Autonomous robot & sp theory of intelligence

AUTONOMOUS ROBOTS &
SP THEORY OF INTELLIGENCE
PRESENTED BY:
CHRISTY ABRAHAM JOY
CHRISTYPONNATTIL@GMAIL.COM

Introduction
Simplify and integrate concepts across artiﬁcial intelligence, mainstream computing and human
perception and cognition, with information compression as a unifying theme
Aim of “The SP theory of intelligence ”
SP THEORY OF INTELLIGENCE 27/19/2015

SP Theory of Intelligence
• Product of an extensive program of development and testing via the SP computer model.
• Knowledge represented with arrays of atomic symbols in one or two dimensions called “patterns”.
• Processing are done by compressing information.
• Via the matching and unification of patterns.
• Via the building of multiple alignments .

Benefits of the SP Theory
• Conceptual simplicity combined with descriptive and explanatory power across several aspects of
intelligence.
• Simplification of computing systems, including software.
• Deeper insights and better solutions in several areas of application.
• Seamless integration of structures and functions within and between different areas of application
4SP THEORY OF INTELLIGENCE7/19/2015

The SP Theory and the SP Machine: A
Summary
• All kinds of knowledge are represented with patterns: arrays of atomic symbols in one or two
dimensions.
• At the heart of the system is compression of information via the matching and uniﬁcation (merging) of
patterns, and the building of multiple alignments
• The system learns by compressing “New” patterns to create “Old” patterns

Multiple Alignment
• The system aims to find multiple alignments that enable a New pattern to be encoded economically in
terms of one or more Old patterns
• Multiple alignment provides the key to:
1. Versatility in representing different kinds of knowledge.
2. Versatility in different kinds of processing in AI and mainstream computing.

The Best Multiple Alignment
SP THEORY OF INTELLIGENCE 7
The best multiple alignment created by the SP computer model with a store of Old patterns like those in rows
1 to 8 (representing grammatical structures, including words) and a New pattern (representing a sentence to
be parsed) shown in row 0.
7/19/2015

Schematic representation of the proposed
SP Machine

Simplification of Computing Systems
Apart from the simplification and integration of concepts in artificial intelligence, mainstream
computing, and human perception and cognition, the SP theory can help to simplify computing
systems, including software.

10
Schematic representations of a conventional computer and an SP machine
SP THEORY OF INTELLIGENCE7/19/2015

Benefits of Overall Simpliﬁcation of
Computing Systems
• Savings in development effort and associated costs. With more intelligence in the CPU there should be
less need for it to be encoded in applications.
• Savings in development time. With a reduced need for hand crafting, applications may be developed
more quickly.
• Savings in storage costs. There may be useful economies in the storage space required for application
code.
117/19/2015 SP THEORY OF INTELLIGENCE

Towards HUMAN-LIKE VERSATILITY
In Intelligence
• Versatility in intelligence - a major strength of the SP system-flows from the goal that has been central
in the development of the theory: to combine conceptual simplicity with descriptive and explanatory
power.
• This strength of the SP system chimes well with what is required in any autonomous robot that is to
function effectively in situations where little or no help can be provided by people.

Towards HUMAN-LIKE VERSATILITY
In Intelligence
A. SIMPLIFICATION AND INTEGRATION
B. NATURAL LANGUAGE PROCESSING
C. PATTERN RECOGNITION
D. INFORMATION STORAGE AND RETRIEVAL
E. VISION
F. REASONING
G. PLANNING AND PROBLEM SOLVING
H. SEQUENTIAL AND PARALLEL PROCEDURES

SIMPLIFICATION AND INTEGRATION
1. SIMPLIFICATION OF STRUCTURES AND FUNCTIONS
• The adoption of one simple format - SP patterns - for the representation of all kinds of knowledge.
• One computational framework, with multiple alignment center-stage, for all kinds of processing.
2. INTEGRATION OF STRUCTURES AND FUNCTIONS
• Syntax and Semantics
• Recognition and Learning
• Knowledge Representation and Learning
• Knowledge Representation and Reasoning
3. DEEPER INSIGHTS AND BETTER SOLUTIONS TO PROBLEMS
• Relatively new insights are the ways in which computational effciency may be improved, with
corresponding savings in the use of energy

NATURAL LANGUAGE PROCESSING
1. PARSING OF NATURAL LANGUAGE
2. PRODUCTION OF NATURAL LANGUAGE
3. THE INTEGRATION OF SYNTAX AND SEMANTICS
4. PARALLEL STREAMS OF INFORMATION
• Vowel sounds, for example, may be analyzed into formants, two or more of which may occur
simultaneously. Vowels, and perhaps other elements of speech, may be represented most
naturally with parallel streams of information
• It does not seem right that the syntactic and semantic aspects of natural language should be
forced into the procrustean bed of a single sequence. As with formants in speech, it seems most
natural to regard syntax and semantics as parallel streams of information.

PATTERN RECOGNITION &
INFORMATION STORAGE AND RETRIEVAL
PATTERN RECOGNITION
• It can recognize patterns at multiple levels of abstraction, with the integration of class-inclusion
relations and part-whole relations.
• It can model ``family resemblance'' or polythetic categories, meaning that recognition does not depend
on the presence absence of any particular feature or combination of features.
• Recognition is robust in the face of errors of omission, commission or substitution in the New pattern or
patterns.
INFORMATION STORAGE AND RETRIEVAL
• The system lends itself to information retrieval in the manner of query-by-example. There is also
potential for information retrieval via the use of natural language or query languages such as SQL.
• The system supports object-oriented concepts such as class hierarchies and inheritance of attributes,
and it provides for the representation of part-whole hierarchies and their seamless integration with
class hierarchies.
7/19/2015

VISION
The main strengths and potential of the SP system are:
• Low level perceptual features such as edges or corners may be identified via the multiple alignment
framework by the extraction of redundancy in uniform areas in the manner of the run-length encoding
technique for information compression
• The system may be applied in the recognition of objects and in scene analysis, with the same strengths
as in pattern recognition
• There is potential for the learning of visual entities and classes of entity and the piecing together of
coherent concepts from fragments
• There is potential, via multiple alignment, for the creation of 3D models of objects and of a robot's
surroundings.

REASONING
The SP system lends itself to several kinds of reasoning:
• One-step `deductive' reasoning.
• Abductive reasoning.
• Reasoning with probabilistic decision networks and decision trees.
• Non-monotonic reasoning and reasoning with default values.
• Reasoning in Bayesian networks, including ``explaining away''.
• Reasoning which is not supported by evidence.
• Inheritance of attributes in an object-oriented class hierarchy or.

TOWARDS HUMAN-LIKE ADAPTABILITY
IN INTELLIGENCE
As with versatility in intelligence, the current generation of robots falls far short of human-like adaptability
in intelligence.
A. PRELIMINARIES
B. UNSUPERVISED LEARNING IN THE SP SYSTEM
C. ONE-TRIAL LEARNING
D. LEARNING LINGUISTIC KNOWLEDGE
E. LEARNING TO SEE
F. HOW A ROBOT MAY BUILD 3D MODELS OF OBJECTS, OF ITSELF, AND OF ITS ENVIRONMENT
G. INTERACTIONS AND OTHER REGULARITIES
H. EXPLORATION, PLAY, AND THE LEARNING OF MINOR SKILLS
I. LEARNING A MAJOR SKILL VIA PRACTICE & DEMONSTRATION
J. CUTTING THE COST OF LEARNING

UNSUPERVISED LEARNING IN THE SP
SYSTEM
In broad terms, the SP70 model processes a set of New patterns (which may be referred to as “I” ) in two
main phases:
1) Create a set of Old patterns that may be used to encode “I”.
2) From the Old patterns created in the first phase, compile one or more alternative grammars for the
patterns in New, in accordance with principles of minimum length encoding
The two phases are described in a little more detail in the following to subsections.
• CREATING CANDIDATE PATTERNS
• COMPILING ALTERNATIVE GRAMMARS

CREATING CANDIDATE PATTERNS
Here, the pattern shown in row 1 is an analogue of something that a child has heard (`t h a t b o y r u n s')
with the addition of code symbols `<', `%1', `9', and `>', while the pattern in row 0 (`t h a t g i r l r u n s') is an
analogue of something that the same child has heard later.

22
From that multiple alignment, the program derives the patterns `t h a t' and `r u n s' from subsequences
that match each other, and it derives `g i r l' and ‘b o y' from subsequences that don't match. In
addition, the program assigns code symbols to the newly-created patterns so that
`t h a t' becomes `< %7 12 t h a t >',
`r u n s' becomes `< %8 13 r u n s >',
and so on. And, using those code symbols, the program creates an abstract pattern,
`< %10 16 < %7 > < %9 > < %8 > >‘
that records the whole sequence. The overall result in this example is the set of patterns. This is
essentially a simple grammar for sequences of the form
`t h a t g i r l r u n s‘ and `t h a t b o y r u n s'.
Patterns derived from the multiple alignment
SP THEORY OF INTELLIGENCE7/19/2015

HOW A ROBOT MAY BUILD 3D MODELS OF
OBJECTS,OF ITSELF, AND OF ITS ENVIRONMENT
• the multiple alignment framework may be applied in creating models of objects (including robots), and
of a robot's environment
• The basic idea is that partially-overlapping images (from the robot's eyes) may be stitched together to
create a coherent whole, in much the same way that partially-overlapping digital photographs may be
stitched together to create a panorama.
Plan view of a 3D object, with each of the five lines around it representing a view of the object, as seen from the side
7/19/2015

INTERACTIONS AND OTHER REGULARITIES
This difference between learning from a one-dimensional stream of information and learning from parallel
streams of information may be accommodated with three refinements of the SP70 model:
• Represent Parallel Streams of Information With 2D Patterns
• Generalize the Sequence Alignment Process to the Matching of 2D Patterns
• Generalize the Process for Building Multiple Alignments to Accommodate 2D Patterns
A multiple alignment produced by the SP computer model showing how two instances of the
pattern `I N F O R M A T I O N' may be detected despite the interpolation of non-matching symbols
throughout both instances.
7/19/2015

Deeper Insights and Better Solutions in
Several Areas of Application
1. Applications in the Processing of Natural Language
• Towards the Understanding and Translation of Natural Language
• Natural Language and Information Retrieval
• Interactive Services
• Going Beyond FAQs
2. Towards a Versatile Intelligence for Autonomous Robots
• Potential for the kind of visual analysis needed to assimilate the many conﬁgurations of balls,
pockets, and cue
• The versatility of the SP framework in the representation and processing of diverse kinds of
knowledge should facilitate the seamless integration of visual information about the table, balls,
and so on, with information about actions by the player and feedback from muscles and from touch.

3. Computer Vision
• It has potential to simplify and integrate several areas in computer vision, including feature
detection and alignment, segmentation, deriving structure from motion, stitching of images
together, stereo correspondence, scene analysis, and object recognition
4. A Versatile Model for Intelligent Databases
• The system would provide a means of storing and managing the data that are gathered in such
investigations, often in large amounts.
• It may help in the recognition of features or combinations of features that link a given crime to
other crimes, either current or past—and likewise for suspects.
• The system’s capabilities in pattern recognition may also serve in the scanning of data to recognize
indicators of criminal activity.

5. Software Engineering
• Procedural Programming, Automatic Programming,
• No Compiling or Interpretation
• Sequential and Parallel Processing
6. Information Compression
7. Medical Diagnosis
8. Managing “Big Data” and Gaining Value from It
9. Other Areas of Application
• Knowledge, Reasoning, and the Semantic Web
• Bioinformatics
• Detection of Computer Viruses
• Data Fusion
• Development of Scientiﬁc Theories & New Kinds of Computes

Conclusion
•The SP theory of intelligence and its realization in the SP machine may facilitate the development
of autonomous robots: by increasing the computational efficiency of computers; by facilitating
the development of human-like versatility in intelligence; and likewise for human-like
adaptability in intelligence.
•The SP system has potential for substantial gains in computational efficiency, with corresponding
cuts in energy consumption and in the bulkiness of computing machinery: by reducing the size
of data
•Autonomous robots will require a non-von revolution - perhaps along the lines of SP-neural -
there is plenty that can be done via modelling with von-Neumann-style supercomputers to
explore the potential of new architectures.

References
Ames Gerard Wolff, “Autonomous Robots and the SP Theory of Intelligence”, IEEE
Access/January 21, 2015
Wolff, J.G. The SP theory of intelligence: An overview. Information 2013, 4, 283–341

Autonomous robot & sp theory of intelligence

Recomendados

Recomendados

Mais conteúdo relacionado

Mais procurados

Mais procurados (20)

Semelhante a Autonomous robot & sp theory of intelligence

Semelhante a Autonomous robot & sp theory of intelligence (20)

Último

Último (20)

Autonomous robot & sp theory of intelligence