At my work we were experimenting with something similar -- Citrix Metaframe. It's a great way to display a desktop remotely, but the only problem you are going to have is getting the mp3 audio over the cell connection. Most cell modem connections aren't that fast (9600bps to 19.2kbps) so the audio would have a hard time streaming from the server. Plus if your connection got dropped the music would stop until you reconnected.
Also, keep in mind that terminal server doesn't send the entire screen, just the parts that change. Also many of the terminal-server like programs are limited in color depth (most are 256 colors) and have a hard time doing bandwidth-intensive things like streaming video and audio.
I had the idea awhile back to have the carputer be an NT or Win2k server and give the passengers their own portable wireless terminal (Honeywell makes a wireless webpad with integrated LCD, etc.). That way each passenger could be listening to/watching something different. Someone in the front could have headphones on and be listening to mp3s, while 2 people in the back could be watching separate movies stored on the hard disk.
...maybe someday when the technology gets better we could do something like that.