Chapter 7 2.4 Scattered memory for cognitive behavior

Any thought breeds puzzling ideas. Because the human body is a collection of specialized organs—the heart pumps, the kidneys clean—it wasn't too surprising to find that the mind also delegates cognitive behavior to different regions of the brain. . In the late eighteenth century, physicians noticed a correlation between damaged brain regions and apparent loss of mental capacity in recently deceased patients just before death.The connection goes beyond academics: Is insanity inherently biological? In 1873, at the West Riding Mental Asylum in London, a skeptical young physician surgically removed small portions of the brains of two live monkeys.In one case, the right limb of the monkey was paralyzed, and in the other case, the monkey was deaf.In all other respects, both monkeys were normal.This experiment shows that the brain must be divided, even if part of it fails, the whole will not suffer catastrophe.

If the brain is divided into departments, in which department are memories stored?In what ways do complex brains share the work?The answer was unexpected. In 1888 a man who had once been fluent in speech and had a quick memory showed up in Dr. Landauert's office in a panic because he could no longer name any letter of the alphabet.While dictating a message, the bewildered man wrote every word perfectly.However, he could not read what was written.Even if you make a mistake, you can't find the wrong place.Dr. Landauert recorded: "Ask him to look at the eye chart, and he can't say a single letter. Although he claims to see clearly...he compares A to an easel, Z to a snake, and P to a Hasp."

When the man died four years later, his dyslexia had turned into complete aphasia.As expected, the autopsy revealed two lesions: an old injury in the area near the occipital lobe (vision), and a new injury, possibly near the speech center. This is strong evidence of the bureaucratization of the brain (that is, dividing it into slices).It implies that different brain regions perform different functions.If you want to speak, this department will handle the corresponding letters; if you want to write, it will be under the control of that department.To say a letter (output), you also need to apply to another place.The numbers are handled by a completely different department in another building.If you want to curse, you have to go down the hall, as the skit "Monty Python: The Flying Circus" suggests.

Early brain researcher John Hewling-Jackson told a story about one of his female patients.The patient is completely aphasic in life.Once, when a pile of rubbish dumped across the street from her ward caught fire, the patient uttered a single word—the only word Hullen-Jackson ever heard her utter— "fire!" How could this be?He felt a little unbelievable, could it be that "fire" was the only word her language center remembered?Could it be that the brain has its own "fire" part? With the further deepening of brain research, the mystery of thinking has shown people its very specific side.In the memory literature, there is a class of people who can normally distinguish between concrete nouns—say "elbow" to them, and they will point to their own elbow—but are very strangely incapable of recognizing abstract nouns— —ask them "freedom" or "talent," and they'll stare blankly and shrug.In contrast, another class of seemingly normal people loses the ability to remember concrete nouns, but can fully recognize abstract things.In his brilliant but unremarkable book The Invention of Memory, Israel Rosenfeld writes:

Ancient philosophers said that memory is a palace, and each room houses a thought.With the clinical discovery and research of very special amnesia, the number of memory rooms is explosively increasing, and it is endless.The fortress of memory, which had been divided into suites, was divided into a huge labyrinth of tiny secret rooms. In one study, four patients could identify inanimate objects (umbrellas, towels) but confused living things, including food!One of the patients could speak unequivocally about inanimate objects, but for him the definition of a spider was "a finder who works for the state." of aphasic patients.I've heard another anecdote (I can't confirm it, but I have no doubts) that patients with a certain disease can distinguish all foods except vegetables.

Borges, a famous South American writer, fabricated an encyclopedia of ancient China called Tianchao Renxue Guanglan in his novel.The taxonomy within it aptly represents the grotesqueness that lurks beneath the memory system. In that ancient encyclopedia, animals were divided into: a) belonging to the emperor, b) embalmed, c) domesticated, d) suckling piglets, e) half-human, half-fish, f) pleasing to the eye , g) dogs that leave home, h) that fall into this category, i) that twitch like crazy, j) that are innumerable, k) that are drawn with camel hair, l) that are otherwise, m) The one that just broke the vase, n) looks like a fly from a distance.

Any classification process has its logical problems, just as far-fetched as the Chinese classification method.Unless each memory can be stored in a different place, there is bound to be confusing overlap.A chattering, mischievous pig, for example, might fall into one of three of the above categories.Although it is possible to insert a thought into one of the three memory slots, it is very inefficient. How knowledge is stored in the brain has become more than an academic question as computer scientists try to create artificial intelligence.So, what does the memory architecture look like in a hive mind?

In the past, most researchers tended to think that (memory storage) is as intuitive and natural as humans manage their self-made filing cabinets: each archived file occupies a place, with multiple cross-references to each other, like a library.The idea that each memory corresponds to a single location in the brain was culminated in a series of famously brilliant experiments by the Canadian neurosurgeon Wilder Penfield active in the 1930s.Through a bold craniotomy, Penfield used electrical stimulation to probe the living cerebellum of the patients in an awake state, and asked them to tell their feelings.Patients can recall very vivid memories.The tiniest movement of electricity can trigger radically different thoughts.While scanning the surface of the cerebellum with the probe, Penfield mapped out where each memory corresponds in the brain.

His first unexpected discovery was that those past events can be replayed, just like playing a tape recorder years later-"press the replay button".Penfield used the word "flashback" when describing a hallucination following a seizure in a twenty-six-year-old woman: "The same flashback occurred several times, all connected with her cousin's home or a trip there— She hasn't been there for ten to fifteen years, but she used to go there as a child." Penfield's exploration of the virgin land of the living brain has made people form a deep-rooted impression: the brain hemispheres are like excellent recording devices, and their wonderful playback functions seem to be better than the popular phonograph.Each of our memories is precisely inscribed on its own disc, faithfully cataloged and filed by an unbiased brain, and can be played like a song on a jukebox by pressing the right button. Can be played back unless damaged by violence.

A closer look at the original transcript of Penfield's experiment, however, reveals that memory is not a very mechanical process.There is an example of the response of a twenty-nine-year-old woman when her left temporal lobe was stimulated at Penfield: "Something came towards me from somewhere. It was a dream." Four minutes later, when the stimulus was completely When the same point: "The scenery seems to be different from just now..." And stimulate the nearby point: "Wait, something flashed over me, something I dreamed about." At the third stimulus point— — Deeper in the brain, “I keep dreaming.” Repeat stimulus to the same point: “I keep seeing things—I keep dreaming things.”

These texts speak not so much of the haphazard reappearance of yesterday plucked from the bottom shelves of the Memory Archives as of the vague flashes of a dream.The masters of these past experiences regard them as fragmentary half-memory fragments.They have a blunt "patchwork" color, floating aimlessly; dreams are born from this-those bits and pieces of stories about the past are reorganized into collages in dreams.There was no sense of déjà vu, no strong sense that "this was the case at the time".No one will be fooled by these replays. Human memory does fail.It doesn't work in very specific ways, like when you're at the grocery store and can't remember the vegetables on your grocery list or simply forget about vegetables.Impairment of memory is often associated with physical damage to the brain, leading us to suspect that memory is somehow tied up in time and space—which is the very definition of reality. Modern cognitive science, however, favors a new view: that memories are events that emerge from the aggregation of many discrete, non-memory-like fragments stored in the brain.These fragments of semi-consciousness have no fixed location, they are scattered throughout the brain.The way it is stored is fundamentally different between different consciousnesses - the mastery of shuffling cards is organized in a completely different way than the knowledge of the capital of Bolivia - and this will vary from person to person , will also vary from previous to next time. Since there are more possible thoughts or experiences than neurons can be combined in the brain, memory must be organized in a way that accommodates as many thoughts as possible than it can store.It cannot have a shelf for all the thoughts of the past, nor a place for every thought that may arise in the future. I remember one night in Taiwan twenty years ago, I was sitting in the back of an open truck on a dusty mountain road.The mountain air was cold, so I put on my jacket.I hitched a ride to reach a peak in the mountains before dawn.I looked up at the stars in the clear air as the truck scrambled up the steep, dark mountain road round and round.The sky was so clear I could see little stars close to the horizon.Suddenly, a meteor whizzed down, and because of my angle in the mountains, I saw it dancing in the atmosphere.It jumped, jumped, jumped, like a pebble. Now, when I recall the scene, that pulsating shooting star is no longer a replay of my memory—as vivid as it was.Its image does not exist in any particular place in my memory.When I recreate the experience, I actually reassemble it, and reassemble it every time I recall it.The material used is the tiny fragments of evidence scattered in my brain: shivering in the cold wind, bumping on a rough mountain road, countless stars twinkling in the night sky, and scenes of reaching out to stop cars by the side of the road.The grain of these records is even finer: the cold, the bumps, the blips, the wait.These are the original impressions we receive through our senses, and from this are combined to form our current perception. It is through these many threads scattered in memory that our consciousness creates the present just as it created the past.Standing in front of a museum exhibit whose parallel lines make me mentally associate it with the concept of a "chair", even though the exhibit has only three legs.I have never seen such a chair in my memory, but it fits everything that is associated with it - it is upright, it has a horizontal seat, it is stable, it has legs - and it follows A visual image is produced.This process is very fast.In fact, I will first notice its general "chairness" before I notice its specific details. Our memories (and our hive minds) are created in similarly vague and haphazard ways.To find the pulsating shooting star (in memory), my consciousness first grasped a trail of moving light and then assembled a cascade of sensations related to stars, cold, jolts.What kind of memory I create depends on what I have put into it recently, including feelings or other things that I added when I reorganized this memory last time.That's why each recall is slightly different, because each time it was a completely different experience in the true sense of the word.The act of perceiving is the same as the act of remembering.Both are about combining many distributed fragments into a naturally emerging whole. Cognitive scientist Douglas Hofstadter said: "Memory is highly reconstructive. Searching in memory requires selecting from a large number of events what is important and what is unimportant, emphasizing the important things, and ignore unimportant things.” This process of selection is actually perception. “I am very, very convinced,” Hofstadter told me, “that the core processes of cognition are very, very closely related to perception.” Over the past two decades, some cognitive scientists have outlined ways to create distributed memories. In the 1970s, psychologist David Marr proposed a new model of the human cerebellum, in which memories are stored randomly throughout networks of neurons. In 1974, computer scientist Penti Canelva proposed a similar mathematical network model.With this model, long strings of data can be randomly stored in computer memory.Canelva's algorithm is an elegant way of storing a finite number of data points into a very large potential memory space.In other words, Canelva pointed out a way to store whatever perception the mind has in its finite memory mechanism.Since there are more possible thoughts in the universe than atoms or particles, what the human mind can touch is only a very sparse part of it. Therefore, Canelva called his algorithm a "sparse distributed memory" algorithm. In a sparsely distributed network, memory is a type of perception.Both the act of recall and the act of perception are a pattern that is needed to probe a very large pattern selection set.When we recall, we actually reproduce the original perceptual behavior, that is, we reposition the pattern according to the process of originally perceiving the pattern. Canelva's algorithm is so concise and clear that a computer whiz can roughly implement it in an afternoon. In the mid-1980s, at NASA's Ames Research Center, Canelva and colleagues designed a very stable, practical version on a computer, fine-tuning his sparsely distributed memory architecture.Canelva's memory algorithm can do some incredible things that rival the human mind.The researchers put several low-quality digital images (1 to 9) drawn in a 20x20 grid into the sparse memory in advance.Memory holds these images.They then fed the memory a lower-quality image of a number than the first samples to see if it could "remember" what the number was.And it did!It realizes the archetypes hidden behind all the low-quality images.Essentially, it recalls images it has never seen before! This breakthrough not only makes it possible to find or recreate the past, but more importantly, it also makes it possible to unearth something from a myriad of possibilities when only the most obscure clues are given.For a memory, it is not enough to recall the face of the grandmother. It should be able to recognize the appearance of the grandmother under different lights and from different angles. A hive mind is a distributed memory capable of simultaneous perception and memory.Human thinking is mostly distributed, at least in artificial thinking distributed thinking is definitely dominant.The more computer scientists think about distributed problems in a hive way, the more plausible they become.They point out that most PCs are not actually being used the vast majority of the time they are turned on.When you're writing a letter on your computer, short bursts of keystrokes interrupt the computer's rest, only to return to idleness when you're composing your next sentence.Overall, the computers in the office that are turned on sit idle most of the day.Information systems managers at large corporations see multimillion-dollar personal computer equipment sitting idle on workers' desks at night and wonder whether they can fully utilize their computing power.All they need is a way to coordinate work and storage in a fully distributed system. However, just solving the idle problem is not the main meaning of distributed computing.Distributed systems and hive thinking have their unique advantages, such as being extremely immune to sudden failures.In the labs of Digital Equipment Corporation in Palo Alto, California, an engineer demonstrated to me the advantages of distributed computing: He opened the door of the cabinet that houses the company's internal computer network, and unplugged it dramatically from inside. a cable.Network routing bypasses the gap without hesitation. Of course, there are times when any hive mind fails.But, because of the nonlinear nature of the network, when it does fail, the failure can be similar to aphasia, which remembers everything but vegetables.A damaged network intelligence may be able to calculate the billionth digit of pi, but not forward mail to a new address; it may detect arcane textbook text classifying African zebra varieties, but not any Reasonable descriptions of animals in general.The overall "forgetfulness" of vegetables is less like a local memory failure, it is more like a failure at the system level. According to its symptoms, it may be that there is a problem with a special connection related to vegetables-like a computer It's like having two separate but conflicting programs on your hard drive create a "hole" that prevents you from printing italics.The storage location of the italics is not corrupted, but the system process that renders the italics is. Some of the obstacles to creating a distributed computational mind can be overcome by building a network of computers in a single box.This deliberately compressed distributed computing is also known as parallel computing, because thousands of computers in a supercomputer are running in parallel.Parallel supercomputers cannot solve the "idle computer on the desk" problem, nor can they aggregate computing power from far and wide; parallel operation is an advantage in itself and internally, but it is worth spending a hundred dollars for that alone. million dollars to build a stand-alone device. Parallel distributed computing is well suited to the fields of perception, vision, and simulation.Parallel mechanisms can handle complexity better than conventional supercomputers based on bulky, superfast serial computers.In a supercomputer with sparsely distributed memory, the distinction between memory and data processing disappears.Memory becomes a representation of perception, indistinguishable from the original cognitive act.Both are patterns that emerge from a mass of interconnected parts.