Like in previous years, the big theme here at ISMAR (the International Symposium of Augmented and Mediated Reality) seems to be registration and tracking — how to detect where objects and people are in the physical world so you can overlay graphics as accurately as possible. AR isn’t my main field, but I’ve had a couple of conversations so far about how we’re really reaching a point of diminishing returns. It’s great that we’re seeing minor incremental improvements in this area, but what we’re really lacking are new, innovative uses of AR to push the field further. Unfortunately, it sounds like at least in part a lot of these new innovations didn’t make the cut for the conference because they lacked in strong evaluation or quantifiable contribution to the field — it’s much easier to judge the quality of a new camera-based image-registration method than it is to judge the usefulness of a brand new application.
The Software Agents field was a response to a similar stagnation in Artificial Intelligence. AI researchers had a lot of good but imperfect tools that had been developed over the years, but kept trying to solve the really hard general problems. Software Agents grew out of the idea that it was OK if your algorithm wasn’t perfect in every condition so long as you cleverly constrained your application domain and designed your user interface to cover for those imperfections. It was a struggle to get acceptance of the idea at first, and in the end a few of the big players in the new domain went and founded their own conference rather than try to fit their own work to the evaluation metrics used for more traditional AI papers. Hopefully it won’t take such a dramatic move on the part of AR researchers to breath new life into this field.