Your Brain


How a 1.5 kg bundle of fat and wires turns life into a movie


Posted 05 Feb 2017    Edited 05 Feb 2017


If your brain got expertly cut out of your skull, it would look like this for a few hours:


Wired article by Jonah Lehrer

An exposed brain decomposes fast, though - way faster than most other human tissues. After a couple hours, your brain would be a mushy pile of fat and pretty much good for nothing except to be eaten by microbes or to be used as lube on a bicycle chain.

It's like one-and-a-half kilograms -> that's about two percent of an adult's body weight; but, it accounts for basically every psychic aspect of who a person is.

When you're like, "John's a good guy. I like him," you're saying that because of the way John's brain is.

Actually...

You're more likely saying that because of the way your brain is. The whole world that we think is around us is really just in our brain. It's like a crazy, super complicated lens through which we experience the universe [a universe which could totally and completely be a construct of the brain itself].

Any brain is amazing, but the physical human mind is particularly itself-blowing.

Every cool invented thing - including [and especially] airplanes - exists because there was a human brain to be there, pull nutrients from the blood, regulate itself, gather information from non-brain things in the body, fuel action potentials, do a bunch of black-boxy magic that we don't really understand to process the sensory data, plan, and spit that information out through the body in ways that made things happen that transcended the mere survival of itself + its host organism [you - me - he/she].

Why There Is A Brain Even


Because evolution.

I actually wrote a pretty good [if I do say so myself] summary of evolution in the first part of my 2001: A Space Odyssey entry - basically emphasizing the fact that life sprang out of not-life.

The earliest living things were bacteria-type thingies. They had cellular activity and metabolism, but no brain. Even more complex single-celled organisms and even multi-celled complex organisms evolved later, still lacking brains. Some electrical signal stuff started happening [even in single-celled guys], but brains didn't emerge until we had worms.

Worms have brains; brains that suck compared to ours, but that are still pretty cool in their own right. There's this one specific worm called Caenorhabditis Elegans. Because of its simplicity of body and mind, it is the most-studied, most-documented, most-understood organism in the world [from a neuroscientific perspective].

Its entire brain is mapped. It consists of only 302 neurons [brain cells], but a complete model of how they're all connected [called a connectome] has been compiled and is available for independent study.

There's a project to simulate this connectome [along with all the rest of the C. Elegans's cells] called OpenWorm

OpenWorm Mission Statement


OpenWorm aims to build the first comprehensive computational model of the Caenorhabditis Elegans (C. Elegans), a microscopic roundworm. With only a thousand cells, it solves basic problems such as feeding, mate-finding and predator avoidance. Despite being extremely well studied in biology, this organism still eludes a deep, principled understanding of its biology.


A wireframe image of the OpenWorm connectome from a 2012 Scientific American article

I've been following the openworm story pretty closely for about two years. Even though the organism being studied and computationally represented is simple, the project is ambitious. The folks at OpenWorm are seeking to fully model an entire organism computationally. And once a simple roundworm is modeled, the next step [for somebody] will be to do the same for a more complex organism.

But, like I said, a worm brain is incredibly simple, as far as brains go. Compare the C. Elegans's 302 neurons against a human brain - which has like eighty-six billion of them.

Perhaps a better measurement of a brain's complexity, though, is not in the number of neurons, but in the number of synapses which are the connections between neurons. The C. Elegans has seven thousand of these. The human brain has well over a hundred trillion!

The emergence of worms probably happened in the five-hundred-to-six-hundred-million-years-ago range. That's how long it took for this

Image credit: Scientific American

To become this


Image credit: MRI Master

And it hit every stop in between.


This is a chart used in a study by Suzana Herculano-Houzel who is a Brazilian neuroscientist. She studies brain evolution and stuff. Probably you should open this image in a new tab or something to see it better

Suzana Herculano-Houzel, the scientist referenced in the image caption above, shares an amazing Ted talk on what makes the human brain special [and what makes it perfectly ordinary]. Her big thing is metabolism. She argues that advanced primate cooking practices are really what led to the emergence of the human brain; that our intake of energy is facilitated so greatly by cooking that we can afford to have little bodies with relatively big and energy-hungry brains.

The human brain, she says, while accounting for only two percent of human body weight, consumes an astounding twenty-five percent of the body's energy allowance. She says something like, "the human brain is freaking the most amazing freaking thing any of you have ever heard of."

I wouldn't hesitate to declare that you and I are vastly superior to any other species on earth, especially worms; and that our brains are amazing compared to theirs. But we gotta admit that our brains are gonna suck compared to what's coming - whether it emerges organically from us, or artificially from an algorithm we've developed [I say "whether" like there's a possibility for significant organic evolution to compete with artificial evolution. There isn't. It will be from an algorithm].

I'm not a crying person, but learning about and thinking about the evolutionary development of the brain makes me feel like I almost might.

What A Brain Is Even


It's like a bunch of fat with real thin electrical cords running all over inside it. And it's foldy.

I guess, to start with the big picture, the brain has four or five main parts:
  • There's the cerebrum. It is the biggest part of the brain and it's situated on top of everything. When we imagine what a brain is, we generally [probably] are visualizing the cerebrum. It's further divided into several lobes: the frontal lobe is the most famous and is most strongly associated with abstract thinking, planning, and decision making. The other lobes do a lot of other stuff, but aren't as cool as the frontal lobe
  • There's the cerebellum, which is latin for "brain jr" because it kinda looks like a brain within the brain. It's below the cerebrum and sits in the back of the skull. The cerebellum takes up only a tenth of total brain volume, but it is home to about eighty percent of the brain's neurons. It is responsible for processing motion data and controlling muscles and stuff throughout the body
  • There's the brainstem. It is the gateway for brain stuff to get down into the body and for body stuff to get up into the brain. When you're like, "Ow! Shit!" when you walk into the corner of a table, you can thank your awesome brainstem for carrying those signals from your legs to your brain. Or you can be mad at your stupid brainstem for carrying the stupid signals down to your legs that made them walk into the stupid table in the first place. The brainstem is our brain's structure that most closely resembles the brains of our evolutionary ancestors. The parts of human nature that make us like animals [instincts to eat, be aggressive, and lie down with the ladies] are all strongly associated with areas in and around the brainstem

Stepping down to a smaller scale, there's a lot going on inside of these main sections that is highly specified and really hard to fully understand, but there are certain regions of the brain that are well-mapped. Neurologist have a couple tricks for finding out what a certain part of the brain does.

One way that is kinda scary is direct stimulation of brain tissue. These guys did an experiment where they electrically stimulated the brains of some people with OCD to help alleviate their symptoms. They sent different levels of voltage into an area of the brain called the anterior limb of the internal capsule and the nucleus accumbens region. Probably not many people outside of the study even know wtf that is, but they observed that patients would smile when shocked with mild voltage and they would straight up laugh out loud when shocked with higher voltages.

This type of testing requires risky surgery, though. I wouldn't do it.

One kinda easy way to learn about brain regions is to just study brain damage documentation. If you read twenty papers about subjects who have had the same tiny part of their brain damaged or surgically removed or something, you can look at the symptoms they exhibit and you can pretty confidently say, "Well, none of these people can recognize color anymore, so I guess this part of the brain helps us do color stuff" or whatever.

Some modern techniques are less invasive like EEGs. They measure magnetic field changes that arise from electrical activity in the brain. EEGs have pretty high accuracy in determining where activity is happening in the brain, but they are really limited in capturing stuff that doesn't occur on or very close to the brain's outer surface.

There are techniques for seeing changes in blood flow in the brain [which corresponds to activity], but they don't provide a ton of precision.

Snapshot of an fMRI [from BYU]. FMRIs use magnetic resonance imaging focused specifically on oxygen signatures; there's no radiation and it's non-invasive.

Plus, even if we could see what's going on in every region of the mind, we're still just scratching the surface of how things work in there. Cause the functional mechanisms of the brain are teeny tiny: it's all about the neurons and synapses, bruh

Little-bitty Brain Bits


All nervous tissue [not just what makes up the brain, but also all the nerves throughout the body] is made of neurons and glial cells. Glial cells don't transport signals like neurons do. Instead, they provide protective and sustaining functions to the neurons. About half the volume of the brain and spinal cord is made outta glial cells.

They contain tons of fat in terms of mass percentage. Fat is squishy and insulates the functional nerve cells from damage. Fat is also an electric insulator, which is really important, because the neurons transmit information to and from the brain and throughout the nervous system using electrochemical signals.

The neurons are the real players in the nervous system. When explaining them, instructors typically use illustrated images, because actual microscopic neural images are kinda a mess.


Image credit: Paul De Koninck, Universite Laval

The image above shows neurons [stained green] and glial cells [red]. The width of the image represents about nine one-hundredths of a millimeter. You can see that the neurons have large central cell bodies. These are called soma. They contain the typical cell elements that are found in other cells types in the body [like a nucleus, mitochondria, Golgi apparati, etc] that keep the cell working and healthy.

The branching structures coming off the soma like hairs are called dendrites. They increase the surface area of the cell and extend its reach to connect with other neurons. Sometimes, one of the long hairs reaches all the way to another neuron. These branches are called axons. They're basically electrical wires that carry electrochemical signals from one cell to another, and they're usually insulated in a covering of myelin which helps their signals propagate more quickly down the shaft.

The point of contact between an axon and another cell's dendrite or soma is called a synapse. A synapse is a connection between neurons, but there's actually a small gap between one cell and the other. This synaptic gap has chemical messengers that "wait" on one side for a signal to cross over to the other side. The receiving side has receptors that listen for the chemical messengers to cross. Once they do, some action is triggered in the receiving cell. The action is usually to send its own signal on to another cell.

This process of signal propagation down a cell's membrane is referred to as an action potential and it is one of the most fundamental concepts in neuroscience. Like, if you wanna talk about neurology at all, you gotta have an understanding of action potentials. And once you do, basically all neurological disorders and most consciousness-altering drug symptoms and lots of other things about the nervous system are going to make a lot more sense.

But it's a too-long conversation to have here. So just follow the link and skim the Wikipedia article if you can. If your brain wants to...

Here's a pretty cool cartoon from that article that shows how all that shit fits [including a zoomed feature on a synapse]:


Diagram of a typical myelinated vertebrate motor neuron by LadyofHats via Wikimedia Commons

So that's kinda like the basic stuff, I guess... You know? about what the brain is.

Myriad Enigmata


The humbling thing about all this is that it doesn't make any sense.

We can totally look at brain cells under a microscope and we can see the different structures and we can scan people's brains while they listen to classical music and we can even cut a lady's head in half and shock different parts of her brain with electrodes and watch her laugh and cry and fart; but we're still a long way away from really getting it.

Even if we had a full connectome of the human brain [like the one we have for the C. Elegans], it doesn't mean we'd understand it at all. It'd just be a mess of wires in a head, not giving us any insight into how memories are stored or where consciousness materializes or how sensory data is processed or anything like that.

We don't even know what a memory is, from a physical standpoint. Like, in a computer, data is stored magnetically as a 1 or a 0.

But the brain isn't like that, as far as we know. Memories seem to have more to do with the strength of pathways in the brain, than with some fixed location that stores letters or numbers or anything like that.

There was a scientist in the fifties who proposed a radical hypothesis about memories: that they may not be located in the brain at all.



James V McConnell, a University of Michigan professor and researcher, hypothesized that memories were stored in some way outside the brain. He did these experiments with flatworms. Basically, he'd train a flatworm to do something well enough that he could prove that it had memorized something. Then, he'd cut it in half and [because it was a flatworm] it would grow a new head with a new brain. The "new" flatworm was observed doing the same tasks it had been trained to do before, even though its original brain had been removed.

The scientific community was [and still is] very sketched out by McConnell's experiments, and they basically reject his findings. A new guy, though, named Michael Levin, is resurrecting the McConnell experiments, seeking to avoid some of the gaps that caused his work to come under such heavy scrutiny. And his results, so far, look similar: flatworms with their brains removed can still remember memorized information.

In this really sweet Verge article about McConnell and Levin, it says:

It’s ... possible that the flatworm’s unique regenerative abilities lie behind its ability to recall memories after growing a new brain. In that case, flatworms may be the only species with a central nervous system that can store memories outside of the brain. "I find that highly unlikely," Levin says, "but it’s a possibility."

Even if storing memory outside of the brain is universal among animals, that storage method might only work for simple pieces of information. Complex memories like the significance of the word "truth" or "caring" might not have a place beyond the brain. But if there is a small chance that the experiment is reproducible, and that this isn’t a trait reserved to some tiny insignificant worm, the impact could be revolutionary.

This section [up to this point] took me like five hours to write because I got lost in a Wikipedia rabbit hole. If you wanna go on an interesting link-journey through Wikipedia, might I suggest: James V McConnell > Ted Kaczynski > Project MKUltra

Beyond memory, another phenomenon about the mind that freaking boggles it is consciousness.

Consciousness is kinda like memory - we only know it exists by subjective, personal experience. We're like, "Well, this movie-like experience of my life is happening. Like, I'm watching it right now, so it must be a thing."

But when we crack into the brain and look for the place where consciousness lives, it's nowhere to be found. Why do we have consciousness at all? Obviously, having memories aids an organism's survival. With the magic of memory, she can learn from her experience, solve problems in the present, and make accurate predictions about the future. But, why does she need to consciously be alive? What biological advantage [if any] results from experiencing existence?

No one really knows. As far as we can tell, the mechanisms of the body and mind could function without our "knowing" it. Luckily they don't.

Consciousness is not [really] in neurological territory, yet. There's this article from a 1999 edition of Neurological Review that attempts to make the connection between neurology and consciousness, but it just makes correlations between aspects of consciousness and brain regions. I'll admit I just kinda skimmed it. But it really doesn't explain what consciousness is or anything too satisfying.

Moral of the story: the brain, at this point, is basically incomprehensible. We don't know what memories are. We don't know what thoughts are. We don't know what consciousness is.

Moral of the story: we don't really know what the brain is.

It's possible, though, that even with an extremely incomplete understanding of the brain, we could create something that could very quickly evolve an intelligence beyond that of the human mind.

Learning: An Algorithm


Traditional computer functions written by people are explicit: meaning that you tell the computer every thing to do and it does it, step-by-step.

My first computer science professor in college said that you have to treat the computer like your ten-year-old Russian nephew who is visiting you in America for the first time. He said that you can't just tell your nephew to go do the laundry. He said you have to first speak to him in his language [cause he's Russian], and you have to tell him every little single thing to do [cause he's a ten-year-old].

Like, if you want to make a really stupid program that multiplies a number by two, you gotta write lines of code that ask the user for input, then you gotta tell the computer to check if the input is something that could be multiplied by two. Then, if it is, you tell the computer to multiply it by two and save the result. Then you tell the computer to display the result to the user.

A special category of algorithms have been developed that surpass the need to explicitly tell the computer what to do, and, instead, give the computer the freedom to determine its own [limited] behavior based on data from which it "learns."

With machine learning algorithms, functions can solve problems that don't require a programmer to account for every possible input scenario or to explicitly code every step of a process. Instead, the programmer can establish functions that learn from data and can alter their behavior based on inputs.

In a machine learning and data mining class that I took my last year in college, I studied these algorithms and coded some up. The first project we worked on used a categorization method called k nearest neighbors. It basically looks at a data set, learns what attributes are most closely related, and then predicts what kind of thing is being shown to it, based on given attributes.

One of the common example data sets we looked at was the iris data sheet. It's a list of one hundred fifty actual iris flowers that exist as three species: Setosa, Virginica, and Versicolor. For each flower, the petal length, petal width, leaf length, and leaf width are measured and listed for the flower. The species is also specified.

The algorithm we developed first shuffles the data up, then it cuts two-thirds of the data out. These two-thirds become the "learning" set. It takes the remaining third [the "prediction" set] and predicts what category [species, in the iris case] the given input will be based only on its attributes [lengths and widths, in the iris case].

For every entry in the prediction set, the attributes are compared against their nearest neighbors: the entries in the learning set that have attributes that most nearly match the value of the entry in the prediction set. The dominant category in the neighbors is assigned as a predicted value for the prediction entry.

The function can then look at the actual value of the entry and see if it predicted the correct category. In this way, the function can calculate and display its own accuracy in predicting values for the dataset.


I coded this project up in the Python programming language. For any of you familiar with Python, here's a copy of the script I wrote. And here's a copy of the iris dataset we used. You can download the Python script even if you don't know Python just to get a feel for the simplicity of the code and what the steps are. You can open it in a text editor like Notepad.

And that is pretty much the simplest instance of machine learning that can be implemented. Modern applications of these methods are pushing the frontiers of computer science. Existing technologies are so cool that... Here's one that I love:

Some developers at Google made an algorithm to learn patterns from visual data: pictures.

Using visual data as input is soooooo much sicker than length and width entries in an Excel spreadsheet. This image algorithm is already used in Google's Reverse Image Search.

Some of the technology that went into this tool led to some innocent experimentation. Developers were like, "We've got this algorithm that can detect patterns in images. That's fine. What if we flipped the code inside-out? What if we allowed the code to write images instead of just reading them in?"

The result was Deep Dream <--seriously, you gotta read the article and look at the example images.

The code can take an image of... whatever... and manipulate its already-existing features to show patterns of other images from the code's "learning" set. Since the version of Deep Dream that is available "learned" on a bunch of images of dogs and a bunch of images of buildings, it tends to insert little freakish dogs and little Dr-Seuss-y pagoda steeples all over in the images supplied to it.

Here's me...



...and here's me after getting my face Deep Dreamed by Google robots...



Look at the [definitely not] cute little doggies all over my shirt and in my hair!

Now, you may or may not think this is appealing. I'm kinda fascinated by Deep Dreamed images. But, you have to admit it's beyond incredible - that an algorithm, with no psychic or conscious understanding of the images it is "seeing" can produce something that seems to be the work of someone that knows what a dog is - the way only people do - and can display it so horrifyingly - the way only [really creepy] people could.

Interestingly, the type of machine learning implemented in Deep Dream is a neural network: a kind of algorithm that mimics the path-strengthening behavior of organic brains.

The Brain: An Algorithm


Deep Dreaming is rad; but, perhaps not particularly useful. What makes machine learning so cool is that its potential is not limited by human imagination or knowledge.

In theory, a powerful enough optimization method could easily do anything a human could do and could [just as easily] do anything a human could never even imagine to do.

I'll let my boy Nick Bostrom take us home [for the blog to finish, this must be watched in its entirety]:


The Times They Are A-changin'

Highlight:

The potential for superintelligence kind of lies dormant in matter.

I hope that the experts Bostrom interviewed are right. I hope I get to witness the singularity. I'll have lived a life of mental superiority, and I'll get to be humbled, then, by becoming absolutely, infinitely inferior. I'll get to know how it feels to be a worm.
    




For Kevin