Brains

I am reading Fall by Neal Stephenson. I'm about half way through. It's kind of interesting but not exactly gripping. The story is about a rich guy who dies and has his brain digitized and uploaded into a giant computer system. My first problem with the story is that they start up a simulation process without connecting any input or output devices, so this 'brain' is just running around loose in this virtual space. It eventually starts creating a world from its memories. Now everything in this world needs to be simulated. I suppose this is reasonable, after all your brain creates a complete environment when you are dreaming, and it holds a model of the real world when you are awake. It's reasonable in that if you could create a digital model of a human brain, that digital model should be able to create a digital simulation of the world. But without any communication with the outside world, you would have no idea what is going on in there. In the book, they are relying on traffic analysis to try and paint a picture of what is going on, but it's kind of like trying to follow politics by looking a satellite image of the Earth using Google Maps. (Yes, the Democrats have set fire to California and cut off the power to the Northern half of the state, which must be populated by Republicans.)

Communicating with digitized brain is going to problematic. Shoot, digitizing a brain is going to problematic. I got the impression that the scan in the story makes a map of all the inter-neuron connections (a 'connectome'). This might be a good way to model a brain, IF you had a good model for a neuron. But neurons are complicated. The standard model says that neurons communicate with electrical signals, but recently someone discovered that they also send bits of RNA to other neurons which can then incorporate that into their structure, kind of like adding an app to your smartphone. It can add a new capability, modify an current one or completely change that neuron's personality. So if you really wanted to digitize a brain, you could not get by just by making a connectome, you would have to scan every stinking atom. And that wouldn't be enough either, because the brain you are scanning is dead, so you would need to clean it up, reverse all the decomposition and any damage done by freezing that was done to delay the decomposition, which is going to be very tricky because once complicated molecules break apart, they don't carry any markers that tell you where the pieces came from. Like if you take a giant Lego castle and smash it. Can you tell what the structure was by looking at the pile of pieces?

Communicating with a digitized brain, assuming you can do that, is not going to be easy either. You would have to model the signals that the various sense organs generate, construct analogous devices and wire them into the connectome. Likewise for output, you would need to model the signals coming out of brain and translate those into something could perform an action. Making a manikin walk around might be easier than making it talk.

In the book, the overall approach to this business is to just keep adding more processors and more memory. Since the people who are getting their brains digitized are all billionaires, they can afford lots of processors. There are like 3 million data centers in the US. Whether that would be enough to simulate a digitzed brain is open to debate. Billionaires could be building every larger data centers, kind of like the Pharoahs built pyramids. Not because they are good or useful, but because they can.