Representation of movement in near extra-personal space in macaque area VIP

1
Intelligence, Cognition, and Adaptive Behavior
2
The Interdisciplinary Nature of Computational
Neuroscience
3
Intelligence, Cognition, and Adaptive Behavior
Is something missing?
4
Warm up
What are the main characteristics of
intelligence and cognition? Any suggestions?
5
What is intelligence?

• Mainly assumed characteristics
• thinking
• logic reasoning
• problem solving

• What is about
• drinking a glass of water?
• recognizing friends?
• preparing a meal?

6
What is intelligence?

• Is a bacteria intelligent?
• Are ants intelligent?
• Ant colonies?
• Mice?
• Cats?
• Babies?
• Machines?

7
What is intelligence?

• Intelligence is a descriptive term.
• Descriptive terms are rather arbitrary (and
  usually relative to your perspective).
• Are there common denominators?

8
Intelligence What is it for?

• Our brain hasn't evolved to solve mathematical
  proofs!
• It controls our behavior to ensure our survival.
• Natural intelligence is always manifested in
  behavior.
• We have to understand behavior and its underlying
  mechanisms to understand intelligence.
• To behave a body is needed -gt embodied
  intelligence.

9
Intelligence Definitions
Are there general definitions about intelligence?
(unfortunately not) Let's have a look in the
history

• Journal of Educational Psychology (1921) asked 14
  experts for definitions of intelligence, and it
  got 14 different answers
• abstract thinking (Terman)
• ability of learning to adjust yourself to the
  environment (Colvin)
• a biological mechanism by which the effects of a
  complexity of stimuli are brought together and
  given a somewhat unified effect in behavior
  (Peterson)
• the capacity to acquire capacity (Woodrow)
• the capacity to learn or to profit by experience
  (Dearborn)

10
Intelligence Commonsense notions

• Thinking and problem solving
• Learning and memory
• Language
• Intuition and creativity
• Consciousness
• Emotions
• Surviving in a complex world
• Perceptual and motor abilities

11
Intelligence Commonsense notions

• Thinking and problem solving
• We never know whether another agent (human,
  animal, robot) is thinking or not!
• We can only speculate about it.
• Learning and memory
• Just memorizing facts out of context is
  pointless!
• Memory for useful knowledge is important.
• Transfer of knowledge, not merely storing it.
• Learning per se is not intelligent, but learning
  to learn!

12
Intelligence Commonsense notions

• Language
• Often seen as the hall mark of intelligence.
• And humans are outstanding in that.
• Combines good learning and memory capacity.
• Abilities many animals have, too!
• Intuition and creativity
• go beyond thinking requires engagement of
  emotions
• intuition come to conclusions without a tracable
  chain of logical thoughts
• creativity
• relative to a particular society's value
• often seen as the highest form of human
  intelligence
• combination of consciousness thought and
  unconsciousness processes
• curiosity? (animals have that)

13
Intelligence Commonsense notions

• Consciousness
• hard to define what it is all about
• required by thinking, language, and creativity
• essential for intelligence???
• Emotions
• richness of emotions are claimed to depend on
  intelligence
• emotional intelligence?

14
Intelligence Commonsense notions

• Surviving in a complex world
• is fulfilled by every living being in the world
• for instance
• social behavior (from bacteria to humans)
• use of tools (birds, primates, etc.)
• Perceptual and motor abilities
• mistakenly often not assumed to be essential for
  intelligence
• What's about
• recognizing kin (e.g. faces)?
• walking on two legs, playing tennis, constructing
  a spider web?

15
Intelligence Can we measure intelligence?
16
Intelligence How useful is an IQ test to
measure intelligence?

• In fact, it is problematic to reduce a complex
  issue as intelligence to a single number!
• For instance, Gardner (1985) proposed a theory of
  multiple intelligence or multiple competences
• linguistic intelligence,
• musical intelligence,
• spatial intelligence,
• bodily-kinesthetic intelligence,
• personal (emotional) intelligence.
• most of those competences can
• not be measured with standard tests,
• not simply mapped onto one dimension.

17
Intelligence A matter of brain size?
508 meters high 282 times the average height of a
man
2,50 meters high / deep 250 times the average
length of an ant
18
Intelligence A matter of brain size?

• Ant brain
• 250,000 neurons
• So, 40,000 ants have the same number of neurons
  as a single human brain

19
Intelligence A matter of brain size?
Myxomycota (slime molds)
No brain and no neurons at all!
20
What are the origins of intelligence?
21
Intelligence The nature-nurture debate

• Is intelligence largely innate, genetically
  predetermined?
• If so, it has to be coded in the genes!
• BUT genes interact with their environment on all
  levels
• "There is virtually no interesting aspect of
  development that is strictly 'genetic'" (Elman et
  al., 1996)
• Solution to the question about the origin of
  intelligence presumably lies in the interaction
  between
• genetic factors (nature) and
• environmental factors (nurture).
• This leads to a new scientific quest
• How does development actually works?
• In which precise way do genetic and environmental
  factors interact?

22
Is there a common denominator for intelligence?
23
Intelligence The diversity-compliance trade-off

• What are the mechanisms enabling organisms to
  adapt to, cope with, environmental changes?
• The dual meaning of adaptivity
• complying with existing rules (compliance
  conservative component)
• rules of information, social rules, rules
  grammar, laws of nature and physics
• generating new behavior (diversity innovative
  component)
• This characterization applies to all levels
• a mathematician's abstract thinking abilities
• a child's conversation to its parents
• an animal's escape from a predator
• Note and keep in mind Terms like adaptation,
  behavior, and generation of behavioral diversity
  imply the existence of a body interacting with an
  environment!

24
Intelligence The diversity-compliance trade-off
Known from psychology (Piaget, 1952) as
Assimilation Accommodation
25
Intelligence
Is it all about information processing?
26
Intelligence The information processing metaphor

• Intelligence as an input-processing-output cycle
  (open-loop systems!)

• Focuses mainly on
• thinking
• abstract reasoning
• abstract problem solving

Wait a second, what have we said about
intelligence?
27
Intelligence What is it for?

• Our brain hasn't evolved to solve mathematical
  proofs!
• It controls our behavior to ensure our survival.
• Natural intelligence is always manifested in
  behavior.
• We have to understand behavior and its underlying
  mechanisms to understand intelligence.
• To behave a body is needed -gt embodied
  intelligence.

28
Intelligence The information processing metaphor

• Intelligence as an input-processing-output cycle
  (open-loop systems!)
• with this view of intelligence it proved that it
  is extremely difficult to get machines do even
  the simplest jobs, like
• moving around
• picking up objects
• bringing them to designated locations
• What's the matter with machines?
• They are an excellent indicator for our current
  understanding of natural phenomena.
• For instance, nowadays we can build an artificial
  heart because we fully understood its function
  and the mechanisms which underlie that functional
  behavior.
• Have you ever seen a machine, where you thought
  "Wow that machine behaves almost as intelligent
  as a baby, a cat, an ant, a bacteria"? Honestly!

29
What is intelligence?

• Is a bacteria intelligent?
• Are ants intelligent?
• Ant colonies?
• Mice?
• Cats?
• Babies?
• Machines?

30
Intelligence The information processing metaphor

• Intelligence as an input-processing-output cycle
  (open-loop systems!)
• with this view of intelligence it proved
  extremely difficult to get machines do even the
  simplest job, like
• moving around
• picking up objects
• bringing them to designated locations
• Brook's (1986) subsumption architecture
• intelligent behavior by using many loosely
  coupled process, functioning asynchronous and in
  parallel
• minimal internal processing is required, if we
  consider an intelligent systems as tightly
  coupled to the environment (closed-loop systems!)
• in this perspective, an agent has a body,
  sensors, motors
• it is embodied and autonomous!

31
We'll skip Brook's (1986) approach, even though
it was a radically new perspective on intelligent
behavior at this time. Instead, we'll take a
look to the history in science, probably there
already is something similar interesting? But
before, let's talk about embodiment, autonomy,
and situatedness.
32
Autonomy

• Loose definition independence from human
  (external) control (robotics).
• More technical
• A non-autonomous system some parameters or
  constraints are independent functions of time.
• e.g., systems driven by some external factor
• Otherwise it is autonomous

33
Biological autonomy
Autonomous systems are organizationally
closed "That is, their organization is
characterized by processes such that (1) the
processes are related as a network, so that they
recursively depend on each other in the
generation and realization of the processes
themselves, and (2) they constitute the system as
a unity recognizable in the space (domain) in
which the processes exist." (F. Varela,
1979) We'll come back to that later when we talk
about Autopoiesis.
34
Embodiment and Situatedness
Embodiment in short The behavior of an agent
arises from the physiology of its brain and body
with which it can perceive and act in its
environment. "In everyday life you usually
remember your "place" largely because the
external world is there to remind you what you
have or haven't done. For instance, you can check
up whether you have already added the vanilla
essence by sniffing or tasting the mixture, or
perhaps by referring to the pencil and paper
representation of the culinary task that you have
drawn up for this mnemonic purpose. A computation
system that solves its problems "in its head"
rather than by perceiving and acting in the real
world, or pencil and paper models of it, has to
have all its memory aids in the form of internal
representations." (M. Boden, 1977)
35
Embodiment and Situatedness

• Embodiment in short The behavior of an agent
  arises from the physiology of its brain and body
  with which it can perceive and act in its
  environment.
• Situatedness
• closely related to embodiment
• a situated agent can acquire information about
  the world and its current state through its
  sensors and own actions.
• the agent's world is its ecological niche!
• important for the perspective issue of the
  frame-of-reference problem

36
Looking back
Now to the history in science. Something
interesting there?
37
Cybernetics

• Cybernetics started in the early 1940s!
• and brought together different fields of
  research
• mathematics (Wiener, von Neumann)
• neuroscience (Rosenblueth, McCulloch)
• social science (Bateson, Mead)
• engineering (Shannon)
• psychology (Frank, Lewin)
• ...

38
Cybernetics

• Greek ??ße???t?? (kybernetes, steersman,
  governor, pilot)
• Aimed at unifying principles.
• Introduced, on abstract levels, concepts as
• control,
• feedback,
• signal
• information
• logic networks
• ...
• If you think you have a great idea, go in the
  library and check the literature on cybernetics
  first. You may wonder...
• Are you searching for interesting questions
  nobody asks?
• Check the cybernetics literature first!

39
Cybernetics Norbert Wiener

• 1894 1964
• mathematician
• published 1948 Cybernetics or Control and
  Communication in the Animal and Machine.
• "The best model of a cat is another or,
  preferably, the same cat." .

40
Cybernetics Norbert Wiener

• introduced the idea of feedback

sensors
nervous system
actuators
sensors
nervous system
actuators
feedback
41
Cybernetics Norbert Wiener

• sensing and acting are tightly coupled
• controversial to the standard view of the time
  (and it is not quite standard even today)

sensors
nervous system
actuators
feedback
42
Cybernetics Grey Walter

• 1910 - 1977
• neurophysiologist and roboticist
• nowadays famous for his tortoise like robots
  Elsie and Elmer

43
Cybernetics Grey Walter's "Machina Speculatrix"
Elsie, and its circuit.

• two sensors, two motors
• a complete physical agent that exploits
  interactions with the environment.

44
Cybernetics Grey Walter
1950s!!!

• behavior integration
• avoid obstacles and
• tropism to light

45
Cybernetics Grey Walter's Elsie

• Autonomy
• recharged its own battery

1950s!!!
46
Cybernetics Grey Walter's Elsie

• Decision making / action selection mechanisms
• choose between alternatives

1950s!!!
47
Cybernetics Grey Walter's Elsie
narcissism
1950s!!!
48
Cybernetics Grey Walter's Elsie and Elmer
love and hunger
1950s!!!
49
Cybernetics It's children

• artificial intelligence
• robotics
• control theory
• complex systems
• neural networks, cellular automaton, genetic
  algorithms
• self-organization
• order from noise
• ultra-stability
• autopoiesis
• information theory
• Most of those are essential parts of modern
  trends of research like, for instance, the
  dynamical systems approach to behavior and
  cognition.

50
The frame of reference problem
51
The frame-of-reference problem
It's probably the most FUNDAMENTAL
PROBLEM!!! Whenever you want to understand an
intelligent / adaptive system (natural or
artifical) never, really never, underestimate the
frame-of-reference problem! It concerns the
relation between the observer (that's probably
you), the designer / modeler (maybe you as well),
the artifact, the environment, and the observed
agent.
52
The frame-of-reference problem

• The main aspects (Clancey, 1991)
• Perspective
• observer's vs. agent's perspective
• descriptions of behavior from an observer's
  perspective must not be taken as the internal
  mechanism underlying the described behavior
• Behavior vs. mechanism
• behavior always results from system-environment
  interactions.
• not explainable by the internal mechanism only.
• Complexity
• the observable complexity of a particular
  behavior does not necessarily reflect the
  complexity of the underlying mechanisms.

53
The frame-of-reference problem

• An example

54
The frame-of-reference problem
Herbert Simon's (1969) ant on the beach
starting point
nest
55
The frame-of-reference problem Perspective
Observer's perspective.
starting point
puddles
pebbles
rocks
nest
56
The frame-of-reference problem Perspective

• Ant's perspective is different.
• because of an entirely different embodiement
  (sensors, brain, body)

starting point
puddles
pebbles
rocks
nest
57
The frame-of-reference problem Behavior vs.
mechanism

• Possible mechanism maybe thought by an observer
• the complex path is stored in the brain, and then
  guide the ant to the nest

starting point
nest
58
The frame-of-reference problem Behavior vs.
mechanism

• The actual mechanism is much simpler
• "if obstacle sensor to the right is active, turn
  left" and vice versa

starting point
nest
59
The frame-of-reference problem Complexity
The seemingly complex trajectory results from
simple behavioral mechanisms.
starting point
nest
60
The frame-of-reference problem Complexity

• The seemingly complex trajectory results from
  simple behavioral mechanisms.
• the complexity of the environment is a
  prerequisite for the ant's behavioral complexity.
• We need to take internal mechanisms, the
  environment, and the interactions between them
  into account.
• Behavior cannot be explained by the internal
  rules alone.

61
The frame-of-reference problem Complexity
Let's make the ant bigger.
62
The frame-of-reference problem Complexity
Let's make the ant bigger. How would the
trajectory look like, now?
63
The frame-of-reference problem Complexity

• The seemingly complex trajectory results from
  simple behavioral mechanisms.
• the complexity of the environment is a
  prerequisite for the ant's behavioral complexity.
• We need to take internal mechanisms, the
  environment and their interactions into account
• Behavior cannot be explained by the internal
  rules alone.
• That is a crucial aspect and it has to be
  carefully considered in building up models of
  biological systems!!!

64
The frame-of-reference problem

• Is that something new?
• to be honest unfortunately to many people it
  still is
• but, actually, it is not new!
• Guess what
• Look to the good old cybernetics!
• Remember Elsie and Elmer?

65
The frame-of-reference problem
hungry?
irresolute?
in love?
1950s!!!
narcissistic?
66
Now, a bit more concrete examples from the
biology.
67
Place cells in mice
for more details see work of Eric Kandels group
at Columbia University
68
Place cells in mice
for more details see work of Eric Kandels group
at Columbia University
69
Place cells in mice
for more details see work of Eric Kandels group
at Columbia University
70
Perceptual development of kittens
Held and Hein, 1967.
71
Perceptual development of kittens

• Held and Hein, 1967.
• visual development depends not only on the
  movement of the body relative to the environment,
  but also on self-actuated movement.
• The behavior, how the kitten perceives the world
  and react to it, does heavily depend on previous
  sensory-motor experiences.

.
72
Sensory motor experience How about you, how do
you perceive the world around you? We are in the
realm of science. Thus, let's do an experiment
(or two)!
.
73
Aplysia
74
Aplysia
75
Aplysia
76
Aplysia
for more details see work of Eric Kandels group
at Columbia University
77
Aplysia
for more details see work of Eric Kandels group
at Columbia University
78
Aplysia

• a perfect system for a reductionist approach
• every individual shows the identical nervous
  system (NS) structure (at least the same amount
  of neurons)
• BUT the strength and structure of single
  synaptic connections is determined by
  environmental contingencies!
• Usually the environment is not static! It is
  dynamic, and unfortunately mostly non-linear.
• Thus, the behavior of a single individual cannot
  solely be explained by the general nervous system
  structure.
• Environmental interactions permanently change the
  NS's structure.
• As listening to this lecture (hopefully) changes
  something in the structure of your brain!
• ... and of course in mine as well!

79
Reductionism, structural changes and couplings,
autopoiesis. If you like, forget everything said
before (except the frame of reference problem, of
course), but take home what comes now!
80
Some remarks to reductionisms

• Scientific reductionism
• There is no clear definition, but let's take that
  one
• When As reduce to Bs, As are nothing but Bs
  (e.g., brains and neurons)
• If I can explain how every single subcomponent of
  a system works,
• I can explain the behavior of the system as a
  whole.
• Where does it come from?
• Descartes, 1637, Discourses.
• animals are like machines

81
Some remarks to reductionisms

• Scientific reductionism
• There is no clear definition, but let's take that
  one
• When As reduce to Bs, As are nothing but Bs
  (e.g., brains and neurons)
• If I can explain how every single subcomponent of
  a system works,
• I can explain the behavior of the system as a
  whole.
• Is it like that?
• Think about the wetness of water
• it clearly is a property of water
• but it clearly is not a property of a H20
  molecule!
• Think about your grandmother
• sure, in your brain specific neurons fire when
  you do that
• but it is not the property of a single neuron
  that let you make up a mental picture of your
  grandmother!
• There is no mystical "above and beyond" thing!
• Nevertheless, both cannot be explained solely by
  the properties of the system's subcomponents.

82
Some remarks to reductionisms

• However, we have to be careful
• One can explain the wetness of water, if one
  knows everything about the molecules and how they
  interact with each other and their environment.
• that's easy (nowadays)!
• One could explain the mechanisms underlying your
  generation of a mental picture of your
  grandmother, if one knows
• everything about all of your neurons,
• how your neurons interact with each other, and
• how that is influenced by your (past, current)
  interaction with the environment.
• that's much more difficult, and far away from
  being disentangled (on every level!)

83
Some remarks to reductionisms

• So, where is the problem?
• I can explain the wetness of water because I know
  the underlying physical laws acting on its
  components, the molecules.
• Surely true, but here comes the real problem
• the water by itself cannot change the interaction
  between the molecules and their interactions with
  the environment.
• And that's exactly why it is
• not a living being
• not intelligent
• not adaptive
• not able to act cognitively
• THAT IS THE CRUCIAL DIFFERENCE!

84
We are not like machines!

• A popular analogy of many 'behavioralists'
• If you want to entirely understand a car engine,
• you can not do so, if you don't put it in a car
  and see how the car behaves when you change
  certain aspects of the engine.
• Surely true, but here comes the important
  warning
• We, and every organism, are not like a mechanic
  duck or a car!

85
We are not like machines!

• You, me, and every living being 'is able' to
  change the way of interactions on almost every
  level
• the sensitivity of sensory elements
• how your neurons are connected
• how you learn certain things, forget other things
• how you play tennis or piano
• how you judge the lecturer from lecture to
  lecture
• how you socialize with other living entities
• On the one hand, that is the foundation of
  intelligence and cognition!
• On the other hand, this makes the whole thing
  difficult to understand and probably impossible
  to do so in a straight forward way.
• However, actually there is nothing mystical about
  it, it's 'just' tremendously complex!

86
A taste of complexity
E.g., part of the metabolic system of E. coli.
When there is cognition, then it is here!
and somewhere over there is cognition
87
A taste of complexity

• The bad news are
• Whenever we strive to fully understand
  intelligent or cognitive behavior of any
  biological system (humans as well as bacteria),
  we have to understand the system of investigation
  as a whole,
• which, unfortunately, dynamically changes its own
  intrinsic structure and the way how it interacts
  with its environment.
• And even worse, the environment is full of such
  permanently changing entities (for instance, your
  seat neighbors or me).
• Two approaches are currently debated
• the building block approach (bottom up)
• figure out the function of each subcomponent
  subsequently and then how they macroscopically
  work together (e.g., "hierarchical reductionism",
  Dawkins, 1986)
• a lot of work
• currently seems to be the most promising and most
  accepted method.
• finding a unified theory
• would be really nice,
• but where to start?

88
A very short introduction to autopoiesis!
Searching for a unified theory. Looking back.
89
Autopoiesis
from the Greek auto - a?t? for self- and
poiesis - p???s?? for creation or production
Humberto Maturana Francisco
Varela
90
Autopoiesis (Maturana, Varela)

• An autopietic system is organized as "a network
  of processes of production (transformation and
  destruction) of components that produces the
  components which..."
• "...through their interactions and
  transformations continuously regenerate and
  realise the network of processes (relations) that
  produced them and"
• "... constitute it ... as a concrete unity in
  the space in which they (the components) exist by
  specifying the topological domain of its
  realisation as such a network."

input
determines
bounded system
molecular components
Adaptation Viability
generates
output
produces
reaction network
91
Autopoiesis (Maturana, Varela)

• Applied to the nervous system
• Neuronal activity depends only on the structure
  of the whole nervous system and its current state
  of activity
• sensory inputs only modulate (e.g. by induced
  structural changes) neuronal activity
• the nervous system interacts with the rest of the
  body (e.g. by hormones, muscle activation)

92
Autopoiesis (Maturana, Varela)

• Applied to social behavior
• Structural coupling organism and environment
  mutually perturb each other without loosing
  autopoiesis (autonomy)
• Coordination Due to the interactions coherent
  behavior arises, which is not immediately obvious
  from the interactive patterns.
• Structural congruence Perturbations induce
  plastic changes
• if chances for further interactions reduce,
  coordination will disrupt
• if these chances enhance, repeated coordination
  will follow.
• after some time, coordination is facilitated
  because the structures of both organisms have
  changed towards a congruent state

93
Take home message(s)

• Intelligence must always be seen with respect to
  a particular ecological niche.
• Intelligence is always manifested in behavior!
• Intelligent and adaptive behavior can be
  characterized by how efficient it solves the
  compliance-diversity trade-off.
• Cognition Life (!?)
• Don't forget to look in the cybernetics
  literature from time to time.
• Always think about the frame-of-reference
  problem!
• A living organism is fundamentally different to
  every machine that exists today! (Autopoiesis)

94
Further reading

• Ashby, W. (1960), Design for a brain. Chapman.
• Clancey, W.J. (1991), The frame of reference
  problem in the design of intelligent machines.
  In Architectures for intelligence.
• Camazine, S. et al. (2001), Biological
  self-organization. Princeton university press.
• Clark, A. (1996) Being there Putting brain,
  body, and world together again. MIT press.
• Haken, H. (1982), Synergetik. Springer.
• Kandel, E. et al. (2000), Principles of neural
  science. McGraw-Hill medical.
• Kelso, J. (1995), Dynamic patterns The
  self-organization of brain and behavior. MIT
  press.
• Maturana, H. and Varela, F. (1992), The tree of
  knowledge. Shambala.
• Pfeifer, R. and Scheier, C. (1999), Understanding
  intelligence. MIT press.
• Searle, J. (2004), Mind A brief introduction.
  Oxford University press.
• Simon, H. A. (1969), The science of the
  artificial. MIT Press.
• Varela, F. (1979), Principles of biological
  autonomy. North-Holland.
• Walter, G. (1963), The living brain. W. W.
  Norton.
• Wiener, N. (1948), Cybernetics or Control and
  Communication in the Animal and Machine. MIT
  press.

95
Adaptivity What is an adaptive systems?

• A system, natural or artificial, that changes due
  to perturbations (e.g. changes in the
  environment).
• It aims at maintaining a invariant (e.g.
  survival).
• It do so
• by altering its properties (behavior, structure)
• or by modifying its environment.