Information Space and Memory

(In progress.  Speculative.)

Memory is kind of spooky.

It’s really weird to think about the idea that memories aren’t physically stored in the brain.  If you really start digging into neural networking, it starts to make a little sense after a while.  When something as complex as the brain starts making sense to me, I start to worry.  If experience is any guide, that means I’m fixing to find out how complicated it really is.

The neural network is based on the brain cell, or neuron.  With the idea being, that the neural network is made up of billions of neurons that are massively interconnected, and massively parallel.  We are pretty sure that’s right.  Anatomical studies have shown that intelligence probably correlates with the number of connections, and we have even identified a gene that regulates the number of connections, which in humans uniquely allows for a very large number of connections.

Let’s look at the neuron.

I think it’s helpful to think of the neuron as being like a little electrical gun.  Pull the trigger and it fires.  Only unlike a real gun, it doesn’t have a trigger; it has a switch.  Because it’s electric.  And it shoots out an electrical impulse instead of a bullet.  So let’s say, we are going to flip a switch.  When we do, that will trigger an electrical impulse down the nerve.  If we are talking about the nerve that connects the brain to your little finger, that’ll make the finger twitch.  If we are talking about brain cells, then the electrical impulse is going to interact with a lot of other electrical impulses to gin up a memory.

But, it’s a tad more complicated.  Let’s keep in mind the concept of electricity and switches.  Suppose you’ve about had it with the lousy lighting in your living room.  You call out an electrician to re-wire it.  After a full day of working on it, he proudly announces he is done.  You go out into the living room, and sure enough you can see now.  But you notice that the electrician has wired in four switches.  What gives?  Well, he calls this the “semi-democratic light switch.”  He noticed that you are a family of four, and so he gave every family member their own switch.  He wired the light so that it goes on as long as two switches are in the “on” position.  But Mom and Dad get partial veto power. They have special switches; each one has three positions.  In the middle position, the switch is off.  Up is on.  Down actually cancels out a switch.  So if the two kids turn their switches on, and the two parents put theirs in the “cancel” position, the light stays off.  But Dad can get outvoted by the rest of the family.  If there are three “on”, and one “cancel,” that nets out to two “on” and the light goes on.

If you get that, then you get the neuron.  The switch end of the neuron actually might have dozens of switches, representing input from other neurons upstream.  Most of those switches are of the normal on/off variety.  Some are “cancel” switches.   The neuron needs a certain number of switches to be in the “on” position, net of those that are set on “cancel,” to fire.

Now, imagine those are all dimmer switches, and you got yourself a neuron.

At this point, you start feeling pretty good about the neural network.  Theoretically you could take the brain apart, see how each neuron is hooked up, see how the switches are configured, and you should be able to figure out how the brain stores memories.  I mean, it’s really just simple algebra, right?

Well, no.

The output of a neural network can be modeled mathematically, but it’s not simply a matter of adding up which switches have been pulled.   The network will settle into a steady state once all the neurons are finished doing their thing, but how it gets into that state is a head-scratcher.  It’s helpful to think of the output side of a neural network as being sort of an energy landscape.

Did you see “The Longest Day”?  Remember how they made the maps of the Normandy beaches in a sandbox?  They molded the hills and valleys in sand, in 3D, so the troops couldARSANDBOX1-660x373 just look at it, instead of having to squint and figure out a topographical map.   Question: how would you find the lowest point on that sandbox map?  Here’s one way.  Sprinkle water on it.  Water flows downhill and with a little luck it’ll settle in the lowest part.  That’s pretty much how we think the neural network works, except instead of water we are looking at the flow of energy, which is seeking it’s lowest level.  That “lowest energy state” defines a memory.

How the heck do you model that?  What kind of math do you need?

I don’t know, but I’ll tell you this.  The problem can’t be solved on a one-dimensional number line.  Give your mathematician as many dimensions as she wants, and she can take a stab at it.  She’ll model the energy landscape as an n-dimensional hypercube, good to go.  Just don’t ask her about those extra dimensions.  “Are you saying those dimensions actually exist in reality?” “Don’t ask me, I’m just doing the math.”

Even so.  Kind of weird, isn’t it, that something we do every day — thinking — is a process that is carried out in a multi-dimensional information space.  Yet the only thing we can see is a three-dimensional glob of brain cells.


We think of the mind as being an emergent property of the brain.  Seems to me, we could just as easily think of the brain as a projection of a multi-dimensional mind onto a three-dimensional space.

To be clear, any time we talk about projecting an object onto a brane that has matter and energy attached, we aren’t talking about an object made of mass or energy.  Those things are stuck to the brane.  I think we need to be thinking in terms of information that interacts with, or manifests in reality.

To give an example, gravity is a form of information that projects from Motown into the brane.  (It’s not really called “Motown,” it’s called “The 5th Dimension” or “The Bulk.”  But when I think “5th Dimension,” I think “Motown.”)   I think we can fairly call gravity information; it is the influence of gravity that forms star nurseries and planetary systems.  Gravity made our home.  There’s information in there somewhere.

To give another example, one solution to the problem of aperiodic crystals, one way to understand their structure, is as a projection of higher-dimensional lattice structure (information) into three-dimensional space.

So my question then, is this.  I wonder if the brain is the same sort of thing?  That the brain, along with its propulsion systems and nutrient-assimilation systems (which is the neurologist’s attitude toward the so-called “body”), is the result of projection of higher-dimensional information space into the brane?  (I can’t decide if I hate or love the term “brane” when I’m talking about “brain.”  I think I love it.)

Who’s right?  I don’t know.  The simplest answer is probably the best,and my answer isn’t the simplest until the neuroscientists say, “Wait a minute… you can’t get there from here.”  If that ever happens, remember where you read it.

Leave a Reply