Fonts on a computer are defined by fixed forms: a 12-pt Times Roman "A" in a given font is typically a bitmap (for the screen) or a series of drawing commands (for a printer). Regardless of the method, the letterforms are fixed; there is no flexibility in them; even a "handwriting" font is lifeless and unexpressive compared to real handwriting.
How could a truly handwritten font emerge from a computer? I'm not sure, but I have an idea for an approach to study: define characters not as forms, but as methods. Each letter would be its own program (or program-like thing), with various input parameters that were sensitive to what was going on around it.
For example, the letter "a" would know from what direction it was arrived at, and with how much force and velocity. It might also know what letter came next, so that it would be able to get there in a graceful way.
The result of this would be a "handwriting" which had a clearly definable style, but in which no letterform was defined beforehand -- it would emerge in the act of "writing."
There are many nice things about this approach. The most obvious one is that letterforms defined this way would be very easy to modify to create variations. It would also be possible to extract features from a person's own handwriting to create a handwriting font which really worked the way theirs did.
Another possibility would be to connect this with voice recognition: the inflections and stresses of your speaking voice would be reflected in the handwriting on a syllable-to-syllable basis.