Sorry in advance if I repeat myself...
Check out
NUI Group to learn all about multitouch tables. It's an open, community effort to examine natural user interfaces. There are lots of DIY tutorials for building multitouch tables, and the community there is excellent. (Yes, neither Apple nor Microsoft "invented" multitouch. MS is being pretty pioneering, though; you can't knock Surface for being a fantastically responsive and nice looking piece of hardware).
I highly recommend the Diffused Illumination method as opposed to FTIR, although FTIR did previously have more attention from the do it yourself crowd. (I suspect that its ridiculous complexity drew many people to think it was better, or at least had them using the method out of pity for the poor soul who designed such a complex system only to have it beaten by something stupidly simple). Diffused Illumination is easier, cheaper and gives better results than FTIR.
A Diffused Illumination table (like Surface) can recognize things that are not living human appendages, which makes it very useful for very cool things that connect physical and computer interaction. For example, using real paint brushes in a paint program, or having the desktop recognize peripherals placed upon it and popping up a quick tutorial for how to use them, or tagging objects and having those tags recognized by the table. (And, of course, photocopies of yourself).
The larger surfaces like these are "optical touchscreens", which are actually ingeniously simple, and fantastically cheap!
The diffused illumination design uses infrared light (what would we do without it?). Simply a webcam with a filter to see only infrared and a few IR lamps shining out of the table (into the operator's hand or really whatever else is on the table). The reflected light is of course seen by the camera.
The real magic happens with the composition of the table's surface. This surface must diffuse the IR light so that objects far from the surface are not being picked up as touches. (Conveniently, though, with this design one could lift his finger slightly off the surface and still have it tracked, and the system could even guess the distance at which his finger is lifted, making for some cool stuff. This allows for natural flicking motions).
That part is actually incredibly inexpensive. $150 for a big ass multitouch table, and you can make that thing really big, with the only real barrier being the size of the glass (and more lamps). Of course, to get the full experience you need a digital projector (and maybe a mirror for the projector)...
The possibilities are indeed limitless. I think Microsoft is being a bit premature with Surface, though, as it looks to be an entirely new kind of desktop. That isn't a bad thing (and really it's the only thing they can get away with since they're doing it for profit), but the technology will take a while to get into the mainstream, and I, for one, want to use multitouch for everything!
The
MPX project is a modified X server for Linux which works with multiple touch points. The nice thing with that is legacy applications continue to work without modification within the same environment as multitouch-capable apps, and indeed the legacy apps do take multiple touches, if you are gentle to them.
This video has some great examples of how multitouch can improve interface.
I think it is really great what these can do for games. I'll admit to being a bit of a first person shooter person, but for once I couldn't care less about first person shooters on this kind of an interface. Since they get so much nicer to use with multiple flexible touch points, there can be some really intuitive but completely abstract games that feel a lot more natural than before.
With a keyboard and mouse we are being very limited by the rules of the computer right from the start, and these rules are very different from the rules that we, as humans, are used to interacting with. In the computer world, we have a single hand with about three fingers and language consists of a lot of buttons. Compared to the real world, this is dog slow and unintuitive. (Keyboard excluded, of course).
With a multitouch interface, those rules are (mostly) our own. We can communicate using any object of the real world, limited by our own limbs instead of the computer's limbs. Sure, it's still a two dimensional flat panel that we're talking to, but it is a significant step forward for computer interfaces opening many doors. Thanks to how much information can be gathered completely transparently by a multitouch setup, software to support multitouch will be very future proof and forward thinking, making far fewer assumptions than the "poking with a stick" interface we have become used to. That kind of design would hopefully mean an easy transition for when that ultimate interface (full 3D input and output please!) finally does roll along.