It looks to me like maybe a different approach to the problem – but still the same problem. The robot must look at different images which are similarly tagged, and then attempt to recognize untagged objects as belonging to the same set. In the Google problem, there’s nothing to stop them from loading their AI with thousands of tagged images to compare to when searching for images that fit into the same set. The challenge is the same – only differing as a matter of scale, since the Scavenger Hunt will of course be much narrower in scope than a Google search algorithm.
This looks like it would be a good conference to attend – especially as Vancouver is such a beautiful city. I expect I may be too busy in July; but if I can wangle a business trip around that time to Seattle, maybe I can go.