In 2002, a lot of interesting stuff had just finally arrived. Between 1992 and 2002, I had been experimenting with the sort of stuff that is now commonplace.... composing an electronic score on the computer, adding live tracks, and calling it a recording. Where none of the equipment involved was especially or particularly expensive, just the stuff you'd already have with a few small pieces of added hardware and software to make things work together.
Around 1992, I found a nice tracker that could do interesting things with 8 bit samples. I had an Apple IIgs, which had the same Ensoniq 5503 that the Mirage and ESQ-1 keyboards had. Around 1994, I got a guitar. Things twisted and turned around and eventually I'd realized that I could record an entire song live if I kept to one vocal part and one guitar part and all the backing tracks on the computer. Eventually, I got a 486, which meant that I could bump up the quality level significantly. I kept at it until, around 1998 or so, I started messing with Buzz, which really helped bridge things together into a useful environment. Suddenly, things that required a lot of effort in an experimental studio were possible at home.
I recorded a fairly polished album. And then got bored with it all.
Upon getting the urge to play with music again, I figured that things would be dramatically improved and different and stuff. But... not really.
I have a mid-90s rackmount guitar multi-FX processor. There are a few things that newer FX processors do. I'm missing a phaser, a pitch shifter, a wah, and probably the ability to chain more effects in a row compared to the newer models. But, of the hundred or so effects that modern processors have, most of them are just variants on the existing effects, generally trying to simulate vintage FX pedals. Had I purchased a mid-end FX processor in 2001, there would be absolutely nothing that new processors could do that mine could not.
A good sampling synth in the earlier days would have a few megs of memory. This caused people to get creative with banks and patches and FX. These days, you can throw a billion samples into a multi-gigabyte archive and come up with a decent-sounding approximation of an instrument that completely misses the point.
There is much to a real physical instrument that defies keyboard control or simple sampling. If you play two notes on a guitar, the result is not the same as a pair of samples taken in isolation. In fact, various studio engineers have carefully dissected and recorded guitar parts string-by-string to get a particular sound unlike what would happen if you'd played the whole part. If you play a violin, where do you switch bowing directions. The player can move the bow up or down the strings to subtly inflect a tonality. There's a wide variety of extended techniques available for most of these instruments as well. Very few packages make even a vague attempt at handling this and, if you want that level of functionality, it's very expensive.
This extends to the user interface. Guitar FX processors started getting more and more dial-centric, because apparently nobody likes buttons who plays guitar. And most of the soft-synths I'm seeing are all fancy mimicry of how things might look on a super-powered synth with a 40x10 character display. It's neither useful nor readable.
There are a few pedals out there, made by "boutique" companies and often times costing a LOT of money, that can actually do something new and different to a guitar signal. I've seen some video demos of the pedals and they are quite impressive. I'm far too cheap to buy one, so I've been playing with making virtual effects on the computer instead.
There are plenty of new, interesting, and outwordly tones to be had this way. In most cases, you could make them in stompbox form, either because it's a straightforward modification of an existing pedal or because it's otherwise a fairly simple circuit. The easiest example is the wah wah pedal. It's one particular sweepable bandpass filter. Must all wah pedals be a bandpass filter? Not really, it's just that everybody has fixated on one particular bandpass filter and the emulation of every possible component swap of the analog implementation of a bandpass filter. A second example is the Pigtronix EP2 Envelope Phaser.
The same goes for our obsession with re-creating physical instruments solely using samples. There's lots to be done with analytical models for things that either defy easy sampling (say every possible permutation of an opened hi-hat) or would be inconvenient to physically create (say something that's half-bowed-string, half-organ).