While it is unclear whether this will actually happen (this is up to the people who will organize TEI '10, which will be hosted by the MIT Media Lab), it marks an interesting shift in the topics adressed, moving away from pure computer science and smart objects, towards a broader range of topics adressed, including design, arts and philosophy (which was also discussed in the closing plenary). I could not agree more, this topic deserves just that.
A big difference of TEI, compared to CHI and UIST, is the focus on 'hands on': The three days of conference include two days of talks and presentations, and one full day of demos.
Regarding the talks, a novelty introduced in TEI are 2-pages papers and 5-minute talks. These appeared to me to be much more dense and 'crispy' than their 15 minute counterparts. Five minutes are surely not much, but they force their authors to be precise and quickly understandable. I believe that the conference benefited from this format, as many more authors were able to communicate their ideas, keeping the TEI's single-track format.
A great portion of the work presented was related to multi-touch interaction on table surfaces. These included such that used different physical widgets (a camera, as well as acrylic sliders and buttons) connected to virtual elements on the table (this work was presented by Jan Borchers' Media Computing Group from Aachen), thin LCD-based multitouch screens (such as the FLATIR by Ramon Hofer et al.) and interfaces purely based on paper tracking and interactively projecting on them (by Do-Lenh et al.) and several authoring systems for live music or video performance (e.g. the MixTUI by Esben Warming Pedersen et al.).
The Royal College of Arts (London) was also present with two projects, one that generated energy by trapping insects, and one that explored domestic robots behaving a 'way that should not harm the user', plus a number of innovative alarm clocks (one of which started to randomly call people from one's phonebook at wake-up time, whereas the guilt would wake one up). Other rather arty projects included dress-up robots, that had to be dressed up in order to program them (ActDresses by Ylva Fernaeus et al.). Further work included research on determining how is a handheld actually held in the hand (HandSense by Raphael Wimmer), using shadow as a display by controlling the permeability of window blinds (by Marcelo Coelho) and a tracking system for fill level of glasses placed on interactive surfaces.
Overall, the spectrum of this conference is broad, and it was a very inspiring time. Especially the focus on hands-on interaction and the new 5-minute format contributed to the good overall feeling. I am much looking forward to TEI '10.