For brain uploading, my money is on pure algorithms. Throw a lifetime of data on top of base human emulation software, possibly influenced by known genetic data, and then iterate through an insane number of possible brains using an insane quantity of CPU until you get a virtual brain that most closely matches up with the provided data. It'll be at best a crude approximation of the original human, but just like with MP3s, nobody will really care as long as their departed loved one feels like the same person.
The "singularity" is just a mathematical abstraction, as with dark energy or string theory. The math is highly likely to describe a real phenomenon, but anyone extrapolating further detail based on that math is basically just guessing.
That is actually a very strong assumption.
AI are created by building immortal human brain simulations, raising them, then cloning the results. Humans are not able to upload and eventually go extinct.
> I'm not saying it's impossible. I'm saying it's so far from likely that being near-immortal due to bio-enhancements is almost certainly more practical. Or you can develop some "scanning ray gun" which does the same thing without killing you first, I suppose.
Seriously, I see near-immortality due to bio-engineering to be way more likely in the remote future (next few hundred years) than uploading your brain. Similarly, I see strong, general AI to be way more likely.
I'll even say that I see full, human brain simulations to be way more likely than the ability to upload your own into the net.
I'm all for being hopeful and optimistic, but sometimes physics is just not on your side.