Raw is, if you like, 3 images. 1 for each colour. The data is linear. Not at all how your eye sees, or for that matter how your computer shows images. It's, well, raw. And that's why its spelt raw, not RAW. It's unprocessed information - just as the sensor caught it.
The main advantage over JPG, is that it is not logarithmically compressed, so the range of data truly represents every photon captured (and noise too, but that's true of JPG's as well).
Since the camera has done no processing (well the Sony's do do some noise removal on raw) the data can played with in many ways most of them reversible (which is not often true of JPG).
The camera settings are tagged into the raw file. So if you shoot a specific WB, contrast, NR, etc. settings, then that's what will show in ACR. But ACR is _interpreting_ the settings, not showing the raw file as recorded - nothing can! The settings only affect the way you _see_ the raw image, they don't change the raw image at all.
Raw lets you correct for light temperature and tint to the full range of source light characteristic. This is much more limited in JPG.
As an example, let's say you shot one evening indoor with the WB set to tungsten. The next day you take a shot of your kids in the park in the sunlight, but you forgot to change the WB to sunlight. The image will be very blue. You can correct the JPG, although far from "true".
OTOH, raw has no care about the source light characteristic, so you can easily set the WB on raw import.
Beyond that, in editing, you're working with the uncompressed data in its full bits/colour. So with each edit function there is less loss than when working with a lower number of bits that have been compressed and re-compressed, sometimes many times. (This has nothing to do, BTW, with the JPG image compression on save, that's another matter).
With PS CS3's raw converter I find that almost all of my basic image corrections for colour, contrast and so on are accomplished before the image gets into PS proper. This is fantastic when all of the images are shot in the same conditions, then I get the settings right and apply the same settings to all of the images as a batch in raw converter. That saves a hell of a lot of work.
Yes, raw files are larger. Often 33 MB or so on the a900. I save computer disk space by converting to DNG when moving the files from the data card to the computer (using Bridge, supplied with CS3/4). There is also a free DNG converter from Adobe. Most editing programs recognize DNG. With DNG those raw files are reduced 40 - 50% in size (Lossless compression).
You can also, depending on your model use sRaw which is lossy-compressed, but I don't know of anyone yet who has seen a real difference in the files. (Memory's cheap - I don't use sRaw).
Raw also gives you much better recovery of underexposed images (2 stops, even more) and overexposed (1/2 stop at least in most cases) so you can recover exposure boo-boo's much easier than with JPG where brightening an under-EXP results in muddy tones and lots of noise and calming down an over-EXP results in horrible blocking. (You may get some posterization on brightening a 2 or 3 stop underexp in raw, but it's not nearly as ugly as recovering an under-exposed JPG).
You can also "expose to the right" in raw, where you move (usually with the aid of the histogram) the exposure to the right of the histo. This lifts shadow detail. Then in PS you "de-expose" the image to restore the tones to the right level. Using the black point you can conserve shadow detail. For example if the entire correctly exposed image was no brighter than middle grey, then you could expose it two stops brighter. It will look over exposed in the camera monitor. Back in the "lab" you re-expose it properly by moving the black point up. This is much less versatile when done on JPG.