The problem is that nowhere does the math say that the cat actually is alive or dead. Again, mathematically this is a consequence of the fact that the Schroedinger equation is linear, so if you start with a superposition of states, that can only evolve into another superposition of states. The math says unambiguously that the cat and all of its observers are in superpositions. In order to extract a single privileged objective classical reality you have to add some additional mathematical constraint, and no one has been able to figure out a way to do that that isn't either ad hoc or at odds with the data.
And it gets even worse than that because of the Bell inequalities, but that's a whole 'nuther kettle o' worms.
There is no such special thing as a "classic observer": all instruments and people are in the end made out of quanta, and are themselves huge quantum systems.
There is no such special thing as a "wave function collapse from observation": what happens is that the observer (a quantum system) becomes entangled with the observed system. That's it. Initial state of "[ignorant observer] * ([cat is dead] + [cat is alive]) / sqrt(2)" which is equal to "([ignorant observer, cat is dead] + [ignorant observer, cat is alive]) / sqrt(2)" evolves into "([observer thinking of a dead cat, cat is dead] + [observer thinking of a living cat, cat is alive]) / sqrt(2)" without breaking the unitarity.
Yes, for "the" observer, her being part of this state, it looks like "the collapse has happened". That's how you get "many worlds": the state of the universe is a huge sum of independent states that continue to evolve independently, each of them evolving further into a sum of independent states.
(Of course, given that the very existence of the Geiger counter or computer may depend on which branch occurred (perhaps there was a bomb on one path?), it's unclear whether you are free to think of there even being an "it" that could have an answer, either.)
At the quantum level, the quantum degrees of freedom in the Geiger counter or the computer still exist even if a bomb goes off. They just have a different relationship in terms of interactions than they would if the bomb didn't go off.
Do note, however, that the wavefunction and its collapse are at odds with special relativity and thus cannot be an object of reality. (Provided you're interested in realism to begin with.) Take two entangled electrons, for instance, which get sent out from the origin 0 in opposite directions towards detectors A and B, respectively, which are placed at equal distance from 0.
Now take two observers, one moving in the direction from 0 to B and one moving in the opposite direction. Then, depending on which observer's POV you take, it is either detector A might makes the wavefunction of the two electrons collapse (because the electron 1 reaches detector A and interacts with it before electron 2 reaches detector B) in which case interaction of electron 2 and detector B doesn't do anything. Or, vice versa, it is detector B which makes the wavefunction collapse but then detector A doesn't do anything special.
Either way, if the collapse were actually something physical, both observers would have to agree on the causality chain of events. But they can't. In fact, varying the above setup a bit, they won't even in general agree on whether the wave function of a given system has or has not yet collapsed.
In my utterly uninformed imagination I've reconciled this by defining observation as the point where the observer (whether it be a simple apparatus like the sensor on a photon counter, or a complex and squishy one like my consciousness) becomes "intertwined" with the observation. i.e. Once the information has reached some particle, that particle's view of the waveform/event has collapsed.
It's easy to forget the actual conveyance of the information has some physical manifestation (photons hitting my retina, electrons bumping against transistors in an IC and eventually manifesting as dots on a screen which send out those photons) and I picture those causal steps as propogating constraints through the universe. Light speed sets a maximal boundary on how much of the universe may be "intertwined" with the observable, but even if the data is sitting on the computer next to me, until I take a peek (or otherwise subject myself to any consequences stemming from them) those constraints haven't yet permeated any of the bits of the universe that make up me.
Akin to how relativity permits two different observers to hold differing views of some phenomena (eg. silmultanaety) this worldview allows me to imagine the cat is both dead and alive even though my computer or the Geiger counter may know the correct answer.
I'd love to hear from real quantum physicists whether this interpretation is bunk or has some validity (and if so, whether someone else arrived at it before me and gave the theory a name).