Jaw-dropping Photosynth demo

http://www.ted.com/index.php/talks/view/id/129 - (7.5mins) Also here - http://youtube.com/watch?v=s-DqZ8jAmv0 - And much more here. My question is, how do they navigate so smoothly with so many hires images? Fast machine with lots o ram? (A: JP2 see below) I also really love the shared meta data capability. It has the potential to change advertising forever. I saw a similar demo back in July of last year: in theory - with 1 photo - you will be able to find where that photo was taken - and look at the surrounding area
http://labs.live.com/photosynth/video.html - I wonder if this would apply to really old photos - of old towns, etc - i guess you could - based upon the natural landscape/elevation. just imagine how this may help gps driving directions etc... you could actually view the drive, landmarks etc before you drive - http://labs.live.com/photosynth/videodemo.html

Similar technology? John Carmack Presents New Technology

UPDATE: My friend Jeff IM'd me: it is unique in a variety of ways....especially if the client-side catches up with the server-side operations and cataloging that is going on. That is why we all need to lobby the browser groups to support the JP2 standard ( which is being used for the picture displays). It is a very practical use of the technology especially the way they have done the algorithm for screen resolution/number photos = slate size and visual data sent to user. Essentially they are only sending data based on viewing area.....not image specific.........zoom in on a region more data for the new resolution is sent...etc, etc. that data is cached (apparently) so if you slowly panned in on every area of the thumnail views of a region....you would have all the data for every image......and there would be no progressive increase in visual clarity.....its all there already -most definitly or a wavelet algorith very similar....with a server-side web process catching the request for the zooms. JP2/JPG2000 - Link 1 / Link 2 /