One idea, I like to think about, is resetting the world as a plot point/twist.

enter image description here

As this is a soft-reboot that restores the planet to the last known savefile, I think I can handwave most of it with nanomachines. But there is one thing that troubles me, bringing back dead people. In order to accomplish that, I "came up" with my version of the Akashic records, a server that houses the collected brain scans of every person on the planet.

I, again, handwaved the brain scan with nanomachines, all I need to do now is to store them. But, this storage has to accomplish several things:

  • Quick access: Since we don't have a Death Note on us, we can't know when a human dies, and as the process of scanning one's brain takes time, the Akashic records update their "active section" (people who are still alive) every day.
  • Damage resistance: While the server sits in the most fortified place ever, it's still important that the data storing system doesn't fail, as lost data would be nigh-impossible to recover.
  • High information density: How much info is stored and can be stored in the brain? How much of it something that you can copy and paste between humans? I mean, why store walking animations when you can either calculate it or store it once and put links in the file? Unused storage space can be used for redundancy and when the required capacity can fluctuate this much, you'd rather not take any chances.

So, what kind of data storing technology should the Akashic records use?

  • $\begingroup$ If the question is "how much information is stored in the brain", then the answer is very little. If the question is "how much data is needed to represent the state of a brain" then the answer is a lot, but we have no idea how much is that lot; all we know is that it is a big heaping lot. If the question is "what storage technology wil be used to store brain scans when and if such brain scans become feasible", then the answer is we don't know, because we are not prophets. Honestly, I am quite certain that if there will be a need to store humongous amounts of data, a way will be found. $\endgroup$
    – AlexP
    Aug 18, 2020 at 3:56

2 Answers 2


In 3001: The Final Odyssey author Arthur C. Clarke postulated that the human brain required approximately one petabyte (250 bytes). Compared to today's technology — say, a common 2 TB hard drive — you'd need 500 hard drives to store one human brain pattern.

How much of that is repeatable data? Neurologists are quick to point out that the human brain is not (not, not, not, not, not) anything at all like a hard drive... or a CPU... or a GPU... or any other aspect of digital computing. And that's despite the fact that we don't really now how the brain stores information in the first place.

That means there could be some incredibly high inefficiencies when it comes to converting brain-patterns to digital patterns, not unlike the pain involved with using an analog-to-digital converter to store sound or video. We're going to ignore this, but it might make a cool story element.

So, for the sake of argument, let's assume some completely arbitrary, pull-it-out-of-thin-air conditions.

  1. 15% of the brain is automatic processing. This includes motor control, heart beat, thermal balance, and the thousands upon thousands of other frankly magic things the brain does without any conscious thought whatsoever.

  2. 25% of the brain is subconscious processing. This includes processing visual, auditory, tactile, and taste information. It's also our ability to walk (including "muscle memory") and otherwise move muscles without conscious regard to what, exactly, we're doing. (Anybody who's driven somewhere and realized they have no significant memory of exactly how they drove the car, made the direction decisions, etc., will know exactly what I'm talking about.) This would also include unconscious processing (aka, dreaming and whatever is actually happening between our ears when we sleep).

  3. 25% of the brain is conscious processing. This is the part of the brain that lets us pat our heads and rub our tummies all at the same time. It's what allows for acrobatics and thinking. It's what allows us to learn. This part of the brain could be where the divine gift of reason resides. Maybe.

  4. That leaves us 35% that's pattern storage. This is important. The brain doesn't store bits. It stores patterns. What that actually means is well beyond my pay grade, but it has everything to do with traditional audio/video/taste storage as well as recognizing symbols and actively remembering why you secretly liked Baywatch.

Starting with this entirely untrustworthy baseline (don't bother complaining, I'm just going to point you to this part of the answer), what we might, maybe be able to assert is...

  • The automatic processing component is 100% repeatable between humans. Oh, there's those odd little tweaks like heart arrhythmia, the tendency toward arthritis, or full blown we-love-Sherlock sociopathy that could exist — but if you think about it, those are just dials that are turned somewhat randomly in our lives.1

It's worth noting that if you have this ability at all, you also have the ability to cure or cause every known mental condition. That's interesting, don't you think?

  • The subconscious processing part is, probably (and IMO), also completely repeatable. Which means you also have control over how people perceive taste, light, sound, etc.

Where things get dicey are those last two percentages. Some of #3 might be repeatable, but #4 is 100% unique.


Remember that the brain appears to store patterns. Those patterns will, of necessity, be based on the unconscious behaviors of the brain. In other words, the information stored in Sherlock's head has as much to do with his sociopathy as it would his training and learning abilities.

Which means you might not be able to repeat #2 and #3 at all.

In fact, there's a philosophical argument to be made that if your memories are based on conscious and subconscious perceptions, then it would be based on the automatic perceptions, too. Like the memory of running flat out for 60 seconds leaving you huffing and puffing — which has a lot to do with #1.

I've just convinced myself that basically nothing of the human brain could be "repeated" or duplicated between people without damaging the core-memory that would define "I think, therefore I am."

So, you need all 500 hard drives per-person. 7.5 billion people on the planet — 3.75 trillion hard drives. And that assumes no growth. Growth is your enemy.

BTW, I'm assuming you're only keeping "snapshots" and not "incremental backups." You've got a problem if you're trying to store incremental backups.

Now that we know a reference point, what can we do about it?

You need much better storage tech than we have today. Stanford suggests you need ~35fJ to store one bit on an SSD today. That's 2.6e11 Joules or 0.26 terawatts of power. If you think about it, in 2014 the world consumed about 109,613 TWH of power. You need your terawatts running 24/7/365, so you need (am I doing the math right, folks?) about 2,300 TWH to run your server or 2% of the planet's total power consumption.

Honestly, that should make you want to pee your pants. That was ONLY the hard drives.

Conclusion: After muttering some arcane sounds and flipping through some ancient tomes you successfully invoke Clarkean Magic and do exactly what you want, the way you want it, without having to worry at all about year 2020 limitations.

Because, baby, if we use the Ginna nuclear power plant as a reference, it generated about 4.7 MWH in 2018. That means you'd need almost 500,000 nuclear power plants just to run the hard drives.

I couldn't possibly have done that math correctly. 500,000 nuclear plants? The only problem is, I have sources for the world power consumption in 2014 and the Ginna plant output in 2018. So unless I missed a factor of 1,000 someplace... somebody tell me I missed a factor of 1,000 someplace....

1I got this idea from a close friend many years ago who taught a class about autism to some teenagers. Obviously she's simplifying for her audience, but the cool thing she said was that you could consider everyone to be autistic. Autism, she explains, was something like a bunch of dials that were set too high. Too much sensitivity to light, sound, etc. All of us are, to one degree or another, "autistic" and what we define as "autism" in society is actually the worst-case circumstances of the very natural human condition. I found that explanation very humanizing and compassionate (whether it was right or wrong). Anyway, that's where the idea came from.


Theres 86 billion neurons in the brain, according to Google.

They're linked together with a weighted graph, I'm assuming its directed, so I'm just going to make some assumptions and then calculate how I'd store that information the same as any other graph. 2 dense tables of data, one for the vertices, one for the edges.

Estimating here that each neuron is linked to average 20 other neurons, each connection is weighted and directed, and the internal details of each neuron (charge, size, orientation, position variance etc) can be summarised in 20 bytes.

Neuron linkage isnt truly any-to-any, they're not going to cross the entire brain in a giant mess of pure randomness, they're going to be from a small set of nearby peers. Assume from one neuron theres about 10 million potential nearby neurons for it to link with. Even with mega handwaving I cant justify conpacting a neural link into 64 bit. Just doesnt seem to have enough data. So probably 10 bytes to describe a link (5 bytes for from index, 3 for the to index, 2 bytes for weight).

The accuracy of the relationship between any two neurons is now increments of 1:65535. That seems high enough, but if the first minds to round trip this system come out complaining that theyve got a song stuck in their head and have forgotten the last line of it, this is the number to crank up in version 2.

That's 1.6 terrabytes for the neuron information. 20 * 86 billion * 10bytes for the connections, 17.2 terabytes.

Call it 20 terabytes total uncompressed. I'd expect it to compress reasonably well, maybe 8tb.

Theoretically two human brains adjacent to each other would compress even better. They both know English. They both can walk. Etc, that's redundant. but I'd expect subtle differences in how these skills were learnt would affect their implementation, which would appear as a random neuron collection in each person, stopping this compression from working.

  • $\begingroup$ welcome to the site. If you want to improve your answer, a neuron has up to 10000 synapses that connect it to other neurons, but the average number is 1000 $\endgroup$ Aug 18, 2020 at 8:05
  • $\begingroup$ not sure but i think brain chemistry may mean there is some level of chemical state within each neuron that i don't think you account for $\endgroup$
    – jk.
    Aug 18, 2020 at 9:16

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .