Glocal Memory: A New Perspective on Knowledge Representation, Neurodynamics, Distributed Cognition, and the Nature of Mind

Glocal Memory: A New Perspective on Knowledge
Representation, Neurodynamics, Distributed Cognition,
and the Nature of Mind
Ben Goertzel

Novamente LLC, 1405 Bernerd Place, Rockville MD 20851 {ben@goertzel.org}
Abstract. We describe a novel conceptual and formal model of memory
structure that combines key aspects of global and local knowledge
representation. Applications of this “glocal memory” model to artificial general
intelligence are discussed, in the context of the “Novamente Cognition Engine”
and OpenCog software systems, and simpler prototype systems constituting
“glocal Hopfield nets.” Building on recent results regarding visual memory
derived from single-neuron recordings, an hypothesis is made regarding how
some instances of glocal memory might be achieved in the human brain. It is
also argued that glocal memory models the concept of “distributed cognition”
according to which an individual human mind is both localized in a brain and
distributed through a network of interactions with tools and other embodied
minds.
1 Introduction
“Memory,” considered in a general sense, is central to intelligence and to complex
systems in general. Without the ability to reflect its past experiences in its present
structure, a system cannot adapt, and hence cannot manifest either intelligence or
complexity.
However, the conceptual tools currently available for modeling and analyzing
memory structures and dynamics (on the conceptual or formal level) are
disappointingly simplistic. In particular, popular approaches to understanding
memory tend to focus either on local memory (in which a memory is stored in one
place within a system) or global memory (in which a memory is stored as some sort
of pattern of activation distributed across a system). My suggestion is that neither
global nor local memory models are sufficiently subtle or flexible to capture the way
memory works in complex intelligent systems like human brains, nor the way
memory should work in AI systems if these systems are to demonstrate robust
intelligence. Here I introduce a new concept of glocal memory, intended specifically
to describe certain sorts of memory structures that combine local and global aspects.
The basic notion of glocal memory is not entirely new; for instance, the hypothesis of
(what is here called) glocal memory in the human cortex was described in detail in
my1997 book From Complexity to Creativity. However, the glocal memory concept

---

has not previously been named, explicitly formalized, nor discussed in such a general
context as is done here.
The central idea of glocal memory is that declarative, episodic or procedural items
may be stored in memory in the form of paired structures that are called (key, map)
pairs. The key is a localized version of the item and records some significant aspects
of the items in a simple and crisp way. The map is a dispersed, distributed version of
the item which, represents the item as a (to some extent, dynamically shifting)
combination of fragments of other items. The map includes the key as a subset;
activation of the key generally (but not necessarily always) causes activation of the
map; and changes in the memory item will generally involve complexly coordinated
changes on the key and map level both.
After presenting a simple formalization of the glocal memory concept, applications
of the idea to various aspects of human and artificial intelligence are discussed. The
hypothesis of glocal memory in the human brain, from (Goertzel, 1997), is updated in
the light of more recent neuroscience thinking, and some implications for the design
of formal neural networks are considered. Next, the manifestation of glocal memory
in AI is reviewed in the context of the Novamente Cognition Engine (Goertzel, 2006)
and OpenCog (Goertzel, 2008) software systems. Finally, attention is then paid to the
application of glocal memory to notions of “distributed cognition” (in which the mind
of a human or other embodied organism is considered as extending beyond its body
into those aspects of the world that regularly and intricately interact with its body).
2 A Simple Formalization of Glocal Memory
To explain the notion of glocal memory more precisely, we will introduce a simple
formal model of a system S that uses a memory to record information relevant to the
actions it carries out. The overall concept of glocal memory should not be considered
as restricted to this particular formal model. The formal model is not intended for
maximal generality, but is intended to encompass a variety of current AI system
designs and formal neurological models.
In this model, we will consider S’s memory subsystem as a set of objects we’ll call
“tokens,” embedded in some metric space. The metric in the space, which we will
call the “basic distance” of the memory, generally will not be defined in terms of the
semantics of the items stored in the memory; though it may come to shape these
dynamics through the specific architecture and evolution of the memory. Note that
these tokens are not intended as generally being mapped one-to-one onto meaningful
items stored in the memory. The “tokens” are the raw materials that the memory uses
to store items.
We assume that each token, at each point in time, may meaningfully be assigned a
certain quantitative “activation level.” Also, tokens may have other numerical or
discrete quantities associated with them, depending on the particular memory
architecture. Finally, tokens may relate other tokens, so that optionally a token may
come equipped with an (ordered or unordered) list of other tokens.
To understand the meaning of the activation levels, one should think about S’s
memory subsystem as being coupled with an action-selection subsystem, that

---

dynamically chooses the actions to be taken by the overall system in which the two
subsystems are embedded. Each combination of actions, in each particular type of
context, will generally be associated with the activation of certain tokens in memory.
Then, as analysts of the system S, we may associate each token T with an
“activation vector” v(T,t), whose value for time t consists of the activation of the
token T at time t.
“Items stored in memory” over a certain period of time, may then be defined as
clusters in the set of activation vectors associated with memory during that period of
time. Note that the system S itself may explicitly recognize and remember patterns
regarding what items are stored in its memory – but, from an external analyst’s
perspective, the set of items in S’s memory is not restricted to the ones that S has
explicitly recognized as memory items.
The “localization” of a memory item may be defined as the degree to which the
various tokens involved in the item are close to each other according to the metric in
the memory metric-space. This degree may be formalized in various ways, but
choosing a particular quantitative measure is not important here. A highly localized
item may be called “local” and a not-very-localized item may be called “global.”
We may define the “activation distance” of two tokens as the distance between
their activation vectors. We may then say that a memory is “well aligned” to the
extent that there is a correlation between the activation distance of tokens, and the
basic distance of the memory metric-space.
Given the above set-up, the basic notion of glocal memory can be enounced fairly
simply. A glocal memory is one:

•
That is reasonably well-aligned (i.e. the correlation between activation
and basic distance is significantly greater than random)
•
In which most memory items come in pairs, consisting of one local item
and one global item, so that activation of the local item (the “key”)
frequently leads in the near future to activation of the global item (the
“map”)

Obviously, in the scope of all possible memory structures constructible within the
above formalism, glocal memories are going to be very rare and special. But, I
suggest that they are important, because they are generally going to be the most
effective way for intelligent systems to structure their memories.
An example of a predominantly local memory structure, in which nearly all
significant memory items are local according to the above definition, is the Cyc
logical reasoning engine (Lenat and Guha, 1990). To cast the Cyc knowledge base in
the present formal model, the tokens are logical predicates. Cyc does not have an in-
built notion of activation, but one may conceive the activation of a logical formula in
Cyc as the degree to which the formula is used in reasoning or query processing at a
certain point in time. And one may define a basic metric for Cyc by associating a
predicate with its extension, and definiting the similarity of two predicates as the
symmetric distance of their extensions. Cyc is reasonably well-aligned, but according
to the dynamics of its querying and reasoning engines, it is basically a local memory
structure without significant global memory structure.

---

On the other hand, an example of a predominantly global memory structure, in
which nearly all significant memory items are global according to the above
definition, is the Hopfield associative memory network (Amit, 1989). Here memories
are stored in the pattern of weights associated with synapses within a network of
formal neurons, and each memory in general involves a large number of the neurons
in the network. To cast the Hopfield net in the present formal model, the tokens are
neurons and synapses; the activations are neural net activations; the basic distance
between two neurons A and B may be defined as the percentage of the time that
stimulating one of the neurons leads to the other one firing; and to calculate a basic
distance involving a synapse, one may associate the synapse with its source and target
neurons. With these definitions, a Hopfield network is a well-aligned memory, and
(by intentional construction) a markedly global one. Local memory items will be
very rare in a Hopfield net.
While predominantly local and predominantly global memories may have great
value for particular applications, my suggestion is that they also have inherent
limitations. If so, it means that the most useful memories are going to be those that
involve both local and global memory items in central roles. However, this is a more
general and less risky claim than the assertion that glocal memory structure as defined
above is important. Because, “glocal” as defined above doesn’t just mean “neither
predominantly global nor predominantly local.” Rather, it refers to a specific pattern
of coordination between local and global memory items – what I have called the
“keys and maps” pattern.
3 Hints of Glocal Memory in the Human Brain
Our understanding of human brain dynamics is still very primitive, one
manifestation of which is the fact that we really don’t understand how the brain
represents knowledge, except in some very simple respects. So anything anyone says
about knowledge representation in the brain, at this stage, has to be considered highly
speculative. Existing neuroscience knowledge does imply constraints on how
knowledge representation in the brain may work, but these are relatively loose
constraints. These constraints do imply that, for instance, the brain is neither a
relational database (in which information is stored in a wholly localized manner) nor a
collection of “grandmother neurons” that respond individually to high-level percepts
or concepts; nor a simple Hopfield type neural net (in which all memories are
attractors globally distributed across the whole network). But they don’t tell us nearly
enough to, for instance, create a formal neural net model that can confidently be said
to represent knowledge in the manner of the human brain.
As an initial example of the current state of knowledge, I’ll discuss here a series of
papers regarding the neural representation of visual stimuli (Quinoga et al, 2005;
2008), which deal with the fascinating discovery of a subset of neurons in the medial
temporal lobe (MTL) that are selectively activated by strikingly different pictures of given
individuals, landmarks or objects, and in some cases even by letter strings. For instance,
in the 2005 paper, it is noted that

in one case, a unit responded only to three completely different images of the

---

ex-president Bill Clinton. Another unit (from a different patient) responded
only to images of The Beatles, another one to cartoons from The Simpson’s
television series and another one to pictures of the basketball player Michael
Jordan.

These empirical results seem quite clear and exciting, yet the authors’ theoretical
interpretation of the data has consistently been much less so. In the 2005 abstract,
they note that their results “suggest that neurons might encode an abstract
representation of an individual.” And indeed, the title of the 2005 paper is the rather
gutsily worded “Invariant visual representation by single neurons in the human
brain.” Yet in the paper’s conclusion the authors tell a somewhat more conservative
story:

How neurons encode different percepts is one of the most intriguing
questions in neuroscience. Two extreme hypotheses are
schemes based on the explicit representations by highly selective
(cardinal, gnostic or grandmother) neurons and schemes that rely on
an implicit representation over a very broad and distributed population
of neurons. In the latter case, recognition would require the
simultaneous activation of a large number of cells and therefore we
would expect each cell to respond to many pictures with similar basic
features. This is in contrast to the sparse firing we observe, because
most MTL cells do not respond to the great majority of images seen
by the patient. Furthermore, cells signal a particular individual or
object in an explicit manner27, in the sense that the presence of the
individual can, in principle, be reliably decoded from a very small
number of neurons.We do not mean to imply the existence of single
neurons coding uniquely for discrete percepts for several reasons:
first, some of these units responded to pictures of more than one
individual or object; second, given the limited duration of our
recording sessions, we can only explore a tiny portion of stimulus
space; and third, the fact that we can discover in this short time some
images—such as photographs of Jennifer Aniston—that drive the
cells suggests that each cell might represent more than one class of
images. Yet, this subset of MTL cells is selectively activated by
different views of individuals, landmarks, animals or objects. This
is quite distinct from a completely distributed population code and
suggests a sparse, explicit and invariant encoding of visual percepts in MTL.


It seems that the alternate title “Invariant visual representation by sparse neuronal
subnetworks in the human brain,” might have better captured their actual conclusions
– and yet, this weaker title would not have communicated the exciting nature of some
of their individual findings, such as the subject who apparently had “Bill Clinton”
neurons which did not fire in response to other test images (though obviously this
doesn’t rule out that those neurons might have fired in a variety of other
circumstances, which indeed I suspect would be the case).
The 2008 paper backed away from the more extreme interpretation in the title as
well as the conclusion, with the title “Sparse but not "Grandmother-cell" coding in the
medial temporal lobe.” As the authors emphasize there,

---

Given the very sparse and abstract representation of visual information by
these neurons, they could in principle be considered as ‘grandmother cells’.
However, we give several arguments that make such an extreme interpretation
unlikely.


...

MTL neurons are situated at the juncture of transformation of percepts into
constructs that can be consciously recollected. These cells respond to
percepts rather than to the detailed information falling on the retina. Thus,
their activity reflects the full transformation that visual information
undergoes through the ventral pathway. A crucial aspect of this
transformation is the complementary development of both selectivity and
invariance. The evidence presented here, obtained from recordings of single-
neuron activity in humans, suggests that a subset of MTL neurons possesses
a striking invariant representation for consciously perceived objects,
responding to abstract concepts rather than more basic metric details. This
representation is sparse, in the sense that responsive neurons fire only to
very few stimuli (and are mostly silent except for theirpreferred stimuli), but it
is far from a Grandmother-cell representation. The fact that the MTL
represents conscious abstract information in such a sparse and invariant way
is consistent with its prominent role in the consolidation of long-term
semantic memories.

It’s interesting to note how inadequate the Quinoga et al data really is for
exploring the notion of glocal memory in the brain. Suppose it’s the case that
individual visual memories correspond to keys consisting of small neuronal
subnetworks, and maps consisting of larger neuronal subnetworks. Then it would be
not at all surprising if neurons in the “key” network corresponding to a visual concept
like “Bill Clinton’s face” would be found to respond differentially to the presentation
of appropriate images. Yet, it would also be wrong to overinterpret such data as
implying that the key network somehow comprises the “representation” of Bill
Clinton’s face in the individual’s brain. In fact this key network would comprise only
one aspect of said representation.
In the glocal memory hypothesis, a visual memory like “Bill Clinton’s face”
would be hypothesized to correspond to an attractor spanning a significant
subnetwork of the individual’s brain – but this subnetwork still might occupy only a
small fraction of the neurons in the brain (say, 1/100 or less), since there are very
many neurons available. This attractor would constitute the map. But then, there
would be a much smaller number of neurons serving as key to unlock this map: i.e. if
a few of these key neurons were stimulated, then the overall attractor pattern in the
map as a whole would unfold and come to play a significant role in the overall brain
activity landscape. In prior publications (e.g. Goertzel, 1997) I have explored this
hypothesis in more detail in terms of the known architecture of the cortex and the
mathematics of complex dynamical attractors.
So, one possible interpretation of the Quinoga et al data is that the MTL neurons
they’re measuring are part of key networks that correspond to broader map networks

---

recording percepts. The map networks might then extend more broadly throughout
the brain, beyond the MTL and into other perceptual and cognitive areas of cortex.
Furthermore, in this case, if some MTL key neurons were removed, the maps might
well regenerate the missing keys (as would happen e.g. in the glocal Hopfield model
to be discussed in the following section).
Related, interesting evidence for glocal memory in the brain comes from a recent
study of semantic memory (Patterson et al, 2007), which probed the architecture of
semantic memory via comparing patients suffering from semantic dementia (SD) with
patients suffering from three other neuropathologies, and found reasonably
convincing evidence for what they call a “distributed-plus-hub” view of memory.
The SD patients they studied displayed highly distinctive symptomology; for
instance, their vocabularies and knowledge of the properties of everyday objects were
strongly impaired, whereas their memories of recent events and other cognitive
capacities remain perfectly intact. And these patients also showed highly distinctive
patterns of brain damage: focal brain lesions in their anterior temporal lobes (ATL),
unlike the other patients who had either less severe or more widely distributed
damage in their ATLs. This led Patterson et al to conclude that the ATL (which is
adjacent to the amygdala and limbic systems, which process reward and emotion; and
the anterior parts of the medial temporal lobe memory system, which processes
episodic memory) is a “hub” for amodal semantic memory, drawing general semantic
information from episodic memories based on emotional salience.
So, in this view, the memory of something like a “banana” would contain a
distributed aspect, spanning multiple brain systems, and also a localized aspect,
centralized in the ATL. The distributed aspect would likely contain information on
various particular aspects of bananas, including their sights, smells, and touches, the
emotions the evoke, and the goals and motivations they related to. The distributed
and localized aspects would influence each other dynamically, but, the data Patterson
et al gathered do not address dynamics and they don’t venture hypotheses in this
direction.
There is a relationship between the “distributed-plus-hub” view and Damasio's
better-known notion of a “convergence zone” (Damasio, 2000), defined roughly as a
location where the brain binds features together. A convergence zone, in Damasio’s
perspective, is not a "store" of information but an agent capable of decoding a signal
(of reconstructing information). He also uses the metaphor that convergence zones
behave like indexes drawing information from other areas of the brain – but they are
not static but rather dynamic indices, containing the instructions needed to recognize
and combine the features constituting the memory of something. The mechanism
involved in the distributed-plus-hub model is similar to a convergence zone, but with
the important difference that hubs are less local: Patterson et al’s semantic hub may
be thought of a kind of “cluster of convergence zones” consisting of a network of
convergence zones for various semantic memories.
What is missing in Patterson’s and Damasio’s perspective is a vision of distributed
memories as attractors. The idea of localized memories serving as indices into
distributed knowledge stores is important, but is only half the picture of glocal
memory: the creative, constructive, dynamical-attractor aspect of the distributed

---

representation is the other half. The closest thing to a clear depiction of this aspect of
glocal memory that seems to exist in the neuroscience literature is a portion of
William Calvin’s theory of the “cerebral code” (Calvin, 1996). Calvin proposes a set
of quite specific mechanisms by which knowledge may be represented in the brain
using complexly-structured strange attractors, and by which these strange attractors
may be propagated throughout the brain. Calvin explores in great detail how a
distributed attractor may propagate from one part of the brain to another in pieces,
with one portion of the attractor getting propagated first, and then seeding the
formation in the destination brain region of a close approximation of the whole
attractor.
Calvin’s theory may be considered a genuinely glocal theory of memory.
However, it also makes a large number of other specific commitments that are not
part of the notion of glocality, such as his proposal of hexagonal meta-columns in the
cortex, and his commitment to evolutionary learning as the primary driver of neural
knowledge creation. We find these other hypotheses interesting and highly
promising, yet feel it is also important to separate out the notion of glocal memory for
separate consideration.
Regarding specifics, our suggestion is that Calvin’s approach may overemphasize
the distributed aspect of memory, not giving sufficient due to the relatively localized
aspect as accounted for in the Quinoga et al results discussed above. In Calvin’s
glocal approach, global memories are attractors and local memories are parts of
attractors. We suggest a possible alternative, in which global memories are attractors
and local memories are particular neuronal subnetworks such as the specialized ones
identified by Quinoga et al. However, this alternative does not seem contradictory to
Calvin’s overall conceptual approach, even though it is different from the particular
proposals made in (Calvin, 1996).
The above paragraphs are far from a complete survey of the relevant neuroscience
literature; there are literally dozens of studies one could survey pointing toward the
glocality of various sorts of human memory. Yet experimental neuroscience tools are
still relatively primitive, and every one of these studies could be interpreted in various
other ways. In the next couple decades, as neuroscience tools improve in accuracy,
our understanding of the role of glocality in human memory will doubtless improve
tremendously.

---

4 Glocal Hopfield Nets
Following up on the ideas of the previous section, it is interesting to explore the
notion of formal neural network models that embody the notion of glocal memory.
My colleagues and I have recently run some interesting simulations with a variation
on Hopfield netural nets that explicitly incorporates the notion of glocality. Our
technical results will be reported elsewhere, but, a brief discussion of the main ideas
would seem appropriate here.
Essentially, we augment the standard Hopfield net architecture by adding a set of
“key neurons.” These are a small percentage of the neurons in the network, and are
intended to be roughly equinumerous to the number of memories the network is
supposed to store. When the Hopfield net converges to an attractor A, then new links
are created between the neurons that are active in A, and one of the key neurons.
Which key neuron is chosen? The one that, when it is stimulated, gives rise to an
attractor pattern maximally similar to A.
The ultimate result of this is that, in addition to the distributed memory of
attractors in the Hopfield net, one has a set of key neurons that in effect index the
attractors. Each attractor corresponds to a single key neuron. In the glocal memory
model, the key neurons are the keys and the Hopfield net attractors are the maps.
This algorithm has been tested in sparse Hopfield nets, using both standard
Hopfield net learning rules and Storkey’s modified palimpsest learning rule (Storkey
and Valabregue, 1999), which provides greater memory capacity in a continuous
learning context. The use of key neurons turns out to slightly increase Hopfield net
memory capacity, but this isn’t the main point. The main point is that one now has a
local representation of each global memory, so that if one wants to create a link
between the memory and something else, it’s extremely easy to do so – one just needs
to link to the corresponding key neuron. Or, rather, one of the corresponding key
neurons: depending on how many key neurons are allocated, one might end up with a
number of key neurons corresponding to each memory, not just one.
In spite of their considerable theoretical power, Hopfield nets are not particularly
useful for practical applications on von Neumann computer hardware (appropriately
inexpensive massively parallel computer hardware would be another story, but that’s
not the direction the computer industry has taken), so the above-described
experiments with glocal Hopfield nets were conducted with a view toward intellectual
exploration – in order to understand the possible nature of glocal memory in the brain
via a concrete computational model; and in order to provide a simple prototype
domain for experimenting with related ideas in the more complex context of

---

integrative AGI systems such as those discussed in the following section.
5 Glocal Memory in Integrative AGI Systems
One of the main motivations for the development of the glocal memory concept
has been the design of artificial memories, which is a task different in many ways
from the analysis of modeling of naturally occurring memories. In our work on the
Novamente Cognition Engine (Goertzel, 1996) and OpenCog (Goertzel, 2008) AI
systems, my colleagues and I have been motivated by the glocal memory concept to
design memory approaches that are explicitly glocal in nature.
The glocality concept hits straight at the center of one of the biggest debates of
theoretical AI: symbolic versus subsymbolic knowledge representation. This
dichotomy is often discussed but rarely drawn in a formal and rigorous way, and I
have argued elsewhere that it is actually a largely bogus dichotomy (Goertzel et al,
2008). Traditionally, logic-based AI systems are viewed as “symbolic”, and neural
net systems are viewed as “subsymbolic.” But this distinction has gotten fuzzier and
fuzzier in recent years, with developments such as

•
logic-based systems being used to control embodied agents (hence using
logical terms to deal with data that is apparently perception or actuation-
oriented in nature, rather than being symbolic in the semiotic sense), see
(Santore and Shapiro, 2003; Goertzel et al, 2008)
•
hybrid systems combining neural net and logical parts, or using logical or
neural net components interchangeably in the same role (Lebiere and
Anderson, in preparation)
•
neural net systems being used for strongly symbolic tasks such as
automated grammar learning (Elman, 1991 plus a great deal of more
recent work)

In my own AI systems referenced above, I have explicitly sought to span the
symbolic/subsymbolic pseudo-dichotomy, via creating integrative systems that
combine logic-based aspects with neural-net-like aspects, not in the manner of
multimodular systems, but via attaching uncertain-logical truth values and neural-net-
like weight and activation values to the same nodes and links in a knowledge-
representation hypergraph. Furthermore, both the logical and neural-net-like features
are used to handle all sorts of knowledge, from the most concrete perception and
actuation related knowledge to the most abstract relationships. The concept of
glocality lies at the heart of this combination, in a way that spans the pseudo-
dichotomy:

•
Local knowledge is represented in abstract logical relationships stored in
explicit logical form, and also in Hebbian-type associations between
nodes and links
•
Global knowledge is represented in large-scale patterns of node and link
weights, which lead to large-scale patterns of network activity, which

---