NOTICE: Posting schedule is irregular. I hope to get back to a regular schedule as the day-job allows.

Thursday, February 15, 2018 [Full link to blog for email clients.]

My Brain is Full
The human brain is not a computer, but yes, it really does store, represent and retrieve information!

About two years ago, Dr. Robert Epstein penned an essay for AEON entitled "The Empty Brain." []  Dr. Epstein is a research psychologist, and former editor-in-chief of Psychology Today.  In the essay, Epstein started off decrying the modern, casual attempt to equate the functions of the human brain with modern digital computers.  A computer has physical memory representation and processes information in a manner prescribed by algorithms. The human brain, in contrast, lacks physically identifiable memory structures, has no fixed algorithms, and cannot perform the functions attributed to computer "processing."

In short, his logic amounts to:  you can open up a computer and find each function physically represented.  I might quibble with that, but he has a valid point as he continues to say that the human brain has no central processor, no "clock," no memory core, and you cannot find a physical representation of any of the information that the brain utilizes in its function. So we should stop using terminology appropriate to a computer in referring to the function of the brain. The article seems to resurface in social media every 4-6 months, and every time it does so, I get called in to comment, or get drawn into a discussion of why the basic premise: "The brain is not a computer" is true, but nearly everything else in the article is at best, misinformed, and at worst, junk science.

Generally speaking, the first thing that happens in these discussions is that commenters tend to denigrate the "soft-science" profession of psychology and the specific qualifications of the author.  I like to avoid that, because psychology is a very useful tool in the neurosciences, it's just that I feel the author has insulated himself from the "hard-science" findings of neurophysiology, neurology and neurosurgery.  If Dr. Epstein had had a colleague in Neuroscience read the essay, he would likely have been told that many of his illustrations and examples of his premise are incorrect or misleading.

On the other hand, it may be that he felt that the actual neurophysiological findings were too technical for a lay audience, and he wished simply to concentrate on the central premise -- the brain is not a computer - with simple explanations.  After all, he was editor-in-chief of Psychology Today, a publication that distills psychological research and development into articles with broad appeal across both the sciences and lay audience.  Such distillation requires that an article be understandable by readers with an education well below the MD or PhD level, and much scientific rigor and accuracy can be lost in translation.

Or it may simply be the approach of psychology as a field.  As a neuroscience researcher, I am interested in the function of the brain -- human or otherwise.  Having been trained initially in physiology & pharmacology, I tend to take a very mechanistic view.  I also work with Neurologists and Neurosurgeons as part of my day-job, and their viewpoint differs from mine.  They may be more mechanism oriented in some ways (what brain area connects with another; what happens when I cut in this place; why is this area acting abnormally), and less detail oriented in others (not necessarily interested in whether this single cell is connected to another). 

Psychology, however, frequently takes a different approach, concentrating on the external evidence of brain function.  Thus, a psychologist would be much more oriented upon whether a patient can remember a string of numbers than whether those numbers actually have some form of representation in the brain.  This is actually an approach often taken by modelers who use a "Black Box" nonlinear systems approach to the brain.  For a nonlinear model, it is not necessary to map every single connection and modulation, but rather to define the input and output patterns of a given system.  From these, it is possible to calculate nonlinear mathematical solutions which can transform input to output.  It's a useful type of model when the actual detail is so utterly computational complicated that a detailed model is unworkable using current technology.  Epstein's essay clearly relies on such a conceptual approach, but ignores that fact that Neuroscientists do know much of the detail of how the brain processes and encodes information. 

I therefore propose that aside from the central premise -- that the human brain is not a computer -- many of the examples and conclusions in the essay are biased by the author's own field and approach to brain function and neglects or ignores experimental evidence to the contrary.  As for my secondary concerns regarding the validity of Dr. Epstein's essay, I will simply state that he should have consulted colleagues in the more biological/organic corners of the Neuroscience field before making his claims and formulating his conclusion.

This is actually familiar territory for me.  SF congoers who attended the LibertyCon 27 convention in Chattanooga, TN in 2014 may recall a rather heated discussion involving myself and physicist and SF author Dr. Travis S. Taylor. Guests tended to gather around the pool in the late evening, and discussion topics range widely... and wildly.  I had been asked by many people to comment on Taylor's "quantum connection" theory of the mind.  The short form of the theory is that some of the "unexplainable" phenomena of the mind (attraction, common interest, prayer/positive thinking, etc.) could be explained by a quantum connection, and further, Taylor proposed that neurofilaments and neurotubules in brain cells could serve as "antennae" to pick up this quantum signal.  Taylor's usual explanation for the phenomenon included the challenge: "Do you like beer?  Why or why not?  You can't explain that.  It's unknowable with conventional science." - my response:  "Yeah, I can pinpoint the preference right down to the specific molecules making up the taste receptors in the tongue. Anything you propose, I can explain with conventional physics and Neuroscience."  It was a fun and amicable argument, and we actually concluded that I think his idea of a quantum connection is interesting and could be true, but largely untestable.  The problem simply being the selection of the wrong examples for his premise.

Which takes us back to the AEON essay:  Dr. Epstein entreats his readers to stop using a flawed analogy for the function of the human brain; thus, the human brain is not a computer.  However, in the process, Epstein uses many other flawed analogies to support his conclusion that there is neither "representation" nor "processing" of information in the brain.  The problem is that while we may not have a discrete identifiable location for memory storage -- the truth is that neuroscientists can observe the phenomena of memory processing in very real ways.  Thus the analogies used to counter a flawed analogy are themselves flawed.  The initial premise is correct (within certain bounds) but his examples and conclusions are faulty.

So let's take this apart:   First, the brain is not a computer.  We actually had this discussion in a graduate student class yesterday.  I agree.  The brain is not a computer - at least, not a digital one.  Rather, the brain is closest to an analog computer, but even that is a flawed analogy.  One of my favorite thought experiments is to consider that the brain acts as if it were a steampunk analog calculator... but that's a topic for another blog!  [] There is no central processing unit, no "memory core", no "clocks" that synchronize computation steps, and no place in the brain where one could "open it up and see the information." 

The first problem to crop up in the analogy is that one could counter that it is not possible to "see" a representation of memory in a computer, either.  Absent the tools to measure current running through semiconductors and resistors, and charge held in capacitors, it is not possible to see the representation of a picture or any other information content in a computer.  Even then, one does not simply see a color, a line or an object.  Instead, the computer stores bits--ones and zeros, represented by presence and/or absence of voltage, current, or charge--and those bits represent other information such as color, shading, presence or absence of lines or shapes.  On its own, the computer representation also not a "picture," although it can be argued that there is a discrete storage location for the information.
As for there not being a "representation" of memory information in the brain, one need only look at the evidence that the visual and auditory cortex of the brain are organized into groups of brain cells that only respond to a single visual (i.e. line, rotation, position, color) or auditory (frequency, movement, location) feature of the respective sense.  These are highly topographic mappings of information onto the structure and function of the brain.

But let's take it one step further:  My colleague Jack Gallant at UC Berkeley [] would certainly argue that the brain does indeed store and represent a recognizable pattern for pictures and visual scenes.  In 2011, Dr. Gallant published a study in which his team identified brain signals that corresponded to visual scenes in memory.  Using an MRI scanner to track blood and oxygen usage in the brain at resolutions down to cubic millimeters of brain tissue, the team created a "library" of brain activity patterns that resulted when a person looked at particular pictures.  Later, the same subjects were told to "daydream" a scene involving some or all of the pictures.  Gallant's team was able to identify a "movie" of the daydream constructed from human brain patterns.  As the subject imagined each scene in their daydream, the team compared the resulting brain patterns with the library of patterns correlated with previously viewed images.  The identified pictures were placed in sequence, and compared to the subject's report of their daydream with a very high correlation between the two.  Thus Gallant and I would argue that here is evidence that the human brain does both store and represent memory information in the brain.
Unlike a computer, the storage location is not fixed, and the "storage medium" is not standardized.  Still, there are identifiable patterns within the brain that correspond to the stored--i.e. remembered--information.  Furthermore, there is ample evidence that such information is "processed."

A particular region of the brain, the hippocampus, is involved in the processing of memory in all species of mammals, with a similar structure fulfilling the same function in reptiles and birds.  Animals and humans with damage to this part of the brain have difficulty formulating new memories, and may also have difficult retrieving memories.  In 1971, researchers John O'Keefe and Jonathan Dostrovsky noticed that certain cells in the hippocampus of rats were active only when the rat was in a particular position in the cage or test chamber [].  The finding led to nearly 50 years of research into "place cells" in the hippocampus, which correlate their activity with various elements of location.  Cells have been identified with preferences to corners, edges, head directions, body directions, and both future and past movements.  The more cells are recorded from a single subject, the more detailed a "cognitive map" of the environment can be created from the activity of these neurons.  Furthermore, the "map" may disappear or reorganize when the subject moves to a new environment, room, chamber or cage, but will reappear in the original form when returned to the original environment.  Thus, a representation of place exists in the hippocampus, and moreover, it is a memory of a representation, since it can completely disappear and be re-formed in the original format.  Place cells and place fields have been identified in mice, rats, gerbils, cats, dogs, monkeys and even humans, demonstrating that this is a function associated with the physical nature of the brain, and not just an "emergent" phenomenon of human cognition.

In 1994, Matt Wilson and Bruce McNaughton demonstrated that place cells were not only involved in a physical representation, but that they were an important component of memory [].  Utilizing a track that limited a rat's movement through the environment to specific lanes--and hence a specific sequence of activating various place cells, the team demonstrated that the same cells were activated during sleep in the same sequence as when the animal had passed through them in the test session.  By making the rat's behavior and reward dependent on remembering a previous sequence of movement, and then manipulating whether this sleep-replay could occur, Wilson, McNaughton and their colleagues were eventually able to demonstrate that the replay was an important phase of memory formation.  If the spatial information was to be remembered, it had to be replayed during sleep!

One of the major questions raised by both the scientists studying place cells, and the outsiders looking in, was that there was (evidently) no clock, no map, and no coordinate system driving the representation of spatial position in the brain.  Then in 2005, Edvard and May-Britt Moser and their teams reported that neurons in another part of the brain--the entorhinal cortex, which lies "upstream" from the hippocampus--had neurons which fired in a regular grid pattern throughout the environment [].  While not a square representation analogous to the Cartesian coordinates of a world map, the "Grid cells" nevertheless formed a triangular or hexagonal spatial mapping which could serve as the basis for the hippocampal place cells!

So, with just a few examples, we shoot down the "no representation," "no storage," "no controller" and even the "no processing" portions of the AEON articles premise.  In yet another example, of identifying computer or electronic-like functions of the brain, Dale Purves published an article in 1996 that began to lend credence to the concept that the human brain had a function analogous to a CPU clock-cycle [].  Most people have viewed a demonstration of the stroboscopic effect: if a bright light is flashed on a moving object, it can appear to be still. Photographers use it to create stop-action images that reveal the wonders of nature: the beat of a hummingbird's wing, popping a balloon, a drop of water. It is the basic mechanism of a movie film and video.  A sequence of images projected in a stroboscopic manner can create the illusion of motion.  Other illusions are also possible, as seen in the "wagon wheel effect" in which video of a moving object appears to rotate backward because the "frame rate" of the video (or the original photography) does not match the rate of rotation. 

Picture a wagon wheel with 12 spokes.  If the motion of the wheel is captured exactly at, say, 4 times the speed of revolution, the spokes will always be in the same orientation, so playback of those photos makes it appear that the wheel is standing perfectly still.  If the capture speed is slightly slower, the spokes turn a slight additional amount with each photo, and the playback looks as if the wheel is rotating forward.  If the capture speed is slightly faster than rotation, the spokes move less with each capture, and in playback, the wheel appears to be moving backward.  It's the same principle that caused video of CRT-style computer screens to have scan lines and dark bands: the screens typically refreshed at 30 or 60 times per second, but pre-HD video capture was at 29.97 times per second.

Purves research demonstrated that even in continuous light--i.e., no strobes, no frames--the human brain can sometimes perceive a "wagon-wheel effect."  To summarize a lot  of further research, the human brain acts as if it has a 10-times-per-second "frame rate."  Furthermore, we know from various Neuroscience experiments, that the mammalian brain relies on many basic "rhythms" (at roughly 4, 6, 12, 20 and 40 Hz) that can act similarly to a clock function for the brain.  They are not quite computer-like, for they are highly variable, and in fact, one of the key flexibilities of the human brain is the ability to alter certain rhythms with conscious and unconscious control.

With these examples, we've pretty much shot down the corollary statements to Dr. Epstein's premise.  But what about his example of the flawed memory of a $1 bill?  We know that there are many artists and individuals who can draw, paint, sculpt and create beautiful detail from memory.  This is perhaps the easiest of his demonstrations to shoot down.  The example shown is simply less effective memory and offers no "proof" to the lack of representation in the brain, whereas I have provided four very real examples above--including literature citations--which disprove Epstein's claims.  As stated from the start, his counter to a flawed analogy is to simply trot out more flawed analogies.

No, the brain is not a computer.  It's better! The brain is a wonderful thing filled with emergent properties and untapped potential.  Thinking of it as a computer is... limited... not flawed.  There are many features of the brain which inspire and direct our current computer technology from parallel computing, to neural networks, quantum computing and a renewed appreciation of analog systems.

But to say that the brain is not a computer, and then to follow that statement with conditions that can easily be disproven by recourse to experimental evidence outside one's own field is parochial, misguided, and misleading.  We live in a society that all too often doubts scientific professionals because of the flaws in communication.

As professionals, we must do better than that. 

No comments:

Post a Comment

Please add comment - no links, spammers will be banned.