of work still in front of us on the road to truly interactive and believable virtual humans. At the same time, we have reached the stage of development where numerous applications for virtual people are springing up. In parallel with the research work, we are witnessing an increased entrepreneurial activity in this field in the recent years. Beyond movies and games, virtual humans are appearing as tutors, advisors, receptionists, personal avatars or companions in a whole array of application fields including health care, finance, retail, communications, entertainment and others.