Showing posts with label mind. Show all posts
Showing posts with label mind. Show all posts

Wednesday, April 24, 2013

Lemme gloat for a second

I recently started taking Coursera's Computational Neuroscience class and I wanted to point out how close I was to how brains actually work in this old post without actually knowing anything.

The one thing I got wrong was that dendrite configurations are important when in fact it is permanent changes in the neurotransmitter receptors that result in learning and memory.

Also interesting and equally incoherent is this post.

Sunday, October 18, 2009

Memory and Evolution

Expanding on the previous post...
I forgot to reference the TED talk that spawned that whole train of thought (aka pipeline):
Henry Markram and his Blue Brain

Anyways... moving on...

The ever increasing complexity in the Universe seems to stem from evolutionary systems. Clearly not just biological evolution... by evolutionary system I mean a system that has 1) a means of storing information, 2) a means of mutating that information and 3) a set of rules that determines which information is lost and which information is retained. In the context of biological evolution, you have genes that encode information, sex (for the most part) that mutates that information, and natural selection to determine which information is retained or lost.

At a lower level, you have molecules that store information with bonds (kind of trivial I know but bear with me now,) this information is mutated by interaction with other molecules, and the laws of physics govern whether bonds are retained. I'm not very familiar with chemistry, but from what I understand, there are various forces that determine whether bonds are retained. In covalent bonds, the attraction of the electrons of two atoms to the nuclei of both atoms holds the nuclei in equilibrium, overcoming the repulsion between nuclei and between the electron clouds. This type of bond clearly only works for certain atoms and some bonds are better retained, allowing larger structures to be built up, like carbohydrates, amino acids, etc. Some information survives... like the bond structure of a protein, while other information is lost or never created... like the formula for awkward pills (UAr7C2Si14.)

At a higher level, we can look at human culture. Information in this system is stored by brains in the short term, but is transferred (at great risk of mutation) and "stored" in the long term by language. As societies have progressed, we've invented less lossy ways of transferring information (books, for the most part, and more recently a series of tubes.) Certain information survives, largely determined by what's interesting and useful. Physics is very useful, so it survives. Phrenology was not useful. One problem human society faces today is that the cost of storing information is so cheap that there is no longer any way of separating the chronic from the schwag.

I am very interested to see what the solution to this information overload is. It could very well be the jump in abstraction level from one evolutionary system to the other. Or, more likely, human beings will be unable to handle this quantity of data and will have to bow out to the next apex predator: SkyNet. *cue Terminator foot crushing skull and sinister music.*

Long story short: what the universe is about is not the survival of matter or energy. All that is fixed and will be dispersed at some point in the distant future. The universe is a big game of survival of the fittest information.

So here's how the brain works

Let's take the reality around us and represent it as a stream of data.

Your sensory cortices read from that stream. Well, read's sort of the wrong word because what's actually happening is that certain patterns in that stream will set off electrical cascades (let's call them pipelines because I like pipes) in your brain. These pipelines propagate through certain paths determined by neuron configuration and they can trigger other pipelines. More importantly, they can change neuron configuration and thus how pipelines will propagate in the future. In essence, the brain can reprogram itself. Well, duh. That's how memories work.

Let's talk a bit more about pipelines. Neurons are your analog bits. So err... not bits at all. They have a function (the action potential function I think) that determines how input impulse maps to output impulse and the properties of this function depend on the type of neuron. Typically, it's pretty spiky (hence the all-or-none principle of neuro-transmission.) Neurons connect with shitloads of other neurons via dendrites.

A pipeline is the macro view of neurotransmission A neuron will set off an entire cascade of electrical signals in some cases, but do very little in other cases. These cascades/pipelines can be very complex and are definitely not independent of one another.

The action potential functions can be changed temporarily by neurotransmitters like seratonin, and possibly permanently (though this is completely unverified.) Permanent change at the neuron level isn't really necessary for this theory though. What's more important is the dendrite configurations. Dendrite connections between neurons need to strengthen for commonly called pipelines and weaken for rarely called pipelines. This is how memory works. A common misconception is that we're actually retrieving stored data like a computer (so is computer memory a retroactive misnomer?) What's actually happening is that some pipeline set off a pipeline that we associate with the idea memory, because whatever things this memory pipeline calls up happened in the past.

What's also apparent is that there is variability in brain design due to the importance of certain variables, some of which we know: the plasticity of connections between neurons, how signals are changed by chemicals, and the action potential function of constituent neurons.

Through evolution, you can see the trends in brain design. The first brains were simple affairs, not very plastic, mostly there to coordinate motor functions and what not. These brains were for the most part static programs that could group certain high-level concepts into an actionable idea. For example: "predator" is associated with "run the fuck away." It's kind of like a hash table that maps stimuli to reactions. Our brainstem is very similar to this oldest brain design.

After handling our instincts, we needed something so that we can change our reactions and we needed a mechanism to control how we change our reactions. So we get emotions (not just talking about feelings... I'm referring to the physiological reaction,) and the limbic system which is basically a big analog switch statement on infinite loop. From my rather limited understanding of neurophysiology, I gleam (synonym for wiki'ed) that emotions are determined by then levels of seratonin, dopamine, and adrenaline. These are controls of your emotional reactions. Different levels will cause you to call different pipelines. The limbic system is somewhat malleable, and apparently developed because mammals needed to raise their kids.

The last stage in brain evolution is the neocortex, which is very, very malleable. This is the source of the majority of your thinking. It's basically designed as a playground of plasticity. You have shitloads of the most plastic nerves imaginable packed in as densely as possible. The neocortex is such a leap forward from the limbic system that I have no possible analog that I can draw from computers. It is so advanced that the evolutionary prerogative for the past 5 million years has been... MOAR.

Anyways, that's what I've been thinking about this weekend.

Saturday, February 14, 2009

Manic Oppressive

From PVCPVCCPVCCCP

The bad news: I'm bipolar.
The good news: Everyone is bipolar.

Let me explain... I guess I don't mean bipolar like the actual disorder. I'm not saying that people who have bipolar disorder don't have a problem. I just mean that I, like everyone else, exhibit the symptoms of bipolar disorder at times. I think that these symptoms explain a great deal about how humans think and process information.

So here's my theory of bipolar disorder:
Being a functional human requires the ability to process a great deal of information very quickly. Sight, sound, touch, chemical responses... etc etc. The only way to do this is shortcuts.

Your brain generates models of the outside world based on the data you collect. It uses those models to interpret new data and uses that new data to revise the model. Yes... revise, not recreate. The problem with this is that your brain also uses those models to choose what data to collect. Yep... you guessed it... feedback loop fail.

Bipolar disorder is the manifestation of the turbulence inherent in poorly controlled feedback systems. You have your world-view. You're supposed to be collecting a good sample of data to revise that world-view so that it changes smoothly. Well... too bad you don't and when some contradictory data slips in that crushes your world-view, you get utterly pwned. Now you have to recreate your world-view in a very short amount of time, based on the limited data you can collect at the time, which will probably be the negative stuff.

The severity of bipolar disorder varies greatly from person to person. I read a bit about it on Wikipedia and found this interesting tidbit. Basically, creative people tend to have problems with bipolar disorder. Correlation does not equal causation. But it is a big fucking hint.

So in my model of model generation... what could possibly make your turbulence more severe? Well whenever your models systemically diverge from reality (i.e. it is a biased estimator,) you're in deep shit. Poor data sampling gets you data that is not necessarily representative of reality. Aggressive extrapolation in model generating leads to models that don't necessarily fit the data. Both of these sound like problems creative people have.

To kick it old-school MBTI stylie... it appears that extroverted intuitives (ENTP, INTP, ENFP, INFP) are the most likely to have bipolar disorder. I would bet a zillion bucks that sensers don't have much of problem with bipolar disorder.

These issues with poorly controlled feedback seem to apply to a lot of complex systems. Economies suffer from business cycles. Extinctions tend to happen as mass extinctions. etc. etc.

Just an interesting connection... not to be taken seriously until I publish :-)