This page describes the organizational forces that limit change.
It explains how to overcome
them when necessary.
This page uses an example to illustrate how:
- A business can gain focus from targeting key customers,
- Business planning activities performed by the whole organization can build awareness, empowerment and coherence.
- A program approach can ensure strategic alignment.
This page uses the example of HP's printer organization freeing
itself from its organizational constraints to sell a printer
targeted at the IBM pc user.
The constraints are described.
The techniques to overcome them are implied.
The constraints are described.
The techniques to overcome them are implied.
Associatively integrated robots
Summary
Haikonen juxtaposes the philosophy and psychology of consciousness with engineering practice to refine the debate on the hard problem of consciousness. During the journey he describes the architecture of a robot that highlights the potential and challenges of associative neural networks.Complex adaptive system (CAS) theory is then used to illustrate the additional requirements and constraints of self-assembling evolved conscious animals. It will be seen that Haikonen's neural architecture,
This page describes the Adaptive Web framework (AWF) test system
and the agent programming framework (Smiley) that supports its
operation.
Example test system statements are included. To begin a test a test statement is loaded into Smiley while Smiley executes on the Perl interpreter.
Part of Smiley's Perl code focused on setting up the infrastructure is included bellow.
The setup includes:
Smiley's Copycat
architecture and molecular biology's intracellular
architecture leverage the same associative properties. Example test system statements are included. To begin a test a test statement is loaded into Smiley while Smiley executes on the Perl interpreter.
Part of Smiley's Perl code focused on setting up the infrastructure is included bellow.
The setup includes:
- Loading the 'Meta file'
specification,
- Initializing the Slipnet,
and Workspaces and loading them
- So that the Coderack can be called.
Consciousness and Robot Sentience
In Pentti O Haikonen's book 'Consciousness and Robot Sentience' he describes his experiments with building robots which are operated by motors integrated with sensors by associative neural networks.Haikonen argues that machines do not understand because they do not operate with meaning. Meaning he asserts, comes from interacting with the everyday environment and learning how things are. We see and understand the environment directly and readily interact with it - because we are conscious:
- Imitating consciousness is not enough
- Philosophers and psychologists have not explained
consciousness. Haikonen raises the mind/body problem is an assumed separation of the mind and body. It has a long history. Descartes's Cartesian dualism assumes the mind and body are two clocks in synchrony but otherwise unrelated. John Locke commented 'It is impossible to conceive that matter, either with or without motion, could have, originally, in and from itself, sense, perception, and knowledge; as is evident from hence, that then sense, perception, and knowledge, must be a property eternally separable from matter and every particle of it.' Chalmers describes this explanatory gap as the hard problem of consciousness.
. Haikonen's hypothesis is that consciousness is a system property. It is neither energy nor matter and can't be explained by physical laws about energy and matter.
Haikonen suggests there is only one real problem of consciousness 'the hard problem' - related to qualia are the direct qualities of percepts according to Haikonen. He argues they do not require interpretation or any evocation of meaning. Colors are colors and pain is pain. The human visual hierarchy seems at odds with this interpretation with meaning being associated with letters by signalling from the letterbox to the frontal lobes and used in the feedback flows that identify and prime morphemes. based perception which he writes is a mandatory precondition of human-like consciousness - which is obtaining the required combination of symbolic and sub-symbolic information processing. Associative information processing with associative neuronal networks and distributed signal representations inherently facilitates the natural transition from sub-symbolic to symbolic.
Sentient robots must also have information integration and sensory motor integration. That Haikonen argues is achieved through architecture with the Haikonen Cognitive is the ability to orchestrate thought and action in accordance with internal goals according to Princeton's Jonathan Cohen.
Architecture (HCA). He compares this architecture to Baar's global workspace is an architecture for consciousness based on a network of neural assemblies performing on a theater stage for an unconscious audience. It is a type of blackboard architecture.
and Shanahan's global workspace is an architecture for consciousness which leverages Baars model but avoids the homunculus issues of a theater. Instead the workspace is a communication infrastructure that interconnects the: sensory cortex, motor cortex, affect, working memory and episodic memory.
.
Haikonen aims to use the development of a sentient robot based on the HCA and constructed using Haikonen neurons to answer three questions:
- How does neural activity appear internally as a subjective
experience? Consciousness is the presence of internal
appearances which become percepts are internal appearences of the external world and the body according to Haikonen. RSS views them as evolved models that are:
- Associated schematically with the signals generated in response to epi-phenomena detected by sensory receptors and
- Acted on by emergent agents.
- Associated schematically with the signals generated in response to epi-phenomena detected by sensory receptors and
- How does the subject become aware of its mental
content? Haikonen argues only percepts can have
internal appearance. So thoughts and imagination are
made virtual percepts via association with feedback loops
from actual percepts. Inner speech is viewed as
particularly powerful.
- How does 'I', the self, arise? Perceptions of the
body and its sensations and mental content are not
changing. Haikonen views these unchanging percepts as
the self.
The problem of consciousness
Haikonen frames his experiments with robots by reviewing the problem of consciousness. Can mental programs be executed by neural hardware? If engineers could design such hardware they would know what a mind is. He asserts that the mind is not an execution of any such program. The real problem of consciousness is that the mind emerges from within the brain. Ideas are not the real material things that they represent.Haikonen laments that philosophy contributed to the confusion. Cartesian dualism is an assumed separation of the mind and body. It has a long history. Descartes's Cartesian dualism assumes the mind and body are two clocks in synchrony but otherwise unrelated. John Locke commented 'It is impossible to conceive that matter, either with or without motion, could have, originally, in and from itself, sense, perception, and knowledge; as is evident from hence, that then sense, perception, and knowledge, must be a property eternally separable from matter and every particle of it.' Chalmers describes this explanatory gap as the hard problem of consciousness.
implies that machines can't think. Instead a homunculus appears in the mind! Property dualism rejects the soul of cartesian dualism. Instead it is argued there exists the physical material substance with its material events. Some of these physical aspects induce supervenient emergent mental properties. But property dualism does not explain the emergent mechanism.
's unexplained emergence similarly provides no help with engineering a sentient robot. Identity theory is Frances Crick's astonishing hypothesis that neurons are the substrate of the mind, with no emergent phenomena needed. , troubles Haikonen since it has not been able to identify the top level conscious neuronal, specialized eukaryotic cells include channels which control flows of sodium and potassium ions across the massively extended cell membrane supporting an electro-chemical wave which is then converted into an outgoing chemical signal transmission from synapses which target nearby neuron or muscle cell receptors. Neurons are supported by glial cells. Neurons include a:
- Receptive element - dendrites
- Transmitting element - axon and synaptic terminals
- Highly variable DNA schema using transposons.
Haikonen sees value in Chalmers's hard problem of consciousness - the explanatory gap between physical processes and subjective experience. Haikonen argues there must be a phenomenon, process or system property that causes some neural activity to appear internally as the subjective experience. Finding out the practical details of what this mechanism is, is the core mind-body problem needed to build sentient robots.
Consciousness and subjective experience
Boltuc described three types of consciousness:- H - Hard subjective phenomena of the explanatory
gap such as the qualia are the direct qualities of percepts according to Haikonen. He argues they do not require interpretation or any evocation of meaning. Colors are colors and pain is pain. The human visual hierarchy seems at odds with this interpretation with meaning being associated with letters by signalling from the letterbox to the frontal lobes and used in the feedback flows that identify and prime morphemes.
of a painful amplifies the aggression response of people by interoceptive signalling of brain regions providing social emotions including the PAG projecting to the amygdala; making aggressive people more so and less aggressive people less so. Pain is the main reason people visit the ED in the US.
stimulus.
- P - Functional aspects such as removing your hand from a
painful stimulus
- A - Access is, argues Stanislas Dehaene, when some attended information eventually enters our awareness and becomes reportable to others.
The internal appearance of neural activity can be detected with instrumentation but not with the mind's subjective attention. Instead that presents subjective experience: seen items, heard sounds, smell etc. The mind externalizes the visual and auditory sensory percepts of the world, so they appear as the external world. Internal percepts are externalized. The internal percepts are only available to the subject. Meanings are associated.
Internal appearances do not emerge automatically. Additional conditions are necessary that are critical to the development of conscious robots.
Perception and Qualia
Haikonen asks what a percept is. Perception provides content to the conscious mind. Receptors, sensitive to specific external stimuli, transduce into common forms of neural signals.The signals are pre-processed into forms of representation that allow subsequent content detection and meaning to be associated.
The output of preprocessing is a series of resolved details which must be filtered by attention, correlated with current context, experience and expectations to identify any match, mismatch or novelty in the set of resolved details possibly resulting in a conscious percept.
Recognition, Haikonen argues, is split into two aspects:
- Identifying continuity,
- The ability to evoke memorized experiences relating to the
perceived object.
Haikonen argues that qualia are the direct qualities of percepts according to Haikonen. He argues they do not require interpretation or any evocation of meaning. Colors are colors and pain is pain. The human visual hierarchy seems at odds with this interpretation with meaning being associated with letters by signalling from the letterbox to the frontal lobes and used in the feedback flows that identify and prime morphemes. exist because of Chalmerd's explanatory gap. So he reviews them in detail:
- Qualia are generated by the sensory systems. He
argues they are generated by the sensors in response to
stimuli.
- In general they are not properties of the real
world. But he reasons they have causal connections to
real world qualities.
- Qualia are physical experience. They are not symbolic representations because they do not require any reference to some common knowledge in order to be understood. In the brain they are a direct way of experiencing information.
- Qualia are private and subjective.
- Qualia are the foundation of percepts are internal appearences of the external world and the body according to Haikonen. RSS views them as evolved models that are:
- Associated schematically with the signals generated in response to epi-phenomena detected by sensory receptors and
- Acted on by emergent agents.
- Associated schematically with the signals generated in response to epi-phenomena detected by sensory receptors and
- Qualia of different senses have different qualities. Haikonen notes the neural representation is positional and associated with the target neurons of the sensory signals as well.
- Amodal
qualia represent details that are shared across sensory
modalities. These aspects may be sharable with
others. Haikonen uses the example of rhythm.
Haikonen notes amodal features offer robot builders natural
sensorimotor integration.
From Perception to Consciousness
Haikonen suggests introspection justifies the claim of percepts are internal appearences of the external world and the body according to Haikonen. RSS views them as evolved models that are:- Associated schematically with the signals generated in response to epi-phenomena detected by sensory receptors and
- Acted on by emergent agents.
Attention is the focusing of our mental resources onto a specific piece of information. Attention uses valuations assigned to each potential object of thought by the basal ganglia.
is an important filter supporting consciousness. It acts as a pinpointing device allowing correct association via temporal coincidence. Mental selection can be sensory (pain amplifies the aggression response of people by interoceptive signalling of brain regions providing social emotions including the PAG projecting to the amygdala; making aggressive people more so and less aggressive people less so. Pain is the main reason people visit the ED in the US. , pleasure etc.) or internal, based on context, emotional states & motivation. Attention mechanisms first engage a percept, then sustain it and finally disengage. They bring objects into perceptual is a dominant function of the dorsolateral prefrontal cortex and the areas it connects with. Prefrontal neurons implement an active memory continuing to fire after the signal is gone for potentially dozens of seconds from the inferior temporal cortex (multi-sensory integration area) and lower level sensory neurons characterized by Hubel & Weisel, while the short-term memory task continues. If the prefrontal cortex gets distracted the memory is lost from consciousness. Earl Miller argues the prefrontal cortex implements the rules that decide which working memory neurons will fire (Spring 2017).
short-term memory in the brain includes functionally different types: Declarative (episodic and semantic), Implicit, Procedural, Spatial, Temporal, Verbal; Hebb noted that glutamate receptive neurons learn by (NMDA channel based) synaptic strengthening. This strengthening is sustained by subsequent LTP. The non-realtime learning and planning processes operate through consciousness using the working memory structures, and then via sleep, the salient ones are consolidated while the rest are destroyed and garbage collected. from where they can be broadcast and memorized.
Haikonen concludes attention, like qualia, are important for robot consciousness. He speculates that conscious perception allows information integration. It activates connections that allow reporting, recalling, evoking of associated meaning and response generation.
Haikonen explores if consciousness is the cross connection of a global workspace. Many consciousness models assume this but Haikonen asserts that qualia based internal appearance is a key requirement because of the explanatory gap.
A conscious agent differs from a non-conscious one in one respect; the subjective qualia based internal appearance. David Hume commented "I can observe nothing but perception." The contents of consciousness & cognitive is the ability to orchestrate thought and action in accordance with internal goals according to Princeton's Jonathan Cohen.
functions are not explanations of consciousness. They don't cross the explanatory gap. Algorithms are not related to qualia based internal experience. Fitting together neural processes and subjective appearances will explain consciousness.
Emotions and consciousness
Consciousness is the feelings of being alive. Physiological symptoms trigger emotions are low level agents distributed across the brain and body which associate, via the amygdala and rich club hubs, important environmental signals with encoded high speed sensors, and distributed programs of action to model: predict, prioritize guidance signals, select and respond effectively, coherently and rapidly to the initial signal. The majority of emotion centered brain regions interface to the midbrain through the hypothalamus. The most accessible signs of emotions are the hard to control and universal facial expressions. Emotions provide prioritization for conscious access given that an animal has only one body, but possibly many cells, with which to achieve its highest level goals. Because of this emotions clash with group goals and are disparaged by the powerful. Evolutionary psychology argues evolution shaped human emotions during the long period of hunter-gatherer existence in the African savanna. Human emotions are universal and include: Anger, Appreciation of natural beauty, Disgust, Fear, Gratitude, Grief, Guilt, Happiness, Honor, Jealousy, Liking, Love, Rage, Romantic love, Lust for revenge, Passion, Sadness, Self-control, Shame, Sympathy, Surprise; and the sham emotions and distrust induced by reciprocal altruism. . Haikonen is not sure why the physiological symptoms would be useful in a robot.Emotions are internal states of the mind. Haikonen follows James with emotional states being equivalent to perceptions of related body sensations. Percepts gain positive, negative or neutral emotional significance. The physiological routes are well understood. LeDoux describes two routes for emotional stimuli: There is a low, fast, direct route from the thalamus has all the main inputs to the cortex passing through it. It is massively supplied with return innervations from the cortical regions it routes too. It does not stand on the route of the main exits from the cortex. The parafascicular nucleus of the rat thalamus contains relatively high levels of D5 dopamine receptors. For human vision the primary system connects to the neocortex via a small part of the thalamus the LGN. to the amygdala contains > 12 distinct areas: Central, Lateral. It receives simple signals from the lower parts of the brain: pain from the PAG; and abstract complex information from the highest areas: Disgust from the insula cortex. It sends signals to almost every other part of the brain, including to the decision-making circuitry of the frontal lobes. It has high levels of D(1) dopamine receptors. During extreme fear the amygdala drives the hippocampus into fear learning. It outputs directly to subcortical reflexive motor pathways when speed is required. Its central nucleus projects to the BNST. It signals the locus ceruleus. The amygdala:
- Promotes aggression. Stimulating the amygdala promotes rage. It converts anger into aggression and when impaired it impacts the ability to detect angry facial expressions.
- Participates in disgust
- Perceives fear promoting stimuli. In PTSD sufferers the Amygdala overreacts to mildly fearful stimuli and is slow to calm down and the amygdala expands in size over a period of months. Fear is processed by the lateral nucleus which serves as the input from various senses, and the central nucleus which outputs to the brain stem (central grey - freezing, lateral hypothalamus - blood pressure, activates paraventricular hypothalamus => crf -> hormone adjustments).
- Has lots of receptors for and is highly sensitive to glucocorticoids. Stress inhibits the GABA interneurons in the basolateral amygdala (BLA) allowing the excitatory glutamate releasing neurons to excite more.
- Is sensitive to unsettling/uncertain social situations where it promotes anxiety. It is also interested in uncertain but potentially painful situations. The amygdala contributes to social and emotional decision making where the BLA supports rejecting an unacceptable offer by injecting implicit mistrust and vigilance, generating an anger driven rejection that is used as punishment. The amygdala is very rapidly excited by subliminal signals from the thalamus of outgroup skin color. The amygdala subsequently tips social emotions against outgroups unless restrained by the frontal lobe or influenced by subliminal priming to prioritize inclusion. The fast path from the thalamus rapidly but inaccurately signals its identified a weapon.
- Promotes male, but not female, sexual motivation when it is an uncertain potential pleasure.
- Responds to the longing for uncertain potential pleasures and fear that the reward will not be worth it if it happens. The amygdala turns off during orgasm.
- Uses but is not directly involved in vision.
Haikonen argues that the physiological symptoms take the form of emotional qualia are the direct qualities of percepts according to Haikonen. He argues they do not require interpretation or any evocation of meaning. Colors are colors and pain is pain. The human visual hierarchy seems at odds with this interpretation with meaning being associated with letters by signalling from the letterbox to the frontal lobes and used in the feedback flows that identify and prime morphemes. . Physiological states (hunger, thirst, temperature, boredom) are internal causes. Pain amplifies the aggression response of people by interoceptive signalling of brain regions providing social emotions including the PAG projecting to the amygdala; making aggressive people more so and less aggressive people less so. Pain is the main reason people visit the ED in the US. and reward are external causes. Emotions are motivational factors.
Digital computers do not need motivation. They have a scheduler that executes whatever is programmed. But Haikonen asserts an artificial cognitive is the ability to orchestrate thought and action in accordance with internal goals according to Princeton's Jonathan Cohen.
agent is not governed by a program. Attention would not be handled by a program. Another means of action planning is necessary. Motivation would be key.
Haikonen notes that consciousness is typically viewed as providing free will is the subjective assessment of one's ability to make decisions and perform independent actions. Philosophers note that causal chains linking physical phenomena with conscious decisions would undermine the idea of independent free will. RSS views the architecture of CAS agency as requiring indirect associations between phenomena and agent's models. Evolution captures these associations within the genetic structures of the emergent agents, removing any epistemological or complementarity constraints.
. Haikonen argues against free will but looks at determinism and decision making. He sees the primary factor in decision making being the situation which he judges causal. The current state of the agent is a secondary factor. And social pressures and the demands of others can provide external influence. The secondary factors provide the appearance of free will. Haikonen concludes a robot could have a general value system that would help the robot choose what to do.
Inner speech and consciousness
Haikonen has developed an associative model of language based on his view of natural languages as symbol systems with real world grounding of meaning.The meaning of words is grounded in percepts of various modalities and an inner situation model based on real world situations rather than an abstract grammar and text alone. Abstract meanings arise from their associative connections to larger context and their style of use.
Inner speech can use feedback loops to allow qualia are the direct qualities of percepts according to Haikonen. He argues they do not require interpretation or any evocation of meaning. Colors are colors and pain is pain. The human visual hierarchy seems at odds with this interpretation with meaning being associated with letters by signalling from the letterbox to the frontal lobes and used in the feedback flows that identify and prime morphemes. associated with percepts are internal appearences of the external world and the body according to Haikonen. RSS views them as evolved models that are:
- Associated schematically with the signals generated in response to epi-phenomena detected by sensory receptors and
- Acted on by emergent agents.
Qualia and machine consciousness
Haikonen compares human and machine consciousness - arguing that in the human mind the perceived world, and internal content, presents itself as qualia are the direct qualities of percepts according to Haikonen. He argues they do not require interpretation or any evocation of meaning. Colors are colors and pain is pain. The human visual hierarchy seems at odds with this interpretation with meaning being associated with letters by signalling from the letterbox to the frontal lobes and used in the feedback flows that identify and prime morphemes. and that conscious execution by cognitive is the ability to orchestrate thought and action in accordance with internal goals according to Princeton's Jonathan Cohen.functions is driven by qualia-based internal appearances. He asserts this is also a good structure for conscious machines. But the qualia of these machines don't have to be similar to human qualia. Contemporary computers don't have qualia and are not conscious.
Haikonen argues machine qualia need to be:
- Direct, so not represented symbolically
- Perceived in the carried information with the carrier transparent
- Representative of amodal qualities
of features
- Externalized with the association of feedback from exploratory actions.
Testing consciousness
Haikonen identifies requirements for effective tests of consciousness:- They must generate responses to the ongoing situation
- Checks for memory in the brain includes functionally different types: Declarative (episodic and semantic), Implicit, Procedural, Spatial, Temporal, Verbal; Hebb noted that glutamate receptive neurons learn by (NMDA channel based) synaptic strengthening. This strengthening is sustained by subsequent LTP. The non-realtime learning and planning processes operate through consciousness using the working memory structures, and then via sleep, the salient ones are consolidated while the rest are destroyed and garbage collected.
- Checks for attention is the focusing of our mental resources onto a specific piece of information. Attention uses valuations assigned to each potential object of thought by the basal ganglia.
- Checks that the system provides the ability to introspect
- Checks for presence of mental content
- Checks for support of direct perception
- Turing test does not check for qualia are the direct qualities of percepts according to Haikonen. He argues they do not require interpretation or any evocation of meaning. Colors are colors and pain is pain. The human visual hierarchy seems at odds with this interpretation with meaning being associated with letters by signalling from the letterbox to the frontal lobes and used in the feedback flows that identify and prime morphemes. .
- Picture
understanding test is one of Koch & Tonini's tests for consciousness. They argue that human brains integrate information allowing understanding of complex situations. The test compares the response to two pictures: one including a monitor and keyboard, the other a computer monitor with a flower pot where the keyboard should be. If the second picture is not seen as having something wrong with it the observer is not conscious.
asserts that conscious humans or machines should understand patterns not just detect them. Pattern recognition is not consciousness. But Haikonen also argues that the test is not rigorous enough missing some of his requirements. - Cross
examination test for consciousness asks a human or robot about its flows of inner imagery and inner speech based on an assumed ability to introspect and report mental content and then cross-examines the responses. Haikonen proposes the questioning focus on:
- Does the machine have a flow of mental content that is about something?
- Is the machine able to report its mental content to itself and recognize ownership of the content
- Can it describe some qualia?
- Can it remember the immediate past?
- Does it feel pain?
Tests of self-consciousness include:
- Mirror test is passed if the system is aware it is itself
that it sees in the mirror.
- Name test is passed if the system is aware of its
name. But Haikonen warns that animals may just be
associating the reward that often follows it with the name
rather than being self-conscious.
- Ownership test checks for awareness that the system is
responsible for the actions. This is a helpful test
since ownership awareness can often be inferred from
subsequent responses.
- The cross examination test can be applied.
Artificial conscious cognition
Over time various analogies have been used as models of artificial cognition:- Switch board
- Computer
- Hierarchical classifier
- Controller
- Predictor
- Simulator
- Search engine
- Time Machine
Haikonen reviews cognitive is the ability to orchestrate thought and action in accordance with internal goals according to Princeton's Jonathan Cohen.
architectures and their capabilities:
- Cognitive architectures depict system arrangements for
producing human-like cognition is the ability to orchestrate thought and action in accordance with internal goals according to Princeton's Jonathan Cohen.
- Combine and implement cognitive functions: Perception,
Attention, Prediction, Learning, Memory in the brain includes functionally different types: Declarative (episodic and semantic), Implicit, Procedural, Spatial, Temporal, Verbal; Hebb noted that glutamate receptive neurons learn by (NMDA channel based) synaptic strengthening. This strengthening is sustained by subsequent LTP. The non-realtime learning and planning processes operate through consciousness using the working memory structures, and then via sleep, the salient ones are consolidated while the rest are destroyed and garbage collected. , Imagination,
Planning, Judgement, Reasoning, General intelligence enables the achievement of goals in the face of obstacles. The goals are sub-goals of genes' survival and reproduction and include:
- Obtaining and eating food
- Sex
- Finding and maintaining shelter
- Fighting for resources - in the preferred hunter gatherer environment loss of resources was critical while possession was often transient.
- Understanding the proximate environment
- Securing the cooperation of others
- Consciousness, self-consciousness, subjective experience
and qualia.
- Produce human-like responses, actions and behavior.
- Know and understand what it is doing
- Should be able to act autonomously in meaningful ways in new situations
- Learn and accumulate experience
architecture:
- High-level symbolic approach - is the traditional
artificial intelligence (A.I.) approach, where cognitive
functions are implemented directly in discrete function
blocks. The blocks are then logically connected
together.
- Low-level sub-symbolic approach - is the connectionist
approach, where artificial neurons and synapses are defined
and the functions of interconnected neurons and neuron
groups are defined to produce a traditional artificial
neural network.
- Associative symbolic neural network approach - is
Haikonen's approach, starts with low level neural components
but uses neurons and neuron groups that are associative,
allowing the association of signals and signal patterns with
each other in a direct way. This allows association of
meaning and symbols and supports the transition from
sub-symbolic to symbolic level.
Associative information processing
Haikonen notes that the challenge for an associative system is stopping everything becoming associated with everything else. He includes a selective mechanism within his associative nodes to constrain association formation.Haikonen reviews three types of association:
- Pavlovian conditioning - builds a simple association
between two signals that are temporarily linked by a signal
such as a reward or threat.
- Hebbian conditioning - builds an association between
neurons that are active together by synaptic
strengthening.
- Auto association and hetero association. These association types allow:
- Sub pattern signals can cause the whole pattern to be
linked to.
- Pavlovian conditioning can have meaning and symbols
associated.
- Temporal associations can form sequences of
patterns.
- A prior temporal sequence can auto associatively
generate a prediction of a later pattern.
- Prior temporal sequences can hetero associate future
different sequences.
Associative information representation
Information can be represented through content detection being associated with meaning by building distributed representations of signals and using learned experience to associate meaning (machine qualia). The distributed representation method will:- Directly tell something about the content
- Allow easy association of meaning
- Allow easy modification and combination of the
representations to enable imagination and creativity.
- Work with imperfect representations
- Tolerate errors and distortion.
Neural realization of associative processing
Haikonen notes that real neurons develop a representation using spiking transmissions. Machine neurons can more easily leverage block signals. And he asserts that simple models of neurons simulating the information model suffice for building information processing logic. To just represent the information flows block signals of varying intensity can be used.Haikonen associative neuron
The Haikonen neuron includes:- Main input signal - meaning is conserved at the
output. In some situations the main signal must also
pass in which case the main input signal is summed to the
synapse outputs and forwarded to the threshold
circuit.
- Output signal
- Associative vector input (one synapse per signal).
It can evoke the output signal dependent on the state of the
synapses.
- Inhibit input
- Synapses that provide coincidence of main signal and
associative input signal to control the output signal.
Synapse includes a coincidence detector (with learning
control input), one bit memory and a synaptic switch.
The associative synapse can execute simple Hebbian learning
based on simultaneous occurrence of the main signal and the
associative signal. With different learning controls
alternative synaptic associations become available:
- Simple Hebbian learning in given groups of neurons takes place only at suitable occasions determined by the focus of attention. This is facilitated by the learning control signal.
- Correlative Hebbian synapse allows for auto and hetero
associative logic.
Associative neuron groups
The Operation of an Associative Neuron Group is used to evoke one signal from many possible ones. In an associative neuron group:- The associative inputs of the neurons in the group are connected together so the same associative input vector appears at the associative inputs of each neuron.
- Each associative input vector may be associated with one
main signal. This main signal may be evoked if an
associative input vector matches closely the original input
vector.
- The output thresholds of the neurons are connected together with WTA. That will allow the output of the main signal evoked by the closest matching associative vector.
- Haikonen recommends that the value of the output threshold
be floating and slightly less than the maximum evocation
strength value.
Auto-associative Memory can be constructed by wiring the main signal vector as the associative vector as well. This will associate the input signal with itself. The auto associative memory utilizes the SOFT-AND which allows the evocation of the whole input vector as the output vector when a small arbitrary subset, even with some distortions, is introduced at the input.
Temporal Sequences are processed by transforming the serial signals into parallel forms so that the signals become available at the same time. Haikonen uses short-term memories and delay lines for this. Associative neuron groups are then used to associate the various delayed temporal parts with a cue vector. The cue vector will evoke the parts in parallel but the delay lines transform them back into a temporal stream.
Designing a cognitive perception system
Haikonen argues a model perception system will be:- Stimuli - phenomena
- Sensor - transduced
- Internal Format - electric signals
- Pre-processing feature detection - Basic meaning, content
detection, requirements from qualia
- Distributed representation - raw percepts
- Feedback combination - Attention, context, prediction,
match/mismatch, introspection applied top down
- Percept signals - Accepted percepts
- Broadcast
A complete cognitive is the ability to orchestrate thought and action in accordance with internal goals according to Princeton's Jonathan Cohen.
system has a number of perception/response feedback loops, which are associatively cross-linked to each other. The input thresholds of the neuron groups determine, which broadcasts, if any, are accepted at the present moment. The groups learn associative connections between the percept signals and the received broadcast signals including temporal sequences. The feedback loop re-circulates and sustains percepts, acting as a short term working memory is a dominant function of the dorsolateral prefrontal cortex and the areas it connects with. Prefrontal neurons implement an active memory continuing to fire after the signal is gone for potentially dozens of seconds from the inferior temporal cortex (multi-sensory integration area) and lower level sensory neurons characterized by Hubel & Weisel, while the short-term memory task continues. If the prefrontal cortex gets distracted the memory is lost from consciousness. Earl Miller argues the prefrontal cortex implements the rules that decide which working memory neurons will fire (Spring 2017).
.
Match, Mismatch and Novelty feedback are resolved at the feedback neuron group comparing the feedback and the sensory feature signal vector. It is necessary for prediction, attention control, search operations, answering yes or no to questions of the form 'is this ??" and for emotions.
Haikonen describes various examples of perception response feedback loops.
The transition to Symbolic Processing
Haikonen considers the inclusion of natural language within a robot to be important for two reasons:- The robot will be far more useful if it can interact with
humans using a shared understanding of language.
- Philosophers consider natural language processing to be a
core aspect of a cognitive is the ability to orchestrate thought and action in accordance with internal goals according to Princeton's Jonathan Cohen.
architecture.
calls for symbolic presentations. Words can be sounded and text visualized.
The vocabulary of a cognitive agent is a set of words with associated meanings conveyed by associative connections. The meanings of the words are grounded to perceived entities and situations of the external world and body. Haikonen concludes there is a requirement of a perceiving system with associative learning and associative processing of information.
Haikonen's strategy is to learn words with point-able simple meanings first. In this process the sub-symbolic signal patterns of the pointed entities and those of the given word are associated with each other, so that afterwards the signal patterns of the word can act as a symbol for the named entity.
Symbols depict entities not inherently related to them. They have representation that extends beyond the direct meaning of the constituting features. The additional meaning is based on convention. It can be included by association.
Associations can integrate meaning and different modalities. Associative cross-connections are used during associative learning to forward percepts from a modality to the others. Simple two way labeling can take place.
Information integration with multiple modules
Haikonen uses the example of 'find cherry' to illustrate the flows between different modules of his cognitive is the ability to orchestrate thought and action in accordance with internal goals according to Princeton's Jonathan Cohen.architecture. Four modules interact; through associative interconnects, iteratively to develop a viable strategy:
- Motor
- Visual color
- Visual pattern
- Linguistic auditory
'Cherry' is of interest to the visual color module obtaining a match with red.
'Cherry' is of interest to the visual pattern module obtaining a match with round.
'Cherry' is of interest to the linguistic auditory module accepting 'cherry'.
'Red' and 'Round' are fed back to the motor module which is able to generate 'find' movements that drive towards the 'red' 'round' cherry.
Sensorimotor Integration
Haikonen explains how he leverages externalization of the appearance of sensory information. It allows seamless sensorimotor integration. Externalization adds the sense of direction and distance to the perceived object. "The perception of the external position of objects within our reach is seamlessly coupled to the neural circuits that control the motion of our hands and body."Since robots use conventional parts, such as electric motors, without associative interfaces they must be specifically integrated. Haikonen describes the feedback control loops that are required. They are arranged in a hierarchy.
Hierarchical Control Loops
Haikonen asserts that control of motor actions is hierarchic. This allows the architecture to overcome the incompatibility between the details of the feedback loops for different perception/response modalities. Intermediate neuron groups can use associative learning to map between the signals and feedback in one modality and another.This structure ensures that proprioceptive is the sense of awareness of the relative position of adjacent body parts. representations can result in actual action by motors integrated with visual percepts allowing seamless sensorimotor integration.
Emotional Significance of Percepts
Haikonen explains how the problem of combinatorial explosion requires the space of choice of stimuli in an associative system to be limited.Attention is the focusing of our mental resources onto a specific piece of information. Attention uses valuations assigned to each potential object of thought by the basal ganglia.
provides both a filter and a mechanism to change context.
Emotional evaluation of percepts are internal appearences of the external world and the body according to Haikonen. RSS views them as evolved models that are:
- Associated schematically with the signals generated in response to epi-phenomena detected by sensory receptors and
- Acted on by emergent agents.
The outline of the Haikonen Cognitive Architecture (HCA)
Haikonen argues the HCA provides the dynamic system for the production of conscious human like cognition is the ability to orchestrate thought and action in accordance with internal goals according to Princeton's Jonathan Cohen..
It includes information flows between:
- Environment includes mechanical self receives effects from
motor actions; sends signals to sensors
- Sensors (Environmental, Self) - receive input from
environment and mechanical self; send signals to perception
and motor actions
- Perception - receives signals from sensors, and
associative feedback from mental process; sends signals to
mental process and motor actions
- Match/mismatch/novelty
- Emotional evaluation
- Mental process - receives signals from perception; sends
associative feedback to perception
- Motor actions - receives signals from sensors and
perception; sends effects to environment sensors and the
mechanical self
- Reflex reaction - between environment, sensors and motor
actions.
- Sub-conscious routines - between environment, sensors,
perception and motor actions.
- Deliberated actions - between environment, sensors,
perception (mental process), and motor actions
- Imagination - between introspective percept (perception)
and associative deliberation (mental process)
- Internal needs
- Pain amplifies the aggression response of people by interoceptive signalling of brain regions providing social emotions including the PAG projecting to the amygdala; making aggressive people more so and less aggressive people less so. Pain is the main reason people visit the ED in the US. & pleasure sensors that signal to the emotional evaluator which signals system reactions
- Touch, visual, auditory, direction sensors which signal feedback neurons which output a percept that is broadcast to neuron groups with associative memory
- Proprioception is the sense of awareness of the relative position of adjacent body parts. which is fed by effectors and signals feedback neurons which output a percept that is broadcast to neuron groups with associative memory
External factors such as an observed opportunity will trigger imagination of execution of the opportunity.
The HCA allows broadcasts to share a coherent view with other modules. But it also includes the facility to form coalitions of modules. This allows independent cooperation of modules to do tasks in parallel. The coalitions of modules form on an ad hoc basis as required by the situation.
The Comparison of Some Cognitive Architectures
Haikonen compares the HCA with Baar's global workspace is an architecture for consciousness based on a network of neural assemblies performing on a theater stage for an unconscious audience. It is a type of blackboard architecture.and Shanahan's global workspace is an architecture for consciousness which leverages Baars model but avoids the homunculus issues of a theater. Instead the workspace is a communication infrastructure that interconnects the: sensory cortex, motor cortex, affect, working memory and episodic memory.
.
Each of the architectures utilizes a number of unconscious specialist modules. Baar's model interacts with them via the 'theater stage'. Shanahan's model uses the global communications infrastructure. In the HCA the modules communicate directly with one another.
The autonomous unconscious modules in Baar's model compete to post information to the global workspace's working memory. A similar competition takes place in the Shanahan model. Shanahan's communication infrastructure has limited bandwidth and constrains the formation of only one coalition of processes at a time. HCA modules broadcast directly to each other. It is the recipient that decides if the information is appropriate. A module may distribute cues that facilitate the broadcast of information it is interested in. HCA allows multiple coalitions to form and operate at the same time. But only one of the tasks may execute in a way that is conscious.
Unlike the other architectures the HCA does not have a global workspace. All the HCA modules operate in the same way independent of whether they are operating in consciousness or not.
Baar and Shanahan models do not explain Chalmers's hard problem of consciousness.
Complex adaptive system
This page introduces the complex adaptive system (CAS) theory
frame. The theory is positioned relative to the natural
sciences. It catalogs the laws and strategies which
underpin the operation of systems that are based on the
interaction of emergent agents.
John Holland's framework for representing complexity is outlined. Links to other key aspects of CAS theory discussed at the site are presented.
(CAS) theory
supports reasoning about consciousness and sentient robotic
systems. John Holland's framework for representing complexity is outlined. Links to other key aspects of CAS theory discussed at the site are presented.
CAS can only
This page discusses the mechanisms and effects of emergence
underpinning any complex adaptive system (CAS). Key research is
reviewed.
emerge when their environment provides
This page discusses the potential of the vast state space which
supports the emergence of complex
adaptive systems (CAS).
Kauffman describes the mechanism by which the system expands
across the space.
many states including various
different potential niches. Haikonen's
robotic meaning corresponds in CAS systems to - The association of environmental This page discusses the physical foundations of complex adaptive systems (CAS). A small set of rules is obeyed. New [epi]phenomena then emerge. Examples are discussed.phenomena with
Plans emerge in complex adaptive systems (CAS) to provide the instructions that agents use to perform actions. The component architecture and structure of the plans is reviewed.schematic sequences and
Plans are interpreted and implemented by agents. This page discusses the properties of agents in a complex adaptive system (CAS).agent based actions to allow
It then presents examples of agents in different CAS. The examples include a computer program where modeling and actions are performed by software agents. These software agents are aggregates.
The participation of agents in flows is introduced and some implications of this are outlined.
This page reviews the potential to benefit from strategy in a complex adaptive system (CAS). The challenges described by Dorner require a careful search of the proximate environment.assessment of the situation,
- Leverage of the agent state and new Agents use sensors to detect events in their environment. This page reviews how these events become signals associated with beneficial responses in a complex adaptive system (CAS). CAS signals emerge from the Darwinian information model. Signals can indicate decision summaries and level of uncertainty.signals to consult the somatic plan to start
- This page describes the specialized codelets that provide life-cycle and checkpoint capabilities for Smiley applications.Prioritizing all the alternate potential chains of actions to respond effectively and
The codelets implement a Shewhart cycle.
The structural schematic nature of the cycle is described.
Transcription factor codelets operate the phase change controls.
How inhibitory agents are integrated into the cycle is described.
An application agent with management and operational roles emerges.
The codelets and supporting functions are included.
- Help This page reviews the implications of selection, variation and heredity in a complex adaptive system (CAS). The mechanism and its emergence are discussed.natural selection in preserving and adding to the plan all that is good.
Haikonen's description of generating a percept is supported by Dehaene's detailed descriptions of how we read a set of words and how the key detailed unconscious signals achieve conscious access to become a perception.
Features are subtle and are difficult for a CAS to identify.
This page reviews the implications of selection, variation and
heredity in a complex adaptive system (CAS).
The mechanism and its emergence are
discussed.
Evolved The agents in complex adaptive
systems (CAS) must model their
environment to respond effectively to it. Samuel
modeling is described as an approach.
learning
uses Plans change in complex adaptive systems (CAS) due to the action of genetic
operations such as mutation, splitting and recombination.
The nature of the operations is described.
genetic algorithm defined The agents in complex adaptive
systems (CAS) must model their
environment to respond effectively to it. Samuel
modeling is described as an approach.
models and experimental activities to
Agents use sensors to detect events in their environment.
This page reviews how these events become signals associated
with beneficial responses in a complex adaptive system (CAS). CAS signals emerge from
the Darwinian information model. Signals can indicate decision summaries and level of
uncertainty.
associate the most effective models
and strategies for the proximate environment, with clusters of
signals that we identify as features. The This page discusses a complex adaptive system (CAS) implementation of a genetic algorithm (GA), Melanie
Mitchell's robot-janitor built as a set of Copycat codelets
integrated using agent-based
programming. The improvement in the operation of the
robots over succeeding generations of applying the GA is graphed.
The CAS that generated, and operated the robot is reviewed, including the implementation details and codelet operational program flow, and the challenges and limitations of this implementation.
The schematic strings which make up the robot's genotype, as well as the signals which are sent to the nucleus of the robot's agents so that the agents can deploy the appropriate response strings (which activate codelets) are listed. The Slipnet configuration required by the system to associate the schematic strings with programmatic forces (codelets) is also listed. The codelets and supporting perl are also listed.
In the conclusion the limitations of the robot-janitor abstraction in studying emergence and creative evolution are discussed and alternative experimental frameworks are proposed. One such, the schematic cell is the subject of a separate page in this web frame.
virtual robot provides a simple
illustration. The CAS that generated, and operated the robot is reviewed, including the implementation details and codelet operational program flow, and the challenges and limitations of this implementation.
The schematic strings which make up the robot's genotype, as well as the signals which are sent to the nucleus of the robot's agents so that the agents can deploy the appropriate response strings (which activate codelets) are listed. The Slipnet configuration required by the system to associate the schematic strings with programmatic forces (codelets) is also listed. The codelets and supporting perl are also listed.
In the conclusion the limitations of the robot-janitor abstraction in studying emergence and creative evolution are discussed and alternative experimental frameworks are proposed. One such, the schematic cell is the subject of a separate page in this web frame.
Haikonen argues that qualia are direct representations of sensory stimuli. CAS theory implies this is unlikely.
This page discusses the mechanisms and effects of emergence
underpinning any complex adaptive system (CAS). Key research is
reviewed.
Emergent
self-assembly is based on a Plans emerge in complex adaptive
systems (CAS) to provide the
instructions that agents use to
perform actions. The component architecture and structure
of the plans is reviewed.
schematic
plan which is shared by every cell in the eukaryotic, a relatively large multi-component cell type from which yeast and multi-celled plants and animals, including humans, is constructed. It contains modules including a nucleus and production functions such as mitochondria. body and
referenced by enzymatic, a protein with a structure which allows it to operate as a chemical catalyst and a control switch. and cellular
Plans are interpreted and implemented by agents. This page
discusses the properties of agents in a complex adaptive system
(CAS).
It then presents examples of agents in different CAS. The examples include a computer program where modeling and actions are performed by software agents. These software agents are aggregates.
The participation of agents in flows is introduced and some implications of this are outlined.
agents to transform It then presents examples of agents in different CAS. The examples include a computer program where modeling and actions are performed by software agents. These software agents are aggregates.
The participation of agents in flows is introduced and some implications of this are outlined.
This page discusses the potential of the vast state space which
supports the emergence of complex
adaptive systems (CAS).
Kauffman describes the mechanism by which the system expands
across the space.
state and Agents use sensors to detect events in their environment.
This page reviews how these events become signals associated
with beneficial responses in a complex adaptive system (CAS). CAS signals emerge from
the Darwinian information model. Signals can indicate decision summaries and level of
uncertainty.
signals
into actions. Indeed the foundation of a CAS depends on
indirect reference as Deacon
describes. Instead CAS theory suggests This page reviews the implications of selection, variation and
heredity in a complex adaptive system (CAS).
The mechanism and its emergence are
discussed.
evolution has ensured that the
selectively beneficial The agents in complex adaptive
systems (CAS) must model their
environment to respond effectively to it. Samuel
modeling is described as an approach.
models that
are qualia are the direct qualities of percepts according to Haikonen. He argues they do not require interpretation or any evocation of meaning. Colors are colors and pain is pain. The human visual hierarchy seems at odds with this interpretation with meaning being associated with letters by signalling from the letterbox to the frontal lobes and used in the feedback flows that identify and prime morphemes. have become
more and more representative of the actual environment they
describe to the agents that use them. Haikonen's idea of a digital computer and its scheduling architecture is widely held but is too constrained. He asserts the computer can not provide the facilities needed for a cognitive agent. But Hofstadter & Mitchell's Copycat architecture provides the equivalent services on a digital computer as Haikonen's hardware architecture, although probably much more slowly. An artificial cognitive is the ability to orchestrate thought and action in accordance with internal goals according to Princeton's Jonathan Cohen.
agent could be built out of Copycat codelets executed stochastically on the Copycat coderack.
Haikonen's robot uses associative and auto-associative neural networks to augment the learning capabilities of classical artificial neural networks are representational models that achieve high performance on difficult pattern recognition problems in vision and speech. But they need specialized training methods such as greedy layerwise pre-training or HF optimization. with action oriented associations. By associating
Agents use sensors to detect events in their environment.
This page reviews how these events become signals associated
with beneficial responses in a complex adaptive system (CAS). CAS signals emerge from
the Darwinian information model. Signals can indicate decision summaries and level of
uncertainty.
sensor and motor data
the robot can develop sophisticated behaviors. The
associative structures appear similar to AWF's This page describes the Adaptive Web framework (AWF) test system
and the agent programming framework (Smiley) that supports its
operation.
Example test system statements are included. To begin a test a test statement is loaded into Smiley while Smiley executes on the Perl interpreter.
Part of Smiley's Perl code focused on setting up the infrastructure is included bellow.
The setup includes:
Smiley associations. Both architectures
have mechanisms to limit combinatorial explosion. Smiley's
evaluator codelets
limit the building of associative structures in the Example test system statements are included. To begin a test a test statement is loaded into Smiley while Smiley executes on the Perl interpreter.
Part of Smiley's Perl code focused on setting up the infrastructure is included bellow.
The setup includes:
- Loading the 'Meta file'
specification,
- Initializing the Slipnet,
and Workspaces and loading them
- So that the Coderack can be called.
This page describes the Copycat
Workspace.
The specialized use of the Workspace by the adaptive web framework's (AWF) Smiley is discussed.
How text and XML are imported into the Smiley Workspace is described.
Telomeric aging of schematic structures is introduced.
The internal data structure used to represent the state of each workspace object is included.
The Workspace infrastructure functions are included.
workspace. The specialized use of the Workspace by the adaptive web framework's (AWF) Smiley is discussed.
How text and XML are imported into the Smiley Workspace is described.
Telomeric aging of schematic structures is introduced.
The internal data structure used to represent the state of each workspace object is included.
The Workspace infrastructure functions are included.
In regard to thinking machines Dawkin's
Richard Dawkin's explores how nature has created implementations
of designs, without any need for planning or design, through the
accumulation of small advantageous changes.
argues that the creation of life was
a hard problem and likely has only occurred once. Deacon Terrence Deacon explores how constraints on dynamic flows can
induce emergent phenomena
which can do real work. He shows how these phenomena are
sustained. The mechanism enables the development of Darwinian competition.
describes a potential mechanism.
Once it had happened the laws of physics and chemistry were
augmented by the laws of CAS. These supported the
emergence of Consciousness has confounded philosophers and scientists for
centuries. Now it is finally being characterized
scientifically. That required a transformation of
approach.
Realizing that consciousness was ill-defined neuroscientist Stanislas Dehaene and others characterized and focused on conscious access.
In the book he outlines the limitations of previous psychological dogma. Instead his use of subjective assessments opened the window to contrast totally unconscious brain activity with those including consciousness.
He describes the research methods. He explains the contribution of new sensors and probes that allowed the psychological findings to be correlated, and causally related to specific neural activity.
He describes the theory of the brain he uses, the 'global neuronal workspace' to position all the experimental details into a whole.
He reviews how both theory and practice support diagnosis and treatment of real world mental illnesses.
The implications of Dehaene's findings for subsequent consciousness research are outlined.
Complex adaptive system (CAS) models of the brain's development and operation introduce constraints which are discussed.
consciousness and Realizing that consciousness was ill-defined neuroscientist Stanislas Dehaene and others characterized and focused on conscious access.
In the book he outlines the limitations of previous psychological dogma. Instead his use of subjective assessments opened the window to contrast totally unconscious brain activity with those including consciousness.
He describes the research methods. He explains the contribution of new sensors and probes that allowed the psychological findings to be correlated, and causally related to specific neural activity.
He describes the theory of the brain he uses, the 'global neuronal workspace' to position all the experimental details into a whole.
He reviews how both theory and practice support diagnosis and treatment of real world mental illnesses.
The implications of Dehaene's findings for subsequent consciousness research are outlined.
Complex adaptive system (CAS) models of the brain's development and operation introduce constraints which are discussed.
Reading and writing present a conundrum. The reader's
brain contains neural networks tuned to reading. With
imaging a written word can be followed as it progresses from the
retina through a functional chain that asks: Are these letters?
What do they look like? Are they a word? What does it sound
like? How is it pronounced? What does it mean? Dehaene
explains the importance of
education in tuning the brain's networks for reading as
well as good strategies for teaching reading and countering dyslexia. But
he notes the reading
networks developed far too recently to have directly evolved.
And Dehaene asks why humans are unique in developing
reading and culture.
He explains the cultural engineering that shaped writing to human vision and the exaptations and neuronal structures that enable and constrain reading and culture.
Dehaene's arguments show how cellular, whole animal and cultural complex adaptive system (CAS) are related. We review his explanations in CAS terms and use his insights to link cultural CAS that emerged based on reading and writing with other levels of CAS from which they emerge.
writing which both allow ideas to
become He explains the cultural engineering that shaped writing to human vision and the exaptations and neuronal structures that enable and constrain reading and culture.
Dehaene's arguments show how cellular, whole animal and cultural complex adaptive system (CAS) are related. We review his explanations in CAS terms and use his insights to link cultural CAS that emerged based on reading and writing with other levels of CAS from which they emerge.
Plans emerge in complex adaptive
systems (CAS) to provide the
instructions that agents use to
perform actions. The component architecture and structure
of the plans is reviewed.
memetic structures and
subject to similar This page reviews the implications of selection, variation and
heredity in a complex adaptive system (CAS).
The mechanism and its emergence are
discussed.
evolutionary
pressures. Hofstadter's theoretical work on
This page discusses the interdependence of perception and
representation in a complex adaptive system (CAS). Hofstadter
and Mitchell's research with Copycat is
reviewed.
perception and
representation and Neuroscientists such as Consciousness has confounded philosophers and scientists for
centuries. Now it is finally being characterized
scientifically. That required a transformation of
approach.
Realizing that consciousness was ill-defined neuroscientist Stanislas Dehaene and others characterized and focused on conscious access.
In the book he outlines the limitations of previous psychological dogma. Instead his use of subjective assessments opened the window to contrast totally unconscious brain activity with those including consciousness.
He describes the research methods. He explains the contribution of new sensors and probes that allowed the psychological findings to be correlated, and causally related to specific neural activity.
He describes the theory of the brain he uses, the 'global neuronal workspace' to position all the experimental details into a whole.
He reviews how both theory and practice support diagnosis and treatment of real world mental illnesses.
The implications of Dehaene's findings for subsequent consciousness research are outlined.
Complex adaptive system (CAS) models of the brain's development and operation introduce constraints which are discussed.
Dehaene's experiments on conscious access,
following Crick and Koch, show how the brain develops to
understand its local environment and then performs conscious
operations. Poggio's
comment "We do not yet understand how the brain gives rise
to intelligence enables the achievement of goals in the face of obstacles. The goals are sub-goals of genes' survival and reproduction and include: Realizing that consciousness was ill-defined neuroscientist Stanislas Dehaene and others characterized and focused on conscious access.
In the book he outlines the limitations of previous psychological dogma. Instead his use of subjective assessments opened the window to contrast totally unconscious brain activity with those including consciousness.
He describes the research methods. He explains the contribution of new sensors and probes that allowed the psychological findings to be correlated, and causally related to specific neural activity.
He describes the theory of the brain he uses, the 'global neuronal workspace' to position all the experimental details into a whole.
He reviews how both theory and practice support diagnosis and treatment of real world mental illnesses.
The implications of Dehaene's findings for subsequent consciousness research are outlined.
Complex adaptive system (CAS) models of the brain's development and operation introduce constraints which are discussed.
- Obtaining and eating food
- Sex
- Finding and maintaining shelter
- Fighting for resources - in the preferred hunter gatherer environment loss of resources was critical while possession was often transient.
- Understanding the proximate environment
- Securing the cooperation of others
The elucidation of these mechanisms illustrates how Bill Gates proposal 'to sequence the human genome and replicate how nature did intelligence enables the achievement of goals in the face of obstacles. The goals are sub-goals of genes' survival and reproduction and include:
- Obtaining and eating food
- Sex
- Finding and maintaining shelter
- Fighting for resources - in the preferred hunter gatherer environment loss of resources was critical while possession was often transient.
- Understanding the proximate environment
- Securing the cooperation of others
Plans are interpreted and implemented by agents. This page
discusses the properties of agents in a complex adaptive system
(CAS).
It then presents examples of agents in different CAS. The examples include a computer program where modeling and actions are performed by software agents. These software agents are aggregates.
The participation of agents in flows is introduced and some implications of this are outlined.
agent is
'relatively' straight forward. Indeed Rob's strategy
studio (RSS) describes a It then presents examples of agents in different CAS. The examples include a computer program where modeling and actions are performed by software agents. These software agents are aggregates.
The participation of agents in flows is introduced and some implications of this are outlined.
This page discusses the interdependence of perception and
representation in a complex adaptive system (CAS). Hofstadter
and Mitchell's research with Copycat is
reviewed.
Copycat
based architecture used to enable the emergence of such agents
that execute on a standard Intel processor. But there are
major challenges:- As explained in our assessment of Searle's John Searle's influential thought experiment implied to him that computers cannot understand. Complex adaptive system (CAS) theory indicates that this is not the case.Chinese Room one challenge is encoding the plan so that it operates analogously to the way we operate.
- The cells from which we emerge are full agents as well. So far our silicon tools require supply chains, factories, transportation, and lots of people to facilitate the equivalent of a cell!
- The efficiency of carbon based cells is orders of magnitude higher than the lowest power neuromorphic processors implement Carver Mead's goal of using analog circuits integrated within VLSI to mimic the biological architectures present in neural circuits. .
- The development
of the environment that the agent operates within was
the most challenging aspect of the RSS is Rob's Strategy Studio models. Today's
life forms did not emerge in this environment. And CAS
theory tells us that due to This page discusses the impact of random events which once they occur encourage a particular direction forward for a complex adaptive system (CAS).frozen accidents starting afresh would generate a different set of scenarios. We would not get life as we know it.
Using the building of a robot limits the exploration of the developmental is a phase during the operation of a CAS agent. It allows for schematic strategies to be iteratively blended with environmental signals to solve the logistical issues of migrating newly built and transformed sub-agents. That is needed to achieve the adult configuration of the agent and optimize it for the proximate environment. Smiley includes examples of the developmental phase agents required in an emergent CAS. In situations where parents invest in the growth and memetic learning of their offspring the schematic grab bag can support optimizations to develop models, structures and actions to construct an adept adult. In humans, adolescence leverages neural plasticity, elder sibling advice and adult coaching to help prepare the deploying neuronal network and body to successfully compete.
aspects of animal systems. There are key
Terrence Deacon explores how constraints on dynamic flows can
induce emergent phenomena
which can do real work. He shows how these phenomena are
sustained. The mechanism enables the development of Darwinian competition.
constraints
and properties that This page discusses the mechanisms and effects of emergence
underpinning any complex adaptive system (CAS). Key research is
reviewed.
emerge during
that development process. In contrast to Haikonen's use of observation and positional sensing to develop self-awareness, systems that are built out of dividing, self-assembling, growing cells, leverage local signals such as chemical gradients and cell surface molecules to set the context of their
Plans emerge in complex adaptive
systems (CAS) to provide the
instructions that agents use to
perform actions. The component architecture and structure
of the plans is reviewed.
schema driven
operations. This origami like developmental activity
results in a physically defined position for all structural cell
types with neurons then seeking out muscle and skin cells based
on 'smell'. Sperry moved some of the structural embryonic
cells and demonstrated that a grown frog's response to touching
their moved skin is to scratch where the itch should be, rather
than the touched position where structural embryonic cells were
moved to. The
This page reviews the implications of reproduction initially
generating a single child cell.
The mechanism and resulting strategic options are
discussed.
single cell developmental bottleneck
ensures that the Plans emerge in complex adaptive
systems (CAS) to provide the
instructions that agents use to
perform actions. The component architecture and structure
of the plans is reviewed.
genetic plan
represents the competitive advantages that have been expressed
and successfully used. This page reviews the implications of selection, variation and
heredity in a complex adaptive system (CAS).
The mechanism and its emergence are
discussed.
Evolved
The agents in complex adaptive
systems (CAS) must model their
environment to respond effectively to it. Samuel
modeling is described as an approach.
learning is explicitly leveraged in
the application of a Plans change in complex adaptive systems (CAS) due to the action of genetic
operations such as mutation, splitting and recombination.
The nature of the operations is described.
genetic algorithm
in the Smiley based This page discusses a complex adaptive system (CAS) implementation of a genetic algorithm (GA), Melanie
Mitchell's robot-janitor built as a set of Copycat codelets
integrated using agent-based
programming. The improvement in the operation of the
robots over succeeding generations of applying the GA is graphed.
The CAS that generated, and operated the robot is reviewed, including the implementation details and codelet operational program flow, and the challenges and limitations of this implementation.
The schematic strings which make up the robot's genotype, as well as the signals which are sent to the nucleus of the robot's agents so that the agents can deploy the appropriate response strings (which activate codelets) are listed. The Slipnet configuration required by the system to associate the schematic strings with programmatic forces (codelets) is also listed. The codelets and supporting perl are also listed.
In the conclusion the limitations of the robot-janitor abstraction in studying emergence and creative evolution are discussed and alternative experimental frameworks are proposed. One such, the schematic cell is the subject of a separate page in this web frame.
virtual robot.
This evolutionary mechanism could apply to Haikonen's robot's
memetic word based plans. The CAS that generated, and operated the robot is reviewed, including the implementation details and codelet operational program flow, and the challenges and limitations of this implementation.
The schematic strings which make up the robot's genotype, as well as the signals which are sent to the nucleus of the robot's agents so that the agents can deploy the appropriate response strings (which activate codelets) are listed. The Slipnet configuration required by the system to associate the schematic strings with programmatic forces (codelets) is also listed. The codelets and supporting perl are also listed.
In the conclusion the limitations of the robot-janitor abstraction in studying emergence and creative evolution are discussed and alternative experimental frameworks are proposed. One such, the schematic cell is the subject of a separate page in this web frame.
While the self-assembled body can depend on relative cellular positions defined in its cell's
Plans emerge in complex adaptive
systems (CAS) to provide the
instructions that agents use to
perform actions. The component architecture and structure
of the plans is reviewed.
genetic
plan and generated developmentally, it is typically
beneficial to be aware of the local external environment.
During development, while the animal is offered some protection
by its parents, it can integrate evolved models of general
percepts with the situations that its sensors experience.
Reading is a clear
example. Other evolved models that are not
encountered over the developmental period are discarded to limit
combinatorial interference and regain resources. And due
to the pressure of competition the integration will leverage Representing state in emergent entities is essential but
difficult. Various structures are used to enhance the rate
and scope of state transitions. Examples are
discussed.
optimized structures ensuring rapid
awareness and response to significant percepts. Consciousness has confounded philosophers and scientists for
centuries. Now it is finally being characterized
scientifically. That required a transformation of
approach.
Realizing that consciousness was ill-defined neuroscientist Stanislas Dehaene and others characterized and focused on conscious access.
In the book he outlines the limitations of previous psychological dogma. Instead his use of subjective assessments opened the window to contrast totally unconscious brain activity with those including consciousness.
He describes the research methods. He explains the contribution of new sensors and probes that allowed the psychological findings to be correlated, and causally related to specific neural activity.
He describes the theory of the brain he uses, the 'global neuronal workspace' to position all the experimental details into a whole.
He reviews how both theory and practice support diagnosis and treatment of real world mental illnesses.
The implications of Dehaene's findings for subsequent consciousness research are outlined.
Complex adaptive system (CAS) models of the brain's development and operation introduce constraints which are discussed.
Conscious access illustrates the two
stage process of: Realizing that consciousness was ill-defined neuroscientist Stanislas Dehaene and others characterized and focused on conscious access.
In the book he outlines the limitations of previous psychological dogma. Instead his use of subjective assessments opened the window to contrast totally unconscious brain activity with those including consciousness.
He describes the research methods. He explains the contribution of new sensors and probes that allowed the psychological findings to be correlated, and causally related to specific neural activity.
He describes the theory of the brain he uses, the 'global neuronal workspace' to position all the experimental details into a whole.
He reviews how both theory and practice support diagnosis and treatment of real world mental illnesses.
The implications of Dehaene's findings for subsequent consciousness research are outlined.
Complex adaptive system (CAS) models of the brain's development and operation introduce constraints which are discussed.
- Discovery/slow repeated practice and
- Deployment of practiced episodic strategies into
unconscious structures associated with the cerebellum is involved with the efficiency of fine movement. It modulates the force and range of motion and is involved in motor coordination and the learning of motor skills. Damage to the cerebellum impairs standing, walking, or performance of coordinated movements. A virtuoso pianist or other performing musician depends on their cerebellum. The cerebellum receives visual, auditory, vestibular, and somatosensory information. It also receives information about individual muscular movements being directed by the brain. The cerebellum integrates this information and modifies the motor outflow, exerting a coordinating and smoothing effect on the movements. However, patients born without a cerebellum have survived reasonably well. The cerebellum is part of the implicit learning mechanism. It is required for the rabbit eye-blink to be classically conditioned to respond to a sound, and puff of air (threat to eye). It integrates the sound and puff and outputs the response to the motor area (blink).
, spine and muscle groups where fast activity can be leveraged.
where these salient, Douglas Hofstadter controlled the amount of attention a Workspace object in Copycat would receive from codelets via its salience. The more descriptions, analogous to geons, an object has and the more highly activated the nodes involved therin, the more important the object is. Modulating this tendency is any relative lack of connections from the object to the rest of the objects in the Workspace. Salience is a dynamic number that takes into account both these factors. In Smiley the instantaneous salience of a Workspace's objects is calculated by itsalience. details are consolidated into the unconscious real time neuronal structures and the rest are destroyed and garbage collected.
Place cells are neurons which fire whenever an animal occupies a certain location in space. Place cells are highly invariant over a variety of sensory cues, and they even maintain their space-selective firing as the animal wanders around in full darkness. They encode where the animal thinks it is. And the same cells associate a time stamp with the memories.
provide a model representation of episodic activity is the memory of conscious experiences. The entorhinal cortex's place cells record both space AND time details so that the memory stream can be reconstructed episodically.
including both time and space in the memory trace recorded. Similarly time and space are represented in the unconscious models built by the dorsal via the parietal cortex is primarily concerned with space and action. It attends to distance, position, speed, and orientation in space. It is not mirror symmetric.
visual pathway.
Haikonen argues that Crick and Koch's identity theory does not explain effectively why some neural processes are perceived as mental content, while others are not. Crick searched unsuccessfully for a set of neural structures that provide the equivalence of conscious mental and neural states. But CAS theory suggests that the
The agents in complex adaptive
systems (CAS) must model their
environment to respond effectively to it. Samuel
modeling is described as an approach.
models that Haikonen views as
percepts need only be present when evolution retained them for
competitive reasons. The neurons, specialized eukaryotic cells include channels which control flows of sodium and potassium ions across the massively extended cell membrane supporting an electro-chemical wave which is then converted into an outgoing chemical signal transmission from synapses which target nearby neuron or muscle cell receptors. Neurons are supported by glial cells. Neurons include a: - Receptive element - dendrites
- Transmitting element - axon and synaptic terminals
- Highly variable DNA schema using transposons.
The agents in complex adaptive
systems (CAS) must model their
environment to respond effectively to it. Samuel
modeling is described as an approach.
structural
bootstrapping for competitor models that we build and
perceive. Haikonen's strategy for a sentient robot learning reading is a basic analog of Dehaene's description of children learning to read. But it is not clear how the robust framework of genetic models recycled to support perception of language and reading in children will be made available to the robot.
Consciousness and Robot Sentience is an insightful book highlighting key attributes of the control networks that integrate the sensors and moving parts of conscious animals and sentient robots.
.
 Politics, Economics & Evolutionary Psychology |
Business Physics |
 |
integrating quality appropriate for each market |
This page looks at schematic structures
and their uses. It discusses a number of examples:
As a working example it presents part of the contents and schematic details from the Adaptive Web Framework (AWF)'s operational plan.
Finally it includes a section presenting our formal representation of schematic goals.
Each goal has a series of associated complex adaptive system (CAS) strategy strings.
These goals plus strings are detailed for various chess and business examples.
Strategy |
Design |
- Schematic ideas are recombined in creativity.
- Similarly designers take ideas and rules about materials and components and combine them.
- Schematic Recipes help to standardize operations.
- Modular components are combined into strategies for use in business plans and business models.
As a working example it presents part of the contents and schematic details from the Adaptive Web Framework (AWF)'s operational plan.
Finally it includes a section presenting our formal representation of schematic goals.
Each goal has a series of associated complex adaptive system (CAS) strategy strings.
These goals plus strings are detailed for various chess and business examples.
This page uses an example to illustrate how:
Program Management |
Home
- A business can gain focus from targeting key customers,
- Business planning activities performed by the whole organization can build awareness, empowerment and coherence.
- A program approach can ensure strategic alignment.
Profiles | Papers | Glossary | E-mail us