http:wetenschap'vpro'nlprogrammanoorderlichtindex'shtml2785571 2848322 3855404 4264831

1
Is sex nodig?
Noorderlicht, zie ZOS repository op blackboard
Bioloog Bert Kempenaers, in 1993 verbonden aan de
Universiteit van Antwerpen, gebruikte 'DNA
fingerprinting' om het gedrag te onderzoeken van
een populatie pimpelmezen in een bos. De
vogeltjes leken aanvankelijk zo aandoenlijk een
paartje komt bij elkaar, ze bouwen een nest en
gezamelijk zorgen ze voor het nageslacht.
Kempenaers ontdekte echter dat de waarheid verre
van romantisch was. De pimpelmeesvrouwtjes
blijken er bepaald niet vies van om even bij de
buurman aan te wippen als hun eigen mannetje niet
oplet. Soms laat ze zich zelfs uitsluitend
bevruchten door de mannetjes uit de aangrenzende
territoria. De eieren die ze dan legt, brengen
dan ook geen kroost voort van haar eigen mannetje
- die de jonge vogeltjes desondanks keurig
verzorgt en van voer voorziet, als ware het zijn
eigen jongen. De bioloog onderzocht
verschillende kenmerken van de volwassen
mannetjes, en kwam tot de slotsom dat de
populairste pimpelmezen gemiddeld iets langere
pootjes hebben. Of dat ook echt het kenmerk is
waar de pimpelmeesvrouwtjes op vallen, is
onduidelijk. Wel blijken de jongen van de
langpoters een grotere kans te hebben om te
overleven. Van monogaam gedrag onder pimpelmezen,
stelt Kempenaers, is dus geen sprake. In de
populatie die hij onderzocht ging zo'n dertig
procent van de vrouwtjes regelmatig vreemd, wat
er toe leidde dat rond de tien procent van alle
jongen een andere biologische vader heeft dan het
mannetje dat hen verzorgt. Overigens heeft later
onderzoek, daterend van na 1993, bij mensen
precies dezelfde cijfers laten zien in ieder
geval in de Verenigde Staten en Groot-Brittannie
is tien procent van de kinderen niet verwekt door
de man die hen als vader opvoedt, meestal zonder
dat de man (en de kinderen) dat zelf weten.
http//wetenschap.vpro.nl/programma/noorderlicht/i
ndex.shtml?2785571284832238554044264831
2
Zelf Organiserende Systemen (ZOS) - Self
Organising Systems - Vrij Universiteit
Amsterdam Lecture 7 Self Organisation and
AI http//www.cs.vu.nl/zos Martijn
Schut schut_at_cs.vu.nl
3
SOS - Lecture 7
Overview of the Lectures
Introduction Insect-Based Computing
I Insect-Based Computing II Cellular
Automata Criticality DNA and Evolution Self
Organisation and Artificial Intelligence Self
Organising Maps Dynamical Systems
4
SOS - Lecture 7
Overview

• Artificial Intelligence and Self Organisation
• Rule Based Agents
• The Echo System (John Holland)
• Prisoner's Dilemma in Echo
• Summary of Self Organisation course
• Insect-Based Computing
• Cellular Automata
• Evolution
• (15.30 - 16.30) question hour about Mystery

5
SOS - Lecture 7
AI - Brief Selected History
40s, 50s
60s, 70s
80s
90s, 00s
6
SOS - Lecture 7
AI - Brief History of AI
1940s, 1950s Turing Machines 1960s,
1970s Expert Machines (GPS) 1980s Reactive
Intelligence 1990s, 2000s Multi Agent Systems
7
SOS - Lecture 7
AI - Multi Agent Systems
order
Jack
Anne
Xerox
www
Pete
packing
truck
Rob
books
http//www.amazon.com
8
SOS - Lecture 7
AI - Multi Agent Systems
Goals
Abilities
Agent
Beliefs
Etc...
DistributedIntelligence
9
SOS - Lecture 7
AI - Multi Agent Systems
Observe

World
Act
10
SOS - Lecture 7
AI - Multi Agent Systems
Whats in here?

IF THEN IF THEN IF THEN IF THEN
IF THEN IF THEN IF THEN
IF ? condition (observe) THEN ? response
(act)
11
SOS - Lecture 7
AI - Rule-Based Agents
IF stimulus s occurs THEN give response r For
example IF I am hungry, THEN go eat something IF
I am tired, THEN go to bed IF I see an apple,
THEN eat it etctera
12
SOS - Lecture 7
AI - Rule-Based Agents
Complex Adaptive Systems (CAS)
13
SOS - Lecture 7
AI - Rule-Based Agents

• Foundations of CAS
• Aggregation
• Tagging
• Nonlinearity
• Flows
• Diversity
• Internal Models
• Building Blocks

Important!
14
SOS - Lecture 7
AI - Echo

• A grid world, where each site contains resources
  and agents.
• Resources are represented as a,b,c,d,. Each
  site may have a fountain that provides resources
  on each time-step.
• Agents have structures represented by stringing
  resource letters together, called chromosomes.
  Additionally, an agent has a reservoir for
  storing resources.
• The chromosome of an agent consists of a tag
  segment and a control segment.
• The tag segment of a chromosome contains three
  tags, namely offense, defense and adhesion.
• The control segment contains three kinds of
  objects conditions, resource transformations,
  and an activity marker. Then there are three
  kinds of conditions exchange, mating and
  replication.

15
SOS - Lecture 7
AI - Echo

• Echo should be as simple as possible
• Echo should be designed so that the actions of
  its agents can be understood in a variety of cas
  domains
• Echo should facilitate experiments on the
  evolution of fitness
• The primitive mechanisms in Echo should have
  ready counterparts in all cas
• Echo should be designed to incorporate
  well-known models of particular cas wherever
  possible
• One should be able to mathematically analyse as
  many aspects of Echo as possible

16
SOS - Lecture 7
AI - Echo

• focus on information
• inputs to the system (observe) are handled by
  detectors
• outputs of the system (act) are handled by
  effectors
• silimar to program language message processing
  rules
• if receive message X, then send message Y
• Rules are building blocks and hypotheses in case
  of conflict
• Strength of a rule is based on the utility of
  the rule in the past

17
SOS - Lecture 7
AI - Echo - Example

• Prisoners Dilemma (Game Theory)
• Two persons charged with committing a robbery
• Each person can either cooperate (refuse to
  confess) or defect (confess) without knowing what
  the other person does.
• If both confess, they each get four years
  prison.
• If both refuse to confess, they get two years.
• If one confesses and the other one refuses, the
  confessor will make a deal and walk free, while
  the refuser gets jailed for five years.

18
SOS - Lecture 7
AI - Echo - Example
Second Player Cooperate Defect Firs
t Player Cooperate 3, 3 0, 5 Defect 5,
0 1,1
19
SOS - Lecture 7
AI - Echo - Example

• Conclusions
• Agents use different strategies
• tit-for-tat Do unto others as they do unto you
• not much interesting happens in the world
  without tags
• without tags, interactions are largely
  defect-defect
• with tags, some interesting patterns arise
• agent employs tit-for-tat, and (2) prefers to
  play with other agents that also employ
  tit-for-tat
• community of these tit-for-tat agents develops

20
SOS - Lecture 7
Economics - Sugarscape

• Developed by Epstein and Axtell
• Social Science bottom-up

"On a small, bagel-shaped planet, tribes of
natives -- collectively known as "agents" -- go
about their lives. They reproduce. They eat. They
travel. They squabble over limited resources, or
trade them if doing so is to their mutual
advantage. They exist entirely in a computer.
Welcome to Sugarscape."
21
SOS - Lecture 7
Economics - Sugarscape
Click here for an introduction to Sugarscape
22
SOS - Lecture 7
Economics - Sugarscape
What if there are two different types of agents?
Click here to see what happens
23
SOS - Lecture 7
Economics - Sugarscape
What if we added an extra resource, spice?
Click here to see what happens
24
SOS - Lecture 7
Economics - Sugarscape
Can we do other things with Sugarscape?
Yes we can, click here
25
SOS - Lecture 7
Economics - Sugarscape
Can Sugarscape be implemented in Starlogo?
Yes, of course, Starlogo can do everything! Click
here
26
SOS - Lecture 7
Self Organisation and AI

• Topics covered
• Rule-Based Agents
• Economics

27
SOS - Lecture 7
Self Organisation - An Overview

• Insect-Based Computing (Lectures 2,3)
• Cellular Automata (Lecture 4)
• Evolution (Lecture 5,6)
• Summary, Questions, Remarks, Conclusions

28
SOS - Lecture 7
SOS - Overview - Insect-Based Computing

• Ant foraging
• Division of Labour
• Brood Sorting
• Cooperative Transport

29
SOS - Lecture 7
SOS - Overview - Insect-Based Computing
Foraging Ants
30
SOS - Lecture 7
SOS - Overview - Insect-Based Computing

• Absence of centralized control
• Global order
• Redundancy
• Adaptation

31
SOS - Lecture 7
SOS - Overview - Cellular Automata

• Life

• A CA is an array of identically programmed
  automata, or cells,
• which interact with one another in a
  neighbourhood and have
• definite state
• Universal Machines

32
SOS - Lecture 7
SOS - Overview - Cellular Automata
transition rules
T 0
T 1

• dies if number of alive neighbour cells lt
  2 (loneliness)
• dies if number of alive neighbour cells gt
  5 (overcrowding)
• lives is number of alive neighbour cells
  3 (procreation)

33
SOS - Lecture 7
SOS - Overview - Cellular Automata

• Ulam
• Turing
• von Neumann
• Conway
• Langton

• Universal Machines
• Turing Machines
• von Neumann Machines
• Game of Life
• Self Reproduction

34
SOS - Lecture 7
SOS - Overview - Evolution

• Reproduction
• Mutation
• Fitness
• Fitness Landscapes

35
SOS - Lecture 7
SOS - Overview - Evolution

• Evolution
• Selection
• Selection with gt 2 species (Huisman)
• Selection and Self Organisation

36
SOS - Lecture 7
SOS - Overview

• Emergence
• Complexity
• Random

37
SOS - Lecture 7
SOS - Overview

• What is part of the system?
• What are the components of the system?
• What is the environment of the system?
• Describe the relation between the components.
• Describe the relation between the system and its
  environment.
• Describe the self-organisation.
• What are the emergent properties?

38
Self Organising Systems

• Master variant of Artificial Intelligence
• Study years 45

39
Karakteriserende vragen

• Hoe kunnen dynamische organisatievormen
  gemodelleerd worden ? Welke methoden en
  technieken worden gebruikt?
• Hoe evolueren organismen om te overleven in
  nieuwe en onbekende situaties (bijvoorbeeld in
  biologie, economie en ecologie)?
• Hoe bereiken we steeds grotere complexiteit met
  behulp van methoden waarmee evolutie-processen
  nagebootst worden (bijv genetische algoritmen)?

40
Uitstraling

• Combineert
• economische, biologische, sociaal-wetenschappelijk
  e basiskennis en vaardigheden
• organisatietheorie, economische modellen,
  populatiebiologie, empirisch denken, spreken en
  schrijven
• hoog beta gehalte
• programmeren,logica, multi-agent systemen

41
Obligatory Courses

• Datamining 4
• Simulatie van Organisaties 3
• Organisatietheorie (SCW) 4
• Zelf-organiserende systemen II 5
• Economische Modellen 5 (Economie)

42
Recommended Courses

• Evolutiebiologie 6
• Evolutionaire Genetica 4
• Neurobiologie van Gedrag 4
• Organisatie en Leiding 5
• Gedrag in Organisaties 4

43
Graduation Projects

• Companies, e.g., TNO, British Telecom, Getronics
  
• At VU-AI department
• Agent Systems Research group
• Computational Intelligence
• Knowledge Management Representation group
• Other universities
• Government