Computer information thread (Starglider stuff)

[Redleader34]

Now I was thinking of writing a short about a basic transhuman style experiment, or someone putting their brain in a massive computer because their present body is dying due to GENERIC SICKNESS, most likely leukemia or some other end stage cancer, and I realized, I have no idea how much information would be in a human brain and body. Would it be in petabytes, zetabytes, and how would a way of scanning the body even work. More or less it was a question of how much "information" is in a human, and how would you get it from meatspace to the computer world (If this is more sci fi I'm sorry), I just had a thought while I was typing up this basic short.

[Starglider]

I have no idea how much information would be in a human brain and body. Would it be in petabytes, zetabytes

A detailed description of human brain structure would be somewhere around 1 exabyte, with a reasonably compact encoding. A functional model sufficient to accurately reproduce their intelligence might be anywhere from two to four orders of magnitude less, depending on exactly how much exact synapse configuration and neuron structure affects overall intelligence (this is still an open research question).

A human genome is 825 megabytes uncompressed (it crunches a lot under lossless compression though, lots of duplication). A full body 3D description depends on exactly how close you need it. If you just want to make a clone body that looks the same, the genome plus a coarse voxel model (a few gigabytes) is sufficient. If you want to make a replica accurate to a molecular level (I don't know why you'd bother), you are talking about yottabytes of data even with aggressive compression.

and how would a way of scanning the body even work.

You can either use very high resolution NMR/CAT techniques, send in swarms of microbots to physically map things, or freeze the body and destructively scan it with a laser vaporisation slice scanner. The last one gets the highest resolution at any given technology level but has the disadvantage of destroying the sample.

[Serafina]

Actually, IIRC, the size of the genome that is unique to an individual human bein is about 10 MB.
The rest is virtually identical in each human being.

[Redleader34]

The starglider signal worked, yay!

Ok, now the short more or less has it that the character's body is going to die in a year tops, so could you use nanobots via an injection to scan the person, or would that be impossible..


Now for the next question, for a virtual world say the size of London, with basic interactivity (All the buildings have a inside, outside, basements, subway exists, train lines go to and from the "EDGE OF THE WORLD", what space requirements would need in terms of computing space.

For the computing medium I'm thinking of, I have a small building out in the boondocks turned into a virtual world, assuming more's law, could this be done by 2060, or am I being too optimistic.

[Starglider]

Ok, now the short more or less has it that the character's body is going to die in a year tops, so could you use nanobots via an injection to scan the person, or would that be impossible.

Given sufficiently advanced nanotechnology, sure. However if they're going to die anyway it would be easier to freeze them immediately after death, then scan via laser ablation / slicing. Also post-death scanning has a marginally lower chance of setting off the 'continuity flaw' whiners.

Now for the next question, for a virtual world say the size of London, with basic interactivity (All the buildings have a inside, outside, basements, subway exists, train lines go to and from the "EDGE OF THE WORLD", what space requirements would need in terms of computing space.

A terabyte should suffice, using fractal compression/smart interpolation/etc for indefinite zoom on textures. Less if you cheat and replicate content, more if you want really high-definition models (or voxel modelling).

[Redleader34]

Oh wow, I thought it would require huge data, fitting it all in a TB means you could run this on a computer 10 years from now, this works better than I thought.

How hard would altering a persona, or the person's shape be in the world, would it be something that you could say do just by thinking, or would this be a GUI thing, sorry if these questions seem small, I'm trying to make the background work with somewhat realistic technology, even if it is speculative. Also, could you in theory have a low level AI work as an antivirus, to solve the issues with modern ones?

[Starglider]

Oh wow, I thought it would require huge data, fitting it all in a TB means you could run this on a computer 10 years from now

You can fit a detailed model of a city onto a current hard drive with room to spare; 'digital city' projects have been going since the mid 90s to do just that. Of course those projects don't usually attempt to model the interiors of anything except major public buildings. Having one avatar wandering around a static/scripted city viewing things at resolutions near the human perceptual threshold is within the capability of a contemporary high-end PC, though it will still look artificial, we don't quite have the power for real-time ray-tracing etc yet (give it another few years). Having multiple avatars or AI entities 'populating' the city is obviously a much more compute intensive proposition.

How hard would altering a persona, or the person's shape be in the world, would it be something that you could say do just by thinking

No. At least, not by default. Adding that feature would be a very difficult mental command decoding problem.

or would this be a GUI thing

Yes, see: any current MMORPG.

Also, could you in theory have a low level AI work as an antivirus, to solve the issues with modern ones?

Amusingly, when antivirus software was first developed, it was considered an example of applied AI. It isn't any more of course; any piece of 'AI' that is commercialised is soon considered boring old non-AI tech. But yes, people are constantly trying to make smarter virus checkers that use more heuristics and modelling, and less brute-force pattern matching.