I notice many different ways of uploading in stories. Since we never UpLoaded anypony before, I cannot rule out any of them. That said, using Occam's Razor, this seems the simplest way to upload:

1. Remove brain including removing it from DuraMatter (the membrane surrounding the brain which contain its BloodVessels) (if the pony is not dead, the pony is dead now).
2. Chop the brain into little cubes (this allows for parallelization of the scanning process).
3. Scan the 1st layer of each cube.
4. Grind away a micrometer of tissue.
5. If tissue remains, go to line # 3; but if else however, stop.

One has just scanned a brain into a computer.

One could directly simulate it, but this would require enormous memory, CPUCycles, and the simulated brain would run slowly.

It is much more efficient to extract the memories, figure out what the brain does, Convert what it does to ObjectCode (the MachineCode running directly on hardware) and directly run the upload on CompterHardWare. Since the computer would need its own operating system, as well has applications for supporting the running of the upload, the upload would technically run in its own virtual machine. This would use vastly fewer resources that simulating the hardware while allowing the uploads to run vastly faster.

1. Scan the 1st layer of each cube.
2. Grind away a micrometer of tissue.

That's presently done with cutting

"It is much more efficient to extract the memories, figure out what the brain does..."

The method of extraction remans to be developed, but if anything ever works, it would likely be based on nanomachines that would infiltrate the brain to the point where the system could capture a "snapshot" of the live operating state of the brain within a very short amount of time. Consciousness is a dynamic system, after all. Without a "vector" component for each datum, the model would be unworkable.

As for figuring out "what the brain does," it's an even larger challenge than that. The brain is very dependent on blood chemistry, which means that any computer model would have to take into account the (simulated) state of various other organs in the body. Hormones do more of our thinking for us that anyone would like to admit.

Since the brain is an evolved and accreted organ, there's also the problem of all the parts that don't need to be modeled. No use wasting clock cycles on something that could be a simple, undynamic structure. The bigger problem is that we don't know for sure which parts fall into this category, and we may have to find out by building a full model and then selectively pull out pieces until it fails. And given that the simulation may soldier on when effectively crippled, it may mask the point (or spectrum) of failure. (Also, not a fun method for any personality being simulated.)

TL:DR— If this was easy, everyone would do it.

1. Inject stimuli to original
2. Record results
3. Train transformer on stimuli result pairings.
4. Discard original

Results may vary.

7911253
My idea:
1. Remove cranium from around the entire brain.
2. Insert a gazillion electric measuring needles all at once. Create an extreme high resolution electroencephalogram.
3. Remove needles. Flash freeze the brain.
4. Bring in a microtome. Remove layers no thicker than the diameter of a neuron. Photograph at enough resolution that you see every axion.
5. Simulate the physical brain in the computer.
6. Extract the software, then dump the simulation.

7911351
And then, of course, after extraction we will have to work on inception.

7911351
Not just hormones - a crazy amount of thinking is affected by your gut bacteria. Heck, look at some people who literally develop temporary disorders because of a UTI.

7913023
Absolutely! Environmental factors as well. To simulate a harried, hangry parent, you'd need tertiary simulators for a bored child, endlessly asking, "Are we there, yet?" and a McDonalds with a broken milkshake machine.