AI and SWARM: Evolutionary Approach To Emergent Intelligence 1st Edition Hitoshi Iba 2024 Scribd Download
AI and SWARM: Evolutionary Approach To Emergent Intelligence 1st Edition Hitoshi Iba 2024 Scribd Download
com
https://textbookfull.com/product/ai-and-swarm-
evolutionary-approach-to-emergent-
intelligence-1st-edition-hitoshi-iba/
https://textbookfull.com/product/swarm-intelligence-algorithms-
modifications-and-applications-1st-edition-adam-slowik/
textbookfull.com
https://textbookfull.com/product/ai-meets-bi-artificial-intelligence-
and-business-intelligence-1st-edition-lakshman-bulusu/
textbookfull.com
The Long and Short Non coding RNAs in Cancer Biology 1st
Edition Erwei Song (Eds.)
https://textbookfull.com/product/the-long-and-short-non-coding-rnas-
in-cancer-biology-1st-edition-erwei-song-eds/
textbookfull.com
Colloquial English Structure and Variation Andrew Radford
https://textbookfull.com/product/colloquial-english-structure-and-
variation-andrew-radford/
textbookfull.com
https://textbookfull.com/product/critical-care-neurology-part-i-eelco-
f-m-wijdicks-and-andreas-h-kramer-eds/
textbookfull.com
https://textbookfull.com/product/education-and-technological-
unemployment-michael-a-peters/
textbookfull.com
https://textbookfull.com/product/who-is-vera-kelly-rosalie-knecht/
textbookfull.com
https://textbookfull.com/product/heidegger-on-death-and-being-an-
answer-to-the-seinsfrage-johannes-achill-niederhauser/
textbookfull.com
Evolution Equations Long Time Behavior and Control London
Mathematical Society Lecture Note Series 1st Edition Kaïs
Ammari (Editor)
https://textbookfull.com/product/evolution-equations-long-time-
behavior-and-control-london-mathematical-society-lecture-note-
series-1st-edition-kais-ammari-editor/
textbookfull.com
AI and SwArm
Evolutionary Approach
to Emergent Intelligence
Hitoshi Iba
Information and Communication Engineering
School of Information Science and Technology
The University of Tokyo, Japan
p,
p,
A SCIENCE PUBLISHERS BOOK
A SCIENCE PUBLISHERS BOOK
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
© 2020 by Taylor & Francis Group, LLC
CRC Press is an imprint of Taylor & Francis Group, an Informa business
No claim to original U.S. Government works
Printed on acid-free paper
Version Date: 20190725
International Standard Book Number-13: 978-0-367-13631-4 (Hardback)
This book contains information obtained from authentic and highly regarded sources. Reasonable efforts have been
made to publish reliable data and information, but the author and publisher cannot assume responsibility for the
validity of all materials or the consequences of their use. The authors and publishers have attempted to trace the
copyright holders of all material reproduced in this publication and apologize to copyright holders if permission to
publish in this form has not been obtained. If any copyright material has not been acknowledged please write and let
us know so we may rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted,
or utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, includ-
ing photocopying, microfilming, and recording, or in any information storage or retrieval system, without written
permission from the publishers.
For permission to photocopy or use material electronically from this work, please access www.copyright.com
(http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers,
MA 01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety
of users. For organizations that have been granted a photocopy license by the CCC, a separate system of payment
has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for
identification and explanation without intent to infringe.
This book is an explanatory text about Artificial Intelligence (AI) and swarm,
that is, “emergence” for AI. These days, we can say that we are in the middle of
the third AI boom. One of the factors responsible for this is deep learning, which
resulted in the evolution of neural networks as well as machine learning based on
statistical theory. However, there remains the question of whether it is possible
to create a true AI (a human-like general-purpose intelligence, the so-called
“strong” AI). Many research results indicate that there is clearly more depth in
human intelligence and cognition than is possible to express by AI techniques
(“weak” AI).
For about 30 years, the author has studied evolution and emergent
computation. The swarm mechanism gives us a glimpse of the principles of
intelligence in humans and living beings. In this book, such topics related to
artificial intelligence are explained in detail, from the theoretical background to
the most recent progress, as well as future topics.
Some readers may wonder if some of the topics related to swarm do really
belong to the realm of AI technology. Indeed, the essence of AI is to find problems.
According to this line of thought, a problem is no longer considered to be AI as
soon as it is solved. Brooks’ joke in the beginning of this text symbolizes this
point. Brooks is a researcher specializing in AI at MIT (Massachusetts Institute
of Technology) since the 1980’s. He is the author of numerous innovative ideas
in the field of AI. He is also famous for having created the robot vacuum cleaner
Roomba (he is the founder of iRobot in the US). From this perspective, is it
possible to say that, because the vacuum cleaner robot is already in the market,
it is no longer AI?
Current, AI seems to be surrounded by an excessively flamboyant
advertisement of its technical applications. Of course, it is worth noting that
iv < AI and Swarm: Evolutionary Approach to Emergent Intelligence
the intention here is not to deny all this. The author himself is involved in such
projects. However, being satisfied with the above is a different issue. Given that
it is possible to link any type of intelligence to AI, we should not ignore basic
research that focuses on the essence of cognition and life.
The objective of the author’s research is to build a model of human cognitive
functions from the perspective of “emergence” in order to achieve strong AI.
In other words, the objective is to understand how intelligence appeared and
interacted with the real world, as well as the causal relationships in terms of actual
physical and chemical mechanisms. In that sense, this book intends not only
to explain intelligent behavior itself, but also understand its causes and effects.
In order to do that, I have explained the emergence of several types of human
cognitive errors, cognitive dissonances, irrational behaviors, and cooperative/
betrayal actions.
Unfortunately, it is difficult to explain such phenomena using machine
learning based on statistical models and deep learning based on big data, which
constitute the core of current AI. On the contrary, a few hypotheses that contradict
such approaches may be needed. In this book, research results related to brain
mechanism are described, which is a factor that directly triggers intelligent
behavior (called “proximate cause” or “physiological cause” in biological
jargon), and the reasons for the evolution of living beings’ behavior (“ultimate
cause,” or “biological/evolutional cause”). A few emergence simulation models
are also introduced for better understanding. These approaches are important
research efforts to clarify the essence of intelligence, in order to achieve true AI.
Most photographs depicting nature (modes of life and shapes of fish, patterns
found in animals, images of Galapagos, Papua New Guinea, etc.) were shot by
the author himself. Unfortunately, it is difficult for us to witness the evolution of
nature on our own. However, it is possible to feel the driving force of emergence
and the significance of diversity. In order to do that, it is suggested that you leave
behind your PC or smartphone and see real things with your own eyes. The
author has indulged in water life observation for about 30 years as a hobby, and
still feels moved by it, reinforcing his conviction that nothing compares to seeing
the real thing. Even though it is now possible to obtain almost any kind of image
from the Internet, there are lots of things to be learned from real nature, and they
are worth seeing.
Numerous things can be learned from actual living beings and from nature
when studying swarms and AI. Therefore, in this book, the author has tried as
much as possible to explain topics related to living beings and real life based on
actual examples. It is the intent of the author that the readers will expand those
topics and tackle new issues related to AI and swarm.
Acknowledgments
Preface iii
Acknowledgments v
Abbreviations xi
1. Introduction 1
1.1 What is AI? – Strong AI vs Weak AI 1
1.2 What is Emergence? 6
1.3 Cellular Automaton and Edge of Chaos 7
3. meta-heuristics 49
3.1 Ant Colony Optimization (ACO) 49
3.1.1 Collective Behaviors of Ants 49
3.1.2 Simulating the Pheromone Trails of Ants 51
3.1.3 Generating a Death Spiral 52
3.1.4 ACO using a Pheromone Trail Model 53
3.2 Particle Swarm Optimization (PSO) 56
3.2.1 Collective Behavior of Boids 56
3.2.2 PSO Algorithm 60
viii < AI and Swarm: Evolutionary Approach to Emergent Intelligence
6.3 How does Slime Solve a Maze Problem? Slime Intelligence 190
6.4 Swarm Robots 197
6.4.1 Evolutionary Robotics and Swarm 197
6.4.2 Transportation Task for Swarm 199
6.4.3 Occlusion-based Pushing (OBP) 199
6.4.4 Guide-based Approach to OBP 203
6.4.5 Let us See How they Cooperate with each other 205
7. Conclusion 211
7.1 Summary and Concluding Remarks 211
References 213
Index 223
Color Section 235
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
Abbreviations
SA : simulated annealing
SPT : shortest processing time
TFT : tit for tat
TSP : traveling salesman problem
UCB : upper confidence bound
WSLS : Win-Stay-Lose-Shift
ZD : zero-determinant strategy
Chapter 1
Introduction
In the same manner, artificial life (AL), which is mentioned later, can be defined
in terms of strong AL and weak AL.
“Weak AI” is an application of AI technologies to enable a high-functioning
system that replicates human intelligence for a specific purpose. In fact, there are
breakthroughs in weak AI.
In this book we consider simulation in the sense of “strong AI”. More pre-
cisely, the rationale behind this approach is that “the appropriately programed
computer really is a mind, in the sense that computers, given the right programs,
can be literally said to understand and have other cognitive states.”
2 AI and Swarm: Evolutionary Approach to Emergent Intelligence
One day, two new users, A and B, join the discussion board.
When a message is sent to A and B, they both return apt responses.
Of A and B, one is human and the other is a computer.
However, it is impossible to determine which is which, regardless of the
questions asked.
If a program passes this test (in other words, the computer cannot be identi-
fied), the program can be said to simulate intelligence (as long as the questions
are valid). A similar contest, named the “The Loebner prize” after its patron,
the American philanthropist Hugh Loebner, is held online2 . Although an award
of 100,000 US dollars and a solid gold medal has been offered since 1990, so
far, not a single machine participating in the contest has satisfied the criteria for
winning.
Nevertheless, a number of problems with the Turing test have been pointed
out, and various critical remarks have been issued about potential implementation
1 Alan Turing (1912–1954): British mathematician. He worked on deciphering the Enigma encryption
used by the German Army during World War II. He is considered the “father of computer science.” The
Turing machine for computing theory and the Turing model for morphogenesis are some of his pioneering
achievements. Apple’s bitten apple logo is allegedly an homage to Turing.
2 http://www.loebner.net/Prizef/loebner-prize.html
Introduction 3
However, these counterarguments are devoid of meaning. The former rejects the
realization of AI in the first place, and the latter cannot be refuted in light of the
possibility that ultra-high-speed parallel computing or quantum computing may
exist in the future. Thus, neither one can serve as the basis of an argument.
One powerful counterargument is based on system theory. Although the per-
son in the room certainly lacks understanding, he constitutes no more than a sin-
gle part of a larger system incorporating other elements, such as the paper and the
database, and this system as a whole does possess understanding. It is the equiv-
alent to the fact that, even for Chinese, cranial nerve cells do not understand the
language by themselves. This point is integral to the complex systems regarded in
this book. The level at which intelligence is sought depends on the observed phe-
nomenon, and if the phenomenon is considered as being an emergent property,
the validity of the above system theory can be recognized. Moreover, a debate is
ongoing about whether intelligence should be thought of as an integrated concept
or as a phenomenon that is co-evolving as a result of evolution.
On the other hand, Searle reargues that “if the manual is fully memorized
and the answers are given without external help, it is still possible not to under-
stand the Chinese language”. Recently, Levesque et al. [67] counterpoint against
Searle’s criticism has been made based on a theory of computation. This is de-
scribed in the “addition room argument” as follows:
For the “addition room,” the following manual immediately comes to mind:
This manual merely lists the calculation results in the same way as a dictionary.
Just like in the Chinese room argument, a human following this manual does not
add and additionally does not understand the calculation at all. So, was Searle’s
claim correct?
Let us consider here the computational complexity of this manual. The chap-
ter corresponding to the first number requires ten to the tenth power. Each chapter
consists of ten to tenth power of sections. Since this is repeated twenty times, it
becomes 10 to 10 power to 20 power = 10200 of items. It is said that the number
of molecules in the universe is around 10100 . Therefore, such a large manual can
never be made. From this, it can be understood that Searle’s claim is mistaken,
from a computational theory. This is true as far as this manual is concerned, but
Searle himself did not specify the composition method of the manual, so there
may be recurring objections.
Recently, machine translation is accomplished using a huge database and sta-
tistical processing. The famous one is Google’s machine translation. Past ma-
chine translation was a classical AI based on natural language understanding.
Unfortunately, it was not always effective. On the other hand, Google’s machine
translation does not understand natural language at all. This software statistically
joins translated words using a huge database (human translations, e.g., minutes
of the United Nations). As a result, it won the 2006 machine learning contest with
an overwhelming difference. This kind of translation method is possible due to
the power of modern computers and internet connection. This approach may be
a solution to the Chinese room argument.
6 AI and Swarm: Evolutionary Approach to Emergent Intelligence
gossip that circulated around Princeton that he slept with a mystery novel as his pillow in the day and
learned mathematics from the devil at night, he conducted pioneering research in various areas, such as
mathematics, physics, engineering, computer science, economics, meteorology, psychology, and political
science, having substantial influence in posterior advances in these areas. He was also deeply involved
with American nuclear policy and was part of the Manhattan project.
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
8 AI and Swarm: Evolutionary Approach to Emergent Intelligence
itself, but could also build machines more complex than itself. The research
stopped because of his death; however, in 1966, Arthur Burks edited and pub-
lished von Neumann’s manuscripts. John Conway, a British mathematician, ex-
panded on the work of von Neumann and, in 1970, introduced the Game of Life,
which attracted immense interest. Some people became “Game of Life hack-
ers,” programmers and designers more interested in operating computers than in
eating; they were not the criminal hackers of today. Hackers at MIT rigorously
researched the Game of Life, and their results contributed to advances in com-
puter science and artificial intelligence. The concept of the Game of Life evolved
into the “cellular automata” (CA), which is still widely studied in the field of
artificial life. Most of the research on artificial life shares much in common with
the world where hackers played in the early days of computers.
The Game of Life is played on a grid of equal-sized squares (cells). Each cell
can be either “on” or “off”. There are eight adjacent cells to each cell in a two-
dimensional grid (above and below, left and right, four diagonals). This is called
the Moore neighborhood. The state in the next step is determined by the rules
outlined in Table 1.1. The “on” state corresponds to a “•” in the cell, whereas
the “off” state corresponds to a blank. The following interesting patterns can be
observed with these rules.
1. Disappearing pattern (diagonal triplet)
•
• ⇒ • ⇒ disappears
•
“Eaters” that stop gliders and “glider guns” that shoot gliders can be defined, and
glider guns are generated by the collision of gliders. Such self-organizing capa-
bilities mean that the Game of Life can be used to configure a universal Turing
machine. The fundamental logic gates (AND, OR, NOT) consist of glider rows
and disappearing reactions, and blocks of stable patterns are used as memory.
However, the number of cells needed in a self-organizing system is estimated to
be about 10 trillion (3 million×3 million). The size would be a square whose
sides are 3 km long, if 1 mm2 cells are used.
Consider a one-dimensional Game of Life, one of the simplest cellular au-
tomata. The sequence of cells in one dimension at time t is expressed as follows:
time t : 0010011010101100
time t+1 : *11111001010001*
An interesting problem is the task of finding the majority rule. The task is to
find a rule that would ultimately end in a sequence of all 1 (0) if the majority of
the cells are 1 (0) with the minimum radius (r) possible for a one-dimensional
binary sequence of a given length. The general solution to this problem is not
known.
A famous example is a majority rule problem with length 149 and radius 3.
The problem is reduced to finding a function that assigns 1 or 0 to an input with
7
7 bits (= 3 + 1 + 3, radius of 3 plus itself); therefore, the function space is 22 .
How can a cellular automaton (CA) obtain a solution to the majority
problem?
10 AI and Swarm: Evolutionary Approach to Emergent Intelligence
One method is to change the color (black or white) of a cell to the majority
of its neighboring cells. However, this method does not work well, as shown in
Fig. 1.4, because it results in a fixed pattern divided into black and white.
In 1978, Gacs, Kurdyumov and Levin found the rules (GLK) regarding this
problem. Lawrence Davis obtained an improved version of these rules in 1995,
and Rajarshi Das proposed another modification. There is also research to find
effective rules through GAs (genetic algorithms) or GP (genetic programming).
The concept of Boolean functions is applied when GP is used. The fitness value is
defined by the percentage of correctly processed sequences out of 1000 randomly
generated sequences of length 149.
Rules determined by various methods are summarized in Table 1.2. Here, the
transition rules are shown from 0000000 to 1111111 in 128-bit form. In other
words, if the first bit is 0,
Figure 1.4: CA carrying out majority voting (Oxford University Press, Inc., [76]).
The two edges of the growing triangular region at the center with a chess-
board pattern grow at the same pace, progressing the same distance per unit time.
The left edge extends until it collides with a vertical boundary. The right edge
barely avoids the vertical boundary at the left (note that the right and left edges
are connected). Therefore, the left edge can extend for a shorter length, which
means that the length of the white region limited by the left edge is shorter than
Introduction 11
the length of the black region limited by the right edge. The left edge disappears
at the collision point, allowing the black region to grow. Furthermore, the two
edges disappear at the bottom vertex and the entire lattice row becomes black,
showing that the correct answer was obtained.
12 AI and Swarm: Evolutionary Approach to Emergent Intelligence
Figure 1.6: Explanation of CA behavior from collision of particles (Oxford University Press,
Inc., [76]).
6 Stephen Wolfram (1959–): British theoretical physicist. At the age of 20, he obtained his Ph.D. de-
gree from California Institute of Technology for his research on theoretical physics. In 1987, he found
the company Wolfram Research. He released the mathematical software Mathematica, which is used for
research in natural sciences and numerous other fields.
Random documents with unrelated
content Scribd suggests to you:
A herczegasszony teljesíté kívánságát, s más nap egy gyönyörű
arany hal úszkála a tóban.
A sárkány felesége előhívatá férjét, s ezt mondá neki: Meghalok,
ha az arany halat szobámba nem hozzák.
A sárkány kész vala teljesíteni a mit felesége kívánt, de a halat
semmiképen meg nem foghaták. Minthogy a sárkány derekasan
tudott úszni, feltevé magában, hogy maga mégyen el érte. Leköté
oldaláról a rozsdás kardot, hogy az gátlására ne legyen az úszásban;
levetette a szennyes inget, mert annak sem vala szabad megázni, s
belé szökött a tóba. De az arany hal egy ugrással a partra veté ki
magát, s Vas Laczi ismét Vas Laczi leve. Hirtelen fölvette az inget,
hirtelen kirántotta a kardot.
Ezt látván a tizenkétfejű sárkány felesége, felűlt a seprő szárára s
elrepűlt.
A sárkánynak eszébe juta, hogy Vas Laczi megint felélede,
mihelyt testét a lóra kötötték; ő is arra kérte tehát Vas Laczit, hogy
ha megöli, testét köttesse fel a lóra. – Vas Laczi egy csapással üté le
mind a tizenkét fejét, s a tizenkét fejet a testtel együtt a lóra kötteté,
s szabadon hagyá menni, valamerre magának tetszend. Ment a ló, s
még ma is mindúntalan megyen.
Vas Laczi levétette két bátyját a kürtő alól, hol egészen
kiaszának, megfeketűltenek, s elvitte a Kígyókirályhoz, hogy hozza
életre. Ment vele testvére is, a csillagruhájú.
Elérvén a Kígyókirályhoz, mellette egy gyönyörű királyleányt
láttanak űlni; csillag vala homlokán, s a Kígyókirály így szóla: Ez
nekem leányom, s őtet te mentéd meg. Vedd feleségűl. Nagy
lakodalmat tarta nekik. A két fiú és a leány elmenének atyjokhoz;
Vas Laczi veszteg maradt, s a Kígyókirály neki sok pénzt, sok
gyöngyöt, gyémántot hagya, s a földnek nem volt gazdagabb királya
mint Vas Laczi.
A SÓBÁNYÁK.