Intellifest 2012 AI researcher PhD candidate Center for Analysis and Design of Intelligent Agents Reykjavik University Helgi Páll Helgason helgiperseptiocom Schedule changes Today ID: 523964
Download Presentation The PPT/PDF document "AGI & Attention" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
AGI & Attention
Intellifest 2012
AI researcher, Ph.D. candidateCenter for Analysis and Design of Intelligent Agents, Reykjavik University
Helgi Páll Helgason
helgi@perseptio.comSlide2
Schedule changes
Today: AttentionFriday November 2nd:
AttentionTuesday November 6th: AutonomyIntellifest 2012Slide3
Today
Importance of attention for AIExample of inspiration from human attention (cognitive psychology)
Attention beyond the human level (for meta-cognitive purposes)Design of an attention mechanismIntellifest 2012Slide4
Attention & AI
To work on AI research, each researcher or team needs to have
A clear and explicit working definition of intelligenceA clear and explicit motivationIntellifest 2012Slide5
What do we mean by machine intelligence?
Displaying human-like behavior
?Solving computationally complex problems (in some unspecified amount of time)?Performing isolated tasks that have conventionally required humans?Adapting to a complex, dynamic environment with insufficient knowledge and resources?
Intellifest 2012Slide6
What is our motvation?
Development and validation of psychological models?
Development and validation of neurological models?Development and implementation of practical, flexible and versatile autonomous system?How accurately do we want to replicate existing biological mechanisms?To what degree are we biologically inspired?
Intellifest 2012Slide7
Intelligence
Intellifest 2012
Nature’s way of dealing with complexity under resource and time constraints
No real-world intelligence exists that does not address the passage of time head onSlide8
A working definition
Intellifest 2012
“Intelligence is the capacity of a system to
adapt to its environment while operating with insufficient knowledge and resources
.”
Pei Wang
(Rigid flexibility:
The Logic of Intelligence. Springer
2006)Slide9
A working definition
Intellifest 2012
Intelligence is a capability of information processing systems
Intelligence is adaptation:
The system’s solution of one problem is not only determined by the problem itself, but also prior experienceSlide10
A working definition
Intellifest 2012
Insufficient knowledge -> system will usually not have the best solution at hand
Insufficient resources -> system can not consider (process) every possibility nor store all informationSlide11
Intelligence
Intellifest 2012
“
If either time or computational resources are infinite,
intelligence is irrelevant
.”
Dr.
Kristinn
R.
ThórissonSlide12
Attention
In the domain of intelligent systems, the management of system resources is typically called “attention
”Biological (Human) Attention:Selective concentration on one aspect of the environment while ignoring othersArtificial Attention:Resource management and control mechanism to assign limited system resources to processing of most relevant or important information
Intellifest 2012Slide13
Attention
Intellifest 2012
Time constraints
Abundant information
Limited resourcesATTENTIONSlide14
Intelligence
Intellifest 2012
Time constraints
Abundant information
Limited resourcesINTELLIGENCESlide15
What do we mean by AI?
“Narrow“ (classical) AI:Systems explicitly designed to solve specific, reasonably well-defined problemsE.g. Deep Blue, Watson, etc.
Artificial General Intelligence (AGI)Systems designed to autonomously learn novel tasks and adapt to changing environments
Intellifest 2012Slide16
Attention and Narrow AI
When tasks and environments are pre-specified, we know in advance…
what information is relevant to system operationhow frequently the system has to sample informationhow frequently the system has to act
the resource requirements of the systemIntellifest 2012Slide17
Attention and Narrow AI
Substantial dynamic adaption to task not required
Information filtering can be pre-programmed if characteristics of relevant information known in advanceResource management and processing hand-tuned for specific tasks and environmentsMajor reduction in complexity (compared to real-world tasks and environments)
End up with limited and closed models of the real-worldIntellifest 2012Slide18
Attention and AGI
When tasks and environments are unknown, we do
not know in advance…what information is relevant to system operationhow frequently the system has to sample information
how frequently the system has to actthe resource requirements of the system
Intellifest 2012Slide19
Attention and AGI
Must assume up-front:Real world environmental complexity
All information is potentially importantNot just limited, but insufficient resources at all timesDynamic tasks, environments and time constraints
Intellifest 2012Slide20
Why attention?
“Narrow” AISubstantial dynamic adaptation to task not required
Data filtering can be pre-programmed if characteristics of useful data known in advanceLower than real world task complexityResource management and processing hand-tuned for specific scenarios→
Attention not required (?)AGIReal world environmental complexity assumed up-frontComputational resources for the AI assumed to be insufficient at all timesComplexity calls for data filtering and intelligent resource allocation
Environments and tasks unknown at implementation timeResource management must be adaptive→ Demands strong focus on resource management and realtime processingIntellifest 2012Slide21
Resource management challenges for AGI systems
Real world is highly dynamic and complex, provides abundance of information
.System resources not only limited, but insufficient in light of amount of available information.Range of time constraints
(many of which are dictated by the environment) must be satisfied.Unexpected events requiring immediate response may occur at any time.
Intellifest 2012Slide22
SOAR
AGI SUMMER SCHOOL 2012Slide23
Intellifest 2012Slide24
Human attention
Intellifest 2012
“Everyone knows what attention is. It is the taking possession by the mind, in clear and vivid form, of one out of what seem several simultaneously possible objects or trains of thought. Focalization, concentration, of consciousness are of its essence. It implies withdrawal from some things in order to deal effectively with others, and is a condition which has a real opposite in the confused, dazed, scatterbrained state which in French is called distraction, and
Zerstreutheit in German.”- William James, 1890Slide25
Attention in Cognitive Psychology
Modern attention research started with the “cocktail party effect” (Colin Cherry, 1953)
Number of attention models have been proposed, most belonging to two classes:Early selection: Selection of information occurs early in sensory pipeline based on shallow, primitive processing with no or limited analysis of meaningLate selection: Selection of information based on deep analysis of meaning, occurs late in sensory pipeline
Many early selection models contradicted by observed human behavior (e.g. cocktail party scenario)
Intellifest 2012Slide26
Broadbent Filter Model (Early Selection)
Intellifest 2012Slide27
Treisman Attenuation Model(Early selection)
Intellifest 2012Slide28
Early selection
Intellifest 2012Slide29
Deutsch-Norman Model(Late selection)
Intellifest 2012Slide30
Knudsen Attention Framework(Late selection)
Intellifest 2012
Knudsen EI: Fundamental components of attention. Annu
Rev Neurosci 2007, 30:57-78.Slide31
Attention in Cognitive Psychology
Two types of attention:Top-down
Deliberate, goal-driven, targets information related to the tasks being performedBottom-upReactive, targets unexpected but potentially important information
Intellifest 2012Slide32
Inspiration from cognitive psychology
Intellifest 2012
Information competes for limited system resourcesAllows processing decisions to be made as late as possible, when resource availability is known
Early selection may be a problematic paradigmIgnoring information without analysis of meaning introduces operational riskTwo simultaneously active functions of attention can allow systems to perform tasks while remaining reactive to unexpected events.Top-down attention may be controlled by active goals and predictions of the system to catch information related to current tasks.
Bottom-up attention can be controlled by novelty and unexpectedness of incoming informationSlide33
Case study
General Attention Mechanisms for Cognitive ArchitecturesMy Ph.D. project
Intellifest 2012Slide34
Research goals
Design a general attention mechanism intended for implementation in AGI systems (cognitive architectures)
In progress: Implementation and evaluation of resulting attention mechanism in state-of-the-art cognitive architecturesWhile work is biologically inspired at high-level, replication of any existing attention mechanism is not a goal
Intellifest 2012Slide35
Design requirements
CompleteTargets all operational information (internal + external)Top-down +
Bottom-upGeneralNo limiting assumptions about tasks, environments or modalitiesUniformData from all modalities treated identically (at some level of processing
)AdaptiveLearns from experience
Intellifest 2012Slide36
General approach to attention
Modality neutralAll modalities treated identically, at some level of
processingIncluding proprioception (internal modalities, self-sensing)Architecture-independent
Intellifest 2012Slide37
Related work
Attention functionality implemented in handful of AGI systemsE.g. NARS, LIDA, CLARION
Limitations:Data-filtering only (control issues ignored)External information only (internal states ignored)Realtime processing not addressed
Intellifest 2012Slide38
Approach
Amount of available information constantly assumed to exceed system processing capacity
Limited system resources must be focused on most relevant or important informationRequires capability to determine degree of information importance, based on:Current operating contextTime constraintsResource availability
Intellifest 2012Slide39
Issues for AGI attention
Intellifest 2012
Is it sufficient to only evaluate relevance for data?What about the relevance of system processes?Just another “box” in the sensory pipeline?Or something more pervasive
?Can we retrofit existing AGI architectures with attention?Can we extend attention capabilities in useful ways for AGI systems?Slide40
Attention: Functional
requirementsIntellifest 2012
Quantify current relevance of dataData relevance:
Goal-relatedNovelty / UnexpectednessQuantify current relevance of processesProcess relevance:Operational experienceAvailable dataSlide41
Methodology
Intellifest 2012
Constructivist AI“From Constructionist to Constructivist AI”, Thórisson 2009, BICA proceedingsSystems manage own growthFrom manually constructed initial state (bootstrap/seed)
Methodology for building flexible AGI systems capable of autonomous self-reconfiguration at the architecture levelSlide42
Attention for self-reconfiguration
Intellifest 2012
Internally, the system and its operation can be viewed as a dynamic and complex environmentSimilar to external task environmentMeta-cognitive functions responsible for system growth must also process information selectively
Resources remain limitedApplying the same attention mechanism to external and internal environments may produce AI systems capable of performing tasks and improving own performance while being subject to realtime constraints and resources limitations.Slide43
Meta-cognitive operation
Intellifest 2012
Introspection, self-growth, self-improvementAs the sum of internal system activity is a vast stream of information…applying attention to the internal environment can lend significant support for meta-cognitive operation by…
helping determine important information and processes for meta-cognitive functions…in the same way it supports task performance in the external environmentSlide44
Attention mechanism design
Intellifest 2012
While design is architecture-independent, some requirements are necessarySlide45
Architectural requirements
Intellifest 2012
Data-driven
All processing is triggered by the occurrence of data
Eliminates the need for fixed control loops, allowing for operation at multiple time scales and greater flexibility
Fine-grained
Data and processing units are small but numerous
Reasoning about small, simple components and their effects is significantly more tractable than for larger, more complex componentsSlide46
Architectural requirements
Intellifest 2012
Predictive capabilities
Capacity to generate predictions and expectations
Necessary control data for top-down attention in addition to goals
Unified sensory pipeline
Data given identical treatment regardless of origin (external, internal)Slide47
Data items
Processes
New data
Sensory devices
Environment
(Real world)
Actuation devices
Commands
Sampled data
Data-driven
: Processes are activated only when paired with compatible data
Fine-grained
: Data and process objects are small and numerous
Unified sensory pipeline
: External (environmental) and internal data handled identically at architecturelevelSlide48
Goals / Predictions
Attentional patterns
Derived
Matching
Data items
Processes
Data biasing
Top-down
Sensory devices
Environment
(Real world)
Actuation devices
Commands
Sampled data
Predictive capabilities
: Predictions are necessary control information for top-down attention
Data and processess have priority values that are assigned by biasing.Slide49
Attentional patterns
Matching
Data items
Processes
Bottom-up attentional
processess
Evaluation
Top-down
Bottom-up
Sensory devices
Environment
(Real world)
Actuation devices
Commands
Sampled data
Data biasing
Goals / Predictions
DerivedSlide50
Attentional patterns
Matching
Data items
Processes
Top-down
Bottom-up
Process biasing
Data -> Process mapping
Sensory devices
Environment
(Real world)
Actuation devices
Commands
Sampled data
Data biasing
Goals / Predictions
Derived
Bottom-up attentional
processess
EvaluationSlide51
Attentional patterns
Matching
Data items
Processes
Top-down
Bottom-up
Contextualized process performance history
Contextual process evaluation
Experience-based
p
rocess activation
Sensory devices
Environment
(Real world)
Actuation devices
Commands
Sampled data
Data biasing
Goals / Predictions
Derived
Bottom-up attentional
processess
Evaluation
Process biasing
Data -> Process mappingSlide52
Resources
Intellifest 2012
Cognitive architectures
NARS
https://sites.google.com/site/narswang
/
LIDA
http://goertzel.org/agiri06/%
5B4%5D%20StanFranklin.pdf
CLARION
https://sites.google.com/site/clarioncognitivearchitecture
/
Publications:
Cognitive Architectures and Autonomy: A Comparative Review
Kristinn R. Thórisson,
Helgi
Páll
Helgason
http://versita.metapress.com/content/052t1h656614848h/?p=4e1d01ba40e04d5d9f51da3977a8be04&pi=0
Attention Capabilities for AI Systems
Helgi
Páll
Helgason
, Kristinn R. Thórisson
http://www.perseptio.com/publications/Helgason-ICINCO-2012.pdf
On Attention Mechanisms for AGI Architectures: A Design Proposal
(to be published)
Helgi
Páll
Helgason
,
Kristinn
R.
Thórisson
, Eric
Nivel
http://www.perseptio.com/publications/Helgason-AGI-2012.pdfSlide53
Assignments
Intellifest 2012
Wiki reading material for attention
Additional paper:
On Attention Mechanisms for AGI Architectures: A Design Proposal
(to be published)
Helgi
Páll
Helgason
, Kristinn R. Thórisson, Eric
Nivel
http://www.perseptio.com/publications/Helgason-AGI-2012.pdf
Post questions to
Proboard