Jeff Hawkins - On Intelligence
Book	Page		Topic
Hawkins - On Intelligence	30		In an auto associative memory you don't have to have the entire pattern you want to  retrieve in order to retrieve it.		#####
Hawkins - On Intelligence	30		You might have only part of the pattern, or you might have a somewhat messed up pattern		0
Hawkins - On Intelligence	30		The auto-associative memory can retrieve the correct pattern, as it was originally stored, even though you start with a messy version of it.		0
Hawkins - On Intelligence	30		Going to a bank with a ripped and unreadable currency bill and the banker says "I think this is a messed up $100 bill. Give me that one, and I will give you this nice crisp $100 bill. “		0
Hawkins - On Intelligence	30		An auto associative memory can be designed to store sequences of patterns.		0
Hawkins - On Intelligence	30		With a sequence of patterns, similar to a portion of a melody, the auto associative memory can remember the entire melody.		0
Hawkins - On Intelligence	30		You might enter the first few notes of "twinkle twinkle little star" and the auto associative memory returns the whole song.		0
Hawkins - On Intelligence	31		People learn practically everything as a sequence of patterns.		1
Hawkins - On Intelligence	65		Your brain is being flooded    with the spatial and temporal patterns    from all of your senses.		34
Hawkins - On Intelligence	66		A neuron collects inputs from  its synapses,    and combines these inputs together    to decide when to output a spike    to other neurons.		1
Hawkins - On Intelligence	66		A typical neuron can cycle its functions and reset itself in about 5 ms, or around 200 times per second.		0
Hawkins - On Intelligence	66		The brain is a parallel computer.    It has billions of cells    all computing at the same time.		0
Hawkins - On Intelligence	68		The brain does not compute    the answers to problems.    It retrieves the answers    from memory.		2
Hawkins - On Intelligence	68		The essence the answers    were stored in memory    a long time ago.		0
Hawkins - On Intelligence	68		The entire cortex is a memory system.    It is not a computer at all.		0
Hawkins - On Intelligence	69		A retrieved memory    is adjusted as it is recalled    to accommodate    the particulars of the moment.		1
Hawkins - On Intelligence	69		The memory of how to catch a ball    was not programmed into your brain;    it was learned over years of repetitive practice,   and it is stored,    not calculated,   in your neurons.		0
Hawkins - On Intelligence	69		The neocortex is not like a computer,    parallel or otherwise.    Instead of computing answers to problems,    the neocortex uses stored memories    to solve problems and produce behavior.		0
Hawkins - On Intelligence	70		Neocortex stores sequences of patterns.		1
Hawkins - On Intelligence	70		Neocortex    recalls patterns    auto-associatively.		0
Hawkins - On Intelligence	70		Neocortex    stores patterns    in an invariant form.		0
Hawkins - On Intelligence	70		Neocortex    stores patterns    in a hierarchy.		0
Hawkins - On Intelligence	70		A story    is stored in your head    in a sequential fashion    and can only be recalled    in the same sequence.		0
Hawkins - On Intelligence	70		It's almost impossible    to think of anything complex    that isn't a series of events or thoughts.		0
Hawkins - On Intelligence	70		In telling a story    some people can't get to the point of it right away.    They seem to ramble on   with irritating details and tangents.		0
Hawkins - On Intelligence	70		They are promulgating the story    as it happened to them,    through time,    and cannot tell it any other way.		0
Hawkins - On Intelligence	71		All memories are like this.    You have to walk through the temporal sequence    of how you do things.		1
Hawkins - On Intelligence	71		Your memory    of the alphabet    is a sequence of patterns.		0
Hawkins - On Intelligence	71		The same thing is goals for the days of the week for the months of the year, your phone number, your countless other things.		0
Hawkins - On Intelligence	71		Your memory for songs    is a great example    of temporal sequences and memory.		0
Hawkins - On Intelligence	72		You can't recall a song backward.		1
Hawkins - On Intelligence	72		Tactile memory for textures. Your memory for the tactical memory of gravel is based on pattern sequences across the pressure and vibration sensing neurons of your skin.		0
Hawkins - On Intelligence	73		All memories    are stored in the synaptic connections    between neurons.		1
Hawkins - On Intelligence	73		Only a limited number of synapses and neurons in your skin    are playing an active role in memory recall    at any one time.		0
Hawkins - On Intelligence	73		An adult human neocortex    has an incredibly large memory capacity    but we can only remember a few    at any time    and can only do so    in a sequence of association.		0
Hawkins - On Intelligence	73		There are thousands of detailed memories    stored in the synapses of our brains    that are rarely used.		0
Hawkins - On Intelligence	73		Most of the information in our brains that we know    is sitting there idly    waiting for the appropriate cues to invoke it.		0
Hawkins - On Intelligence	73		Autoassociative nature of memory.		0
Hawkins - On Intelligence	73		Patterns    associated with themselves.		0
Hawkins - On Intelligence	73		An autoassociative memory    is one that can recall complete patterns    when given only a partial    or distorted input.		0
Hawkins - On Intelligence	73		Autoassociative memory can work    for both spatial    and temporal patterns.		0
Hawkins - On Intelligence	74		If you recall a small detail    about something that happened long ago,    the entire memory sequence    can come flooding back into your mind.		1
Hawkins - On Intelligence	74		During conversation    we often can't hear all the words    we are in a noisy environment.    Our brains fill in    what they miss    with what they expect to hear.		0
Hawkins - On Intelligence	74		It is well established that we don't actually hear    all the sounds    we perceive.		0
Hawkins - On Intelligence	74		Some people    complete of the sentences    of others aloud,    but in our minds    all of us    are doing this constantly.		0
Hawkins - On Intelligence	74		For the most part we are not aware    that we are constantly completing patterns,    but it's a ubiquitous and fundamental feature    of how memories are stored in the cortex.		0
Hawkins - On Intelligence	74		At any time,    a piece    can activate    the whole.    This is the essence of auto associated memories.		0
Hawkins - On Intelligence	74		Your neocortex    is a complex biological autoassociative memory.		0
Hawkins - On Intelligence	74		During each waking moment,    each functional region    of your neocortex    is essentially waiting vigilantly    for familiar patterns    or pattern fragments.		0
Hawkins - On Intelligence	74		The mere appearance of your friend    forces your brain    to start recalling patterns    associated with her.		0
Hawkins - On Intelligence	82		Memories are stored    in a form that captures the essence of relationships,    not the details of the moment.		8
Hawkins - On Intelligence	82		When you see, feel, or hear something,    the cortex takes the detailed highly specific input   and converts it    to an invariant form.		0
Hawkins - On Intelligence	82		It is the invariant form    that is stored in memory,    and it is the invariant form    of each new input pattern    that it gets compared to.		0
Hawkins - On Intelligence	82		Memory storage,    memory recall,    and memory recognition    occur at the level of invariant forms.		0
Hawkins - On Intelligence	82		An important function    of the neocortex    is to use its memory    to make predictions.		0
Hawkins - On Intelligence	83		When you see your friends face,    your cortex    fills in and predicts    the myriad details.		1
Hawkins - On Intelligence	83		Your cortex    makes these predictions    with great specificity.		0
Hawkins - On Intelligence	84		Your brain does this    by combining a memory    of the invariant structure of her face    with the particulars of your immediate experience.		1
Hawkins - On Intelligence	84		The combining    of invariant representations    and current input    to make detailed predictions    is exactly what is happening.		0
Hawkins - On Intelligence	84		This combining    is a ubiquitous process    that happens in every region of  cortex.		0
Hawkins - On Intelligence	84		We are able to predict    not only the words others will say,    but also in what tone of voice    they will say them.		0
Hawkins - On Intelligence	84		Three properties of cortical memory    (storing sequences,    autoassociative recall,    and invariant representations)    are necessary ingredients to predict the future    based on memories of the past.		0
Hawkins - On Intelligence	86		Our brains use stored memories    to constantly make predictions    about everything we see,    feel,    and hear		2
Hawkins - On Intelligence	86		The majority of predictions    occur outside of awareness.		0
Hawkins - On Intelligence	87		Brain constantly makes predictions    about the very fabric of the world we live in.		1
Hawkins - On Intelligence	87		The predictions    are made in parallel    and will just as readily detect    an odd texture,    a misshapen nose,    or an unusual motion.		0
Hawkins - On Intelligence	87		What we perceive    is a combination    of what we sense    and of our brain's memory=derived predictions.		0
Hawkins - On Intelligence	88		All regions of the neocortex    are simultaneously trying to predict    what their next experience will be.		1
Hawkins - On Intelligence	88		Visual areas make predictions about edges,    shapes,    objects,    locations,    and motions.		0
Hawkins - On Intelligence	88		Auditory areas make predictions about tones,    direction to source,    and patterns of sound.		0
Hawkins - On Intelligence	89		somatosensory areas make predictions about touch,    texture,    contour,    and temperature.		1
Hawkins - On Intelligence	89		Correct predictions result in understanding.		0
Hawkins - On Intelligence	89		Prediction    is the primary function of the neocortex,    and the foundation of intelligence.		0
Hawkins - On Intelligence	90		Rodolfo Llinas, I of the Vortex, the capacity to predict the outcome of future events is most likely, the ultimate and most common of all global brain functions.		1
Hawkins - On Intelligence	90		There is an entire subfield of mathematics devoted to Bayesian networks.		0
Hawkins - On Intelligence	90		Bayesian networks use probability theory to make predictions.		0
Hawkins - On Intelligence	91		Prediction is pervasive    and the basis for how you understand the world.		1
Hawkins - On Intelligence	91		When you listen to a favorite melody,    you hear the next note in your head    before it occurs.		0
Hawkins - On Intelligence	92		Neurons in your head will fire when you hear that next out fire in advance of your actual hearing it, and so you hear this song in your head.		1
Hawkins - On Intelligence	92		When listening to people speak,    you often know what they're going to say     before they finished speaking.		0
Hawkins - On Intelligence	92		People tend to use     common phrases or expressions    in much of their conversation.		0
Hawkins - On Intelligence	92		Prediction   is not always exact.    Rather, our minds work to make probabilistic predictions    concerning what is about to happen.		0
Hawkins - On Intelligence	92		At times,    our expectations    are distributed   among several  possibilities		0
Hawkins - On Intelligence	93		In Western music,    the brain automatically predicts    beats,    repeated rhythms,    completion of phrases,    and end of songs.		1
Hawkins - On Intelligence	93		The basis    of these mostly unconscious predictions    is a set of memories    that are stored in your cortex.		0
Hawkins - On Intelligence	93		Your brain can't say    exactly what will happen next,    but it nevertheless predicts which note  patterns    are likely to happen    and which aren't.		0
Hawkins - On Intelligence	93		We see    what we expect to see.		0
Hawkins - On Intelligence	97		Intelligence    is measured by the capacity to remember and predict    patterns in the world    including language,    mathematics,    physical properties of objects,    and social situations.		4
Hawkins - On Intelligence	97		The brain    receives patterns from the outside world,    stores them in memories,    and makes predictions    by combining    what it has seen before    and what is happening now.		0
Hawkins - On Intelligence	99		The human cortex    is the same thickness    and is very nearly the same structure    as the cortex of our mammal relatives.		2
Hawkins - On Intelligence	99		When evolution makes something big     very quickly,    as it did with the human cortex,    it does so by copying an existing structure.		0
Hawkins - On Intelligence	99		Humans got smart    by adding many more elements    of a common cortical algorithm.		0
Hawkins - On Intelligence	99		The human neocortex    is a relatively new structure.		0
Hawkins - On Intelligence	99		Memory and prediction    are the keys    to unlocking the mystery of intelligence.		0
Hawkins - On Intelligence	99		Start with the reptilian brain with no cortex.		0
Hawkins - On Intelligence	99		If evolution tacks on a memory system (neocortex) to the sensory path of the primitive brain, the animal gains and ability to predict the future.		0
Hawkins - On Intelligence	103		The back part of the cortex    contains the sections where the eyes,    ears,    and touch    inputs arrived.		4
Hawkins - On Intelligence	104		Reptiles with sophisticated senses and sophisticated but relatively rigid behaviors.		1
Hawkins - On Intelligence	104		Nature discovered that by adding a memory system    and feeding the sensory stream into it,    the animal could remember past experiences.		0
Hawkins - On Intelligence	104		When the animal found itself in the same or similar situation, the memory will be recall, leading to a prediction of what was likely to happen next.		0
Hawkins - On Intelligence	104		Intelligence and understanding    started as a memory system    that fed predictions into the sensory stream.		0
Hawkins - On Intelligence	104		To know something means that you can make predictions about it.		0
Hawkins - On Intelligence	104		With humans,    the cortex has taken over most of the motor behavior.		0
Hawkins - On Intelligence	104		Instead of just making predictions based on behavior of the old brain.    The human neocortex    directs behavior    to satisfy the prediction.		0
Hawkins - On Intelligence	105		To make predictions of future events,    the neocortex must store sequences of patterns.		1
Hawkins - On Intelligence	105		To recall the appropriate memories,    the brain has to retrieve patterns    by their similarity to past patterns (auto associative recall).		0
Hawkins - On Intelligence	105		Memories have to be stored in an invariant form    so that the knowledge of past events    can be applied to new situations    that are similar but not identical    to the past.		0
Hawkins - On Intelligence	113		Information also flows    from higher to lower regions    via a network of feedback connections.		8
Hawkins - On Intelligence	113		As many if not more feedback connections in visual cortex as there are feedforward connections.		0
Hawkins - On Intelligence	113		The cortex's core function is to make predictions.		0
Hawkins - On Intelligence	113		Prediction requires a comparison    between what is happening    and what you expect to happen.		0
Hawkins - On Intelligence	113		What is actually happening flows up, and what you expect to happen flows down.		0
Hawkins - On Intelligence	113		The same feedforward=feedback process    is occurring in all your cortical areas    involving all your senses.		0
Hawkins - On Intelligence	114		The transformation    from fast-changing to slow changing    and from spatially specific to spatially invariant,    is well documented for vision.		1
Hawkins - On Intelligence	115		For hearing, when someone speaks to you, the changes in sound pressure occur very rapidly; the patterns entering the primary auditory area change just as rapidly.		1
Hawkins - On Intelligence	115		Patterns received by the first auditory area    can vary widely.    A word can be spoken    with different accents,    and different pitches, or at different speeds.		0
Hawkins - On Intelligence	115		Higher up in the cortex,     those low level features don't matter;     a word is a word    regardless of the acoustic details.		0
Hawkins - On Intelligence	115		We see the same kind of feedback,    prediction,    and invariant recall    in auditory cortex    as we see in the visual system.		0
Hawkins - On Intelligence	116		There are very different inputs arriving at different locations in the primary battle centers somatosensory cortex. Again we would find selves and regions several steps removed from the primary input that respond to an object.		1
Hawkins - On Intelligence	117		Information    flows up and down    sensory areas of the cortex.		1
Hawkins - On Intelligence	117		The downward flow    fills in the current input    and makes predictions   about what we will experience next.		0
Hawkins - On Intelligence	117		Something I hear    can lead to a prediction    of what I should see or feel.		0
Hawkins - On Intelligence	117		Information flows up    the auditory hierarchy    to an association area    that connects vision with hearing.		0
Hawkins - On Intelligence	117		The representation then flows back down     the auditory and visual hierarchies,    leading to both auditory    and visual predictions.		0
Hawkins - On Intelligence	117		This kind of     multisensory prediction    is occurring all the time.		0
Hawkins - On Intelligence	118		Information    simultaneously flows up and down    the sensory hierarchies    to create a unified sensory experience    involving prediction in all senses.		1
Hawkins - On Intelligence	119		All the little sensations    are fully integrated    into our perceptual predictions.		1
Hawkins - On Intelligence	119		These predictions    can only come about    by massive coordination    of patterns streaming up and down    the cortical hierarchy.		0
Hawkins - On Intelligence	119		The entire neocortex,    all the sensory and association areas,    act as one.		0
Hawkins - On Intelligence	119		We have an overarching sensory system,    sights,    sounds,    touch,    and more    combined,    all flowing up and down    a single multi-branched hierarchy.		0
Hawkins - On Intelligence	119		All predictions    are learned by experience.		0
Hawkins - On Intelligence	119		You are not born with    any of this knowledge;    you learned it through the incredibly large capacity    of your cortex    to remember patterns.		0
Hawkins - On Intelligence	120		If there are consistent patterns    among the inputs flowing into your brain,    your cortex will use them    to predict future events.		1
Hawkins - On Intelligence	120		An input in a sensory area    can flow to an association area    which can lead to    a pattern flowing down the motor cortex    resulting in behavior.		0
Hawkins - On Intelligence	120		We interpret    these downward flowing patterns    as predictions.		0
Hawkins - On Intelligence	120		In the motor cortex    we interpret the downward flowing patterns    as motor commands.		0
Hawkins - On Intelligence	120		As Mountcastle pointed out,    the motor cortex    looks like    the sensory cortex.		0
Hawkins - On Intelligence	120		The way the cortex processes    downward-flowing sensory predictions    is similar to    how it processes    downward-flowing motor commands.		0
Hawkins - On Intelligence	120		There are    no pure sensory    or pure motor areas    in the cortex.		0
Hawkins - On Intelligence	120		Sensory patterns    simultaneously flow in anywhere and everywhere,    and then flow back down    in any area of the hierarchy,    leading to predictions or motor behavior.		0
Hawkins - On Intelligence	120		Although the motor cortex    has some special attributes,    it is correct to think of it    as just part of one large hierarchical memory-prediction system.		0
Hawkins - On Intelligence	120		Seeing,    hearing,    touching,    and acting    are profoundly intertwined.		0
Hawkins - On Intelligence	122		V1,V2, and V4,    each is a collection    of many smaller subregions.		2
Hawkins - On Intelligence	122		The largest region by far    is V1,    the primary visual area.		0
Hawkins - On Intelligence	122		Next would be V2.    They are large    compared to most regions.		0
Hawkins - On Intelligence	122		V1 is made up of numerous separate little cortical areas     that are only connected to their neighbors indirectly,    through regions higher up in the hierarchy.		0
Hawkins - On Intelligence	122		V1 would have    the largest number of small subregions    of any visual area.		0
Hawkins - On Intelligence	122		V2 would be composed    of fewer,    slightly larger subregions.		0
Hawkins - On Intelligence	122		The same would be true    for V4.		0
Hawkins - On Intelligence	122		The top region IT    would have only a single region,    which has a birds eye view    of the entire visual world.		0
Hawkins - On Intelligence	123		The cortex    now looks similar    everywhere.		1
Hawkins - On Intelligence	123		Pick any region    and you will find many lower regions    providing converging sensory input.		0
Hawkins - On Intelligence	123		The receiving region    sends projections back    to the input regions,    telling them what patterns    they should expect to see next.		0
Hawkins - On Intelligence	123		Higher association areas    unite information    from multiple senses    such as vision and touch.		0
Hawkins - On Intelligence	123		A lower region    like a subregion of V2    unites the information    from separate subregions    within V1.		0
Hawkins - On Intelligence	123		A region doesn't know – indeed it can't know – what any of those inputs mean.		0
Hawkins - On Intelligence	123		An association area    doesn't need to know    that it is handling visual input    from multiple pairs of V1.		0
Hawkins - On Intelligence	123		An association area    doesn't need to know    that it is handling input    from vision and hearing.		0
Hawkins - On Intelligence	123		The job of any cortical region    is to find out how its inputs are related,    to memorize the sequence of correlations between them,    and to use its memory    to predict how the inputs    will behave in the future.		0
Hawkins - On Intelligence	124		The same process    is happening everywhere:     a common cortical algorithm.		1
Hawkins - On Intelligence	124		This hierarchical depiction    helps us understand    the process of creating invariant representations.		0
Hawkins - On Intelligence	124		The most important result    of this depiction of cortical hierarchy    is that every region of cortex    forms invariant representations.		0
Hawkins - On Intelligence	124		Now we can say that invariant representations are ubiquitous.    Invariant representations are formed in every cortical region.		0
Hawkins - On Intelligence	124		Every region    forms invariant representations    drawn from the input areas    hierarchically below it.		0
Hawkins - On Intelligence	124		Thus the subregions of V4, V2, and V1    create invariant representations    based on what flows into them.		0
Hawkins - On Intelligence	125		Association regions    above IT    form invariant representations    of patterns from multiple senses.		1
Hawkins - On Intelligence	125		All regions of the cortex    form invariant representations    of the world beneath them in the hierarchy.		0
Hawkins - On Intelligence	125		Why is the neocortex built as a hierarchy?    Because the cortex    has built a model of the world.		0
Hawkins - On Intelligence	125		The cortex's hierarchical structure    stores a model    of the hierarchical structure of the real world.		0
Hawkins - On Intelligence	125		The real world's nested structure   is mirrored    by the nested structure of the cortex.		0
Hawkins - On Intelligence	126		Patterns from the retina    entering your primary visual cortex    are being combined    to form components of visual objects.		1
Hawkins - On Intelligence	126		The function of the cortex and the method by which it learns naturally discover the hierarchical relationships in the world.		0
Hawkins - On Intelligence	127		You are not born with    knowledge of language,    houses,    or music.		1
Hawkins - On Intelligence	127		The cortex has a clever learning algorithm     that naturally finds    whatever hierarchical structure exists    and captures it.		0
Hawkins - On Intelligence	127		Higher regions of the cortex    are keeping track of the big picture    while lower areas are actively dealing with the fast-changing small details.		0
Hawkins - On Intelligence	127		Since we can only touch,    hear,    and see    a very small part of the world     any moment in time,    information flowing into the brain    naturally arise as a sequence of patterns.		0
Hawkins - On Intelligence	127		The cortex functions to learn     those sequences    that occur over and over again.		0
Hawkins - On Intelligence	128		Each region of cortex    sees a stream of patterns.		1
Hawkins - On Intelligence	128		If the patterns are related    in such a way    that the region can learn to predict    what pattern will occur next,    the cortical region    forms a persistent representation,   or memory,    for the sequence.		0
Hawkins - On Intelligence	128		Learning sequences    is the most basic ingredient    for forming invariant representations    of real-world objects.		0
Hawkins - On Intelligence	128		Predictability    is the very definition of reality.		0
Hawkins - On Intelligence	129		The brain can be said to store sequences of sequences.    Each region of the cortex    learns the sequences,      develops what Hawkins calls “names” for the  sequences it knows,    and passes these names      to the next regions    higher in the cortical hierarchy.		1
Hawkins - On Intelligence	129		As information moves up    from primary sensory regions    to higher levels,    we see fewer and fewer    changes over time.		0
Hawkins - On Intelligence	129		In primary visual areas like V1,    the set of active cells    is changing rapidly    as new patterns fall on the retina    several times each second.		0
Hawkins - On Intelligence	129		In visual area IT,    self firing patterns    are more stable.		0
Hawkins - On Intelligence	129		Each region of cortex    has a repertoire    of sequences it knows.		0
Hawkins - On Intelligence	129		Regions    store  these sequences    about anything and everything.		0
Hawkins - On Intelligence	129		Each cortical region    has a name    for each sequence    it knows.		0
Hawkins - On Intelligence	129		The "name"    is a group of cells    whose collected firing    represents the set of objects    in the sequence.		0
Hawkins - On Intelligence	129		These cells remain active    as long as the sequences playing,    and it is this “name”    that is passed up    to the next region in the hierarchy.		0
Hawkins - On Intelligence	130		As long as the input patterns    are part of a predictable sequence,    the region presents a constant “name”    to the next higher region.		1
Hawkins - On Intelligence	130		We can imagine region IT    at the top of the visual hierarchy    relaying to an association area above it, "I am seeing a face."		0
Hawkins - On Intelligence	130		In this way, a predictable    sequence of events    gets identified    with a "name."		0
Hawkins - On Intelligence	130		This happens over and over again    as we go up    in the hierarchical pyramid.		0
Hawkins - On Intelligence	130		One region    might recognize a sequence of sounds    that comprise phonograms.		0
Hawkins - On Intelligence	130		The next higher region    recognizes sequences the phonems    to create words.		0
Hawkins - On Intelligence	130		The next higher region    recognizes sequences of words    to create phrases,    and so on.		0
Hawkins - On Intelligence	130		By collapsing predictable sequences    into named objects    at each region in the hierarchy,    we achieve more and more stability    the higher we go.		0
Hawkins - On Intelligence	130		This creates invariant representation.		0
Hawkins - On Intelligence	130		The opposite effect happens    as a pattern moves back down the hierarchy:    stable patterns    get unfolded into sequences.		0
Hawkins - On Intelligence	131		At this point the unfolding pattern splits    and travels down both the auditory section of cortex    and the motor section of cortex.		1
Hawkins - On Intelligence	131		Following the motor path,    each word is unfolded    into a memorized sequence of phonems.		0
Hawkins - On Intelligence	131		In the final bottom region,    each phonem is unfolded    into a sequence of muscle commands    to make sounds.		0
Hawkins - On Intelligence	131		The lower you look in the hierarchy,    the faster the patterns are changing.		0
Hawkins - On Intelligence	131		A single constant pattern    at the top of the motor hierarchy    eventually leads to    a complex and lengthy sequence    of speech sounds.		0
Hawkins - On Intelligence	131		If you want to type the Gettysburg address    you start with the same pattern    at the top of the hierarchy.		0
Hawkins - On Intelligence	131		Words are unfolded into letters,    and the letters are unfolded into muscle commands    to your fingers    for typing.		0
Hawkins - On Intelligence	131		A single memory of the speech    can take various behavioral forms.		0
Hawkins - On Intelligence	131		At any region,    an invariant pattern    can bifurcate    and follow a different path down.		0
Hawkins - On Intelligence	131		Representations of simple objects    at the bottom of the hierarchy    can be reused over and over    for different high-level sequences.		0
Hawkins - On Intelligence	131		A hierarchy of nested sequences    allows the sharing and reuse    of lower-level objects – words,    phonems,    and letters    being  but a few examples.		0
Hawkins - On Intelligence	132		It is a remarkably efficient way    to store information about the world and its structure    and very different    from how computers work.		1
Hawkins - On Intelligence	132		The same unfolding of sequences    occurs in the sensory    as well as the motor regions.		0
Hawkins - On Intelligence	132		The unfolding of sequences    allows you to perceive and understand objects    from different views.		0
Hawkins - On Intelligence	132		The way you memorize sequences    and represent them by name    as information goes up and down the cortical hierarchy    may remind you of the hierarchy of military command.		0
Hawkins - On Intelligence	133		If something goes wrong that cannot be handled by subordinates down the chain of command, then the issue rises up the hierarchy until someone knows  what to do next.		1
Hawkins - On Intelligence	133		What was an unanticipated problem to subordinates is just the expected next task for the officer in charge.		0
Hawkins - On Intelligence	133		If lower regions of cortex    failed to predict what patterns they are seeing,    they consider that an error    and pass the error up the hierarchy.     This is repeated    until some region does anticipate the pattern.		0
Hawkins - On Intelligence	133		Every cortical region    attempts to    store and recall    sequences.		0
Hawkins - On Intelligence	133		The bottom-up inputs to a region of cortex    are input patterns    carried on thousands or millions of axons.		0
Hawkins - On Intelligence	133		These axons    come from different regions    and contain    all sorts of patterns.		0
Hawkins - On Intelligence	133		The number of possible patterns    that can exist on even 1000 axons    is larger than    the number of molecules in the universe.		0
Hawkins - On Intelligence	133		A region will only see a tiny fraction of these possible patterns in a lifetime.		0
Hawkins - On Intelligence	134		The brain    must classify patterns.		1
Hawkins - On Intelligence	134		You are going to look at all the input patterns    coming in from lower cortical regions,     classify them,    and then look for sequences.		0
Hawkins - On Intelligence	134		Both steps,    classification and sequence formation,    are necessary    to create invariant representations,     and each region of cortex    does them.		0
Hawkins - On Intelligence	135		If you know    the most likely sequence    for a series of inputs,    you will use this knowledge    to decide    how to classify    the ambiguous input.		1
Hawkins - On Intelligence	135		You use the context of known sequences    to resolve ambiguity		0
Hawkins - On Intelligence	135		When people speak    their individual words    very often cannot be understood    out of context.		0
Hawkins - On Intelligence	135		Most of the time    you are aware    that you are filling in    ambiguous or incomplete information    from your memories or sequence		0
Hawkins - On Intelligence	135		You hear what you expect to hear and see what you expect to see – at least when what you hear and see fits into past experience.		0
Hawkins - On Intelligence	135		Memory of sequences allows you not only to resolve ambiguity and current input, but also to predict which input should happen next.		0
Hawkins - On Intelligence	135		By recognizing a sequence of patterns,    a cortical region will predict its next input pattern    and tell the region below    what to expect.		0
Hawkins - On Intelligence	135		A region of cortex    not only learns familiar sequences,    it also learns how    to modify its classifications.		0
Hawkins - On Intelligence	136		In cortical regions,    bottom-up classifications    and top-down sequences    are constantly interacting,    changing throughout your life.		1
Hawkins - On Intelligence	136		This is the essence of learning.		0
Hawkins - On Intelligence	136		Forming new classifications    and new sequences    is how you remember the world.		0
Hawkins - On Intelligence	136		Another part of the cortical job    is to relay the name of the sequence you are seeing    to the next level up.		0
Hawkins - On Intelligence	137		The hierarchy of the cortex    ensures that memories of objects are distributed over the hierarchy;    they aren't located in a single spot.		1
Hawkins - On Intelligence	137		Because each region of the hierarchy    forms invariant memories,    what a typical region of cortex learns is sequences of invariant. representations,    which are themselves sequences of invariant memories		0
Hawkins - On Intelligence	137		When you think about the world you are recalling sequences of patterns that correspond to the way the objects of the world are and how they behave.		0
Hawkins - On Intelligence	137		The order in which you experience parts of the world is determined by the world structure.		0
Hawkins - On Intelligence	137		An invariant representation in any region of the cortex can be turned into a detailed prediction of how it will appear on yours senses by propagating the pattern down the hierarchy.		0
Hawkins - On Intelligence	137		Similarly, an invariant representation in the motor cortex can be turned into detailed and situation specific motor commands by propagating the pattern down the motor hierarchy.		0
Hawkins - On Intelligence	138		We will start with a description of what a cortical region looks like.		1
Hawkins - On Intelligence	138		Cortical regions    vary greatly in size,    the largest being the primary sensory areas.		0
Hawkins - On Intelligence	138		Let's assume that a typical cortical area    is the size of a small coin.		0
Hawkins - On Intelligence	139		The density and shape    of the cells in the cortex    vary as you move from top to bottom.		1
Hawkins - On Intelligence	139		These differences    define the layers.		0
Hawkins - On Intelligence	139		Layer 1, the top layer    is the most distinct of the six layers.    It has very few cells    consisting primarily of a mat of axons    running parallel to the cortical surface		0
Hawkins - On Intelligence	139		Layers 2 and 3 looks similar.    They contain many,    tightly packed. pyramidal cells.		0
Hawkins - On Intelligence	139		Layer 4 has a type of star-shaped cell.		0
Hawkins - On Intelligence	139		Layer 5 has regular pyramidal cells    as well as a class of extra big pyramidal-shaped cells.		0
Hawkins - On Intelligence	139		Layers 6,  the bottom layer   also has several types of unique neurons.		0
Hawkins - On Intelligence	139		Columns of cells    run perpendicular to the layers.    You can think of columns    as being vertical units of cells    that work together.		0
Hawkins - On Intelligence	139		The layers within each column    are connected by axons    that run up and down,    making synapses along the way.		0
Hawkins - On Intelligence	139		Columns do not have clear boundaries    but their existence can be inferred    from several lines of evidence.		0
Hawkins - On Intelligence	139		One reason is that it vertically aligned cells    in each column    tend to become active    for the same stimulus.		0
Hawkins - On Intelligence	139		The cells within each column    are strongly connected.		0
Hawkins - On Intelligence	140		Activity spreads up and down    within a column of cells.		1
Hawkins - On Intelligence	140		In an embryo,    single precursor cells    migrate from an inner brain cavity    to where the cortex takes take shape.		0
Hawkins - On Intelligence	140		Each of these cells    divides    to create about 100 neurons,    called a microcolumn.		0
Hawkins - On Intelligence	140		The human cortex    has an estimated several hundred million    microcolumns.		0
Hawkins - On Intelligence	140		Imagine a single microcolumn is the width of a human hair.		0
Hawkins - On Intelligence	140		The brush like mat    is a simplistic model    of the coin size cortical region.		0
Hawkins - On Intelligence	140		And information flows    mostly in the direction of the hairs:    horizontally in layer 1    and vertically in layers 2 through 6		0
Hawkins - On Intelligence	140		At least 90% of the synapses    on cells within each column    come from places outside the column itself.		0
Hawkins - On Intelligence	140		Some connections    arrived from neighboring column.		0
Hawkins - On Intelligence	140		Other connections    come from halfway across the brain.		0
Hawkins - On Intelligence	140		Vernon Mountcastle argued there is a single cortical algorithm,    he also proposed a cortical column is the basic unit of computation in the cortex.		0
Hawkins - On Intelligence	141		It is believed that the column is the basic unit of prediction.		1
Hawkins - On Intelligence	141		Converging inputs from lower regions always arrive at layer 4, the main input layer.		0
Hawkins - On Intelligence	141		Layer 4 cells then send projections up to cells in layer 2 and layer 3    within their column.		0
Hawkins - On Intelligence	141		Layer 6 cells are the downward projecting output cells    from a cortical column    and project to layer 1    in the region hierarchically below.		0
Hawkins - On Intelligence	153		The sum of all these mechanisms    allows the cortex to learn sequences,    make predictions,    and form constant representations,    or "names," for sequences.		12
Hawkins - On Intelligence	153		How does a region of the cortex    make specific predictions    from invariant memories?		0
Hawkins - On Intelligence	153		We have to combine    feedforward information (actual input)    with feedback information (a prediction in an invariant form).		0
Hawkins - On Intelligence	156		Every moment of your waking life,    each region of the cortex is comparing    a set of expected columns driven from above    with the set of observed columns driven from below.		3
Hawkins - On Intelligence	156		Where the two sets intersect    is what we perceive.		0
Hawkins - On Intelligence	156		If we had perfect input from below    and perfect predictions,    then the set of perceived columns    would always be contained    in the set of predicted columns.		0
Hawkins - On Intelligence	156		We often don't have such agreement.    The method of combining partial prediction    with partial input    resolves ambiguous input,    it fills in missing pieces of information,    and it decides between alternate views.		0
Hawkins - On Intelligence	156		It is how we decide whether a picture is of a vase or two faces.		0
Hawkins - On Intelligence	156		In addition to projecting to lower cortical regions,    layer 6 cells    can send their outputs back    into layer 4 cells    of their own column.		0
Hawkins - On Intelligence	156		When they do, our predictions become the input.		0
Hawkins - On Intelligence	156		This is what we do when daydreaming or thinking.		0
Hawkins - On Intelligence	156		It allows us to the consequences of our own predictions.		0
Hawkins - On Intelligence	156		We do this many hours a day as we plan the future, worry about events to come, or just "imagining."		0
Hawkins - On Intelligence	160		Confusion occurs    when the cortex can't find any memory    that matches with the input.		4
Hawkins - On Intelligence	160		Your eyes scan everywhere on the picture.		0
Hawkins - On Intelligence	160		New inputs race all the way up the cortical hierarchy.		0
Hawkins - On Intelligence	160		High-level cortex tries lots of different hypotheses but, as these predictions raced down the hierarchy, each and every one conflicts with the input and the cortex is forced to try again.		0
Hawkins - On Intelligence	160		During this time of confusion your brain is totally occupied with understanding the picture.		0
Hawkins - On Intelligence	160		Finally, you make a high-level prediction that is the right one.		0
Hawkins - On Intelligence	161		The prediction    starts at the top of the cortical hierarchy    and succeeds in propagating all the way to the bottom.		1
Hawkins - On Intelligence	161		In less than a second,    each region is given a sequence    that fits the data.		0
Hawkins - On Intelligence	161		No more errors    rise to the top. You understand the picture, you see a Dalmatian dog.		0
Hawkins - On Intelligence	168		Hippocampus on top of it all		7
Hawkins - On Intelligence	168		Three large brain structures    lie under the neocortical sheet and communicate with it. They are the basal ganglia,    cerebellum,    and the hippocampus.		0
Hawkins - On Intelligence	168		All three existed prior to the neocortex.		0
Hawkins - On Intelligence	168		The basal ganglia were the primitive motor system,    the cerebellum learned precise timing relationships of events,    and the hippocampus stored memories of specific events and places.		0
Hawkins - On Intelligence	168		The neocortex is responsible for all complex motor sequences    and can directly control your limbs.		0
Hawkins - On Intelligence	170		Connections between the hippocampus and the neocortex    suggest that the hippocampus is the top region of the neocortex.		2
Hawkins - On Intelligence	170		The hippocampus occupies the peak of the neocortical pyramid.		0
Hawkins - On Intelligence	170		The hippocampus not only sits at the top of the cortical pyramid,    but it still connects directly to many older parts of the brain.		0
Hawkins - On Intelligence	170		Think about the information flowing from     your eyes,    ears,    and skin    into the neocortex.		0
Hawkins - On Intelligence	170		Each region of the neocortex    tries to understand the input    in the in terms of the sequences it knows.		0
Hawkins - On Intelligence	170		If it does understand the input    it does not pass on the details    to higher levels of the hierarchy.		0
Hawkins - On Intelligence	170		If a region does not understand the current input,    it passes it up the hierarchy    until some higher region    does understand it.		0
Hawkins - On Intelligence	170		A pattern that is truly novel    will escalate further and further up the hierarchy   until some higher region    does understand it.		0
Hawkins - On Intelligence	170		The net effect is that when you get to the top of the cortical pyramid,    what you have left is information that can't be understood    by prior experience.		0
Hawkins - On Intelligence	171		You are left with    the part of the input    that is truly new and unexpected.		1
Hawkins - On Intelligence	171		It is these unexplained and unanticipated remainders,    the new stuff,    that enters the hippocampus    and is stored there.		0
Hawkins - On Intelligence	171		This new, fresh information    won't be stored forever.    Either it will be transferred into the cortex    or it will eventually be lost.		0
Hawkins - On Intelligence	171		The hippocampus has a heterogeneous structure    with several specialized regions.    It's good at the unique task of quickly storing    whatever pattern it sees.		0
Hawkins - On Intelligence	171		You can instantly remember a novel event in the hippocampus,    but you will permanently remember something in the cortex   only if you experience it over and over,    either in reality or by thinking of it.		0
Hawkins - On Intelligence	171		Alternate Path up the Hierarchy.		0
Hawkins - On Intelligence	171		The cortex has a second major pathway    for passing information    from region to region,    up the hierarchy.		0
Hawkins - On Intelligence	172		The alternate path starts with cells in layer 5    which project to the thalamus    and then up to the next higher region    of cortex.		1
Hawkins - On Intelligence	172		As we move up the cortical hierarchy,    there is a direct path    between two regions    and an indirect path    through the thalamus.		0
Hawkins - On Intelligence	173		The alternate pathway    through the thalamus    is likely the mechanism    by which we attend to details    that normally we would not notice.		1
Hawkins - On Intelligence	173		It bypasses the grouping of sequences    in layer 2,    sending the raw data to the next higher region    of cortex.		0
Hawkins - On Intelligence	174		In this way unusual events    quickly rise    to your attention.		1
Hawkins - On Intelligence	174		This is why   we can't avoid focusing on deformities    and other unusual patterns.		0
Hawkins - On Intelligence	174		Often however, errors aren't strong enough    to open the alternate pathway.    This is why we sometimes don't notice    if a word is misspelled as we read it.		0
Hawkins - On Intelligence
Hawkins - On Intelligence
Hawkins - On Intelligence
Hawkins - On Intelligence