Non-Symbolic AI Lecture 12

Slides:



Advertisements
Similar presentations
Homeostasis Walter Cannon 1932 The Wisdom of the Body Jame Lovelock ~1969 Gaia hypothesis (Lovelock, J.E.; Margulis, L. (1974). "Atmospheric homeostasis.
Advertisements

EASy 15 Sept 2004Artificial Life 9, Boston1 Artificial Life 9 Homeostasis and Rein Control: From Daisyworld to Active Perception Inman Harvey Evolutionary.
Daisy World an introduction to systems and equilibria.
An Introduction to Systems 1. What are systems? What are feedback loops? What are equilibrium states? Does viewing Earth as a system allow for deeper.
A state of balance in the body
Stimulus/Response.
Buffers and Feedback Loops Class Notes 1:6. Buffers and Feedback Loops Buffers are agents or regulatory mechanisms that reduce or minimize fluctuations.
Wiring the nervous system On average, a single neuron: - makes ~1000 contacts - receives ~10000 contacts.
GAIA HYPOTHESIS  Created by James Lovelock in 1969 (but not published until 1979)  Named after the Greek Goddess Gaia who was the Earth Goddess.
Systems 3. Feedbacks Coffee System Tc TRTR The hotter the coffee is The faster it cools Tc-T R Cooling Rate The higher the cooling rate The colder the.
THE GAIA HYPOTHESIS EXPLORATION OF DAISYWORLD. What is the Gaia Hypothesis? Life itself is responsible for maintaining the stability of Earth’s climate.
EASy 9 March 2005activate.d workshop1 activate.d Workshop Homeostasis and Dynamical Representations Inman Harvey Evolutionary and Adaptive Systems Group.
What is the importance of climate as a direct and indirect influence on the environment and people?
NANIA 2D - Daisyworld Graeme Ackland (physicist) Tim Lenton (ecologist) Michael Clark (project student) A model planet showing coupling between life and.
Earth Systems Science or Gaia. A new/different kind of ecology Holistic science, is an approach to research that emphasizes the study of complex systems.
Daisyworld.
Improving Stable Processes Professor Tom Kuczek Purdue University
Chapter 2: Theory and Research 1. Theories and our Understanding Psychoanalytic Theory - Freud Psychosocial Theory – Erikson Object Relations Theory Behavioral.
Daisyworld.
Modeling the Gaia Hypothesis: Daisyworld Phillipa Sessini.
The GAIA hypothesis: a New Look at Life on Earth CSCI 1210 Fall 2003.
Systems and Models What is a system? What is a model? Feedback Mechanisms Transfer vs. Transform Laws of Thermodynamics.
Introduction to Systems /Daisyworld
1 MET 112 Global Climate Change MET 112 Global Climate Change - Lecture 9 Daisyworld Eugene Cordero San Jose State University Outline  Introduction 
Daisyworld life regulating climate Temperature Time In the beginning of Daisyworld, God created a young star and a barren planet, cold and distant.
What is Control System? To answer this question, we first have to understand what a system is Simon Hui Engineer Control and Informatics, Industrial Centre.
An Introduction to Systems. The Climate System We will often refer to the “Climate System” Can you name the components of the climate system?
University of Windsor School of Computer Science Topics in Artificial Intelligence Fall 2008 Sept 11, 2008.
Systems and Models What is a system? What is a model?
An Introduction to Systems
Essential Questions What is biology? What are possible benefits of studying biology? What are the characteristics of living things? Introduction to Biology.
GAIA A Totally New Look at Life on Earth: The Earth as an Organism.
 Method for studying the natural world.  Nature works on rules –simple ones and complex ones.  If we study patterns, science can be applied to everything.
بسم الله الرحمن الرحيم وبه نستعين
A state of balance in the body Whoa Notes. Definition of Homeostasis homeo = same; stasis = standing Homeostasis is the term we use to describe the constant.
CHAPTER 4 Designing Studies
INTRODUCTION TO PHYSIOLOGY
Open loop vs closed loop
Life in the earth system
Introductory Physiology Biol 141
Some General Concepts of Point Estimation
HOMEOSTASIS.
Equilibrium, Positive and Negative Feedback
ECE 382. Feedback Systems Analysis and Design
Terra as a self-regulating living entity
Biology: The Study of Life
Background Information
Simple Linear Regression
Simple Linear Regression
GAIA HYPOTHESIS Created by James Lovelock in 1969 (but not published until 1979) Named after the Greek Goddess Gaia who was the Earth Goddess.
Model the non-negative even integers
GAIA HYPOTHESIS the idea of the Earth as a single living superorganism
Environmental Systems
Get Started Abiotic Factor: A Non-living chemical and physical part of the environment that affects living organisms and the functioning of ecosystems.
The scientific study of life
Significance Tests: The Basics
Chapter 02 Lecture Outline
What is fusion and how is it a factor for the life on earth?
Chapter 10: Estimating with Confidence
Non-Symbolic AI Lecture 13
THE CHARACTERISTICS OF LIFE
Environmental Systems
An Introduction to Systems
Organisms & Life.
Homeostasis Read page 423 and brainstorm answers to questions.
Biology 12 An Introduction.
History of Earth.
Homeostasis.
What is a System? Definition: A system is a group of different components that interact with each other Example: The climate system includes the atmosphere,
Some General Concepts of Point Estimation
Presentation transcript:

Non-Symbolic AI Lecture 12 Double bill with lec 13: This lecture -- Homeostasis, The Dynamics of Daisyworld, Enactive perception. Next lecture – The Dynamics of Robots and the Rights and Wrongs of Representations Non-Symbolic AI lecture 12 Summer 2004

Things versus Processes For 100 years or more, there has been a clash of views, or rather of starting points, between “A perception is a thing” versus “Perceiving is a process” (Also substitute in this phrase:- Thought vs. thinking, memory vs. remembering, representation vs. representing etc etc) Non-Symbolic AI lecture 12 Summer 2004

Non-Symbolic AI lecture 12 Enactive Perception I take the starting point of Enactive Perception to be: “Perceiving happens when an interested agent engages with a world it is interested in” So firstly, think dynamics rather than things (dynamical systems approach). Secondly, think seriously about what an agent, a world, engagement is – don’t take anything for granted Non-Symbolic AI lecture 12 Summer 2004

A Rough Hierarchy “What is a an Object, that a Creature may perceive it, and a Creature, that it may perceive an Object” (adapted from Warren McCulloch) All organisms react to perturbations here and now Some organisms engage with objects distant in space/events distant in time Some organisms engage in a social world “Static Vegetables”, Homeostasis “Mobile Animals”, Perception “Social Humans”, Representation Non-Symbolic AI lecture 12 Summer 2004

Non-Symbolic AI lecture 12 Lecture Plan Lec 12: Homeostasis – a new and simplified daisyworld model, with Rein Control. Lec 12: Perception -- extending the underlying mathematical insight to a novel form of (en-)active perception. Lec 13: Dynamics of Robots. Lec 13: Representation – the rights and wrongs of Representation-language in the context of understanding perception. Non-Symbolic AI lecture 12 Summer 2004

Non-Symbolic AI lecture 12 Homeostasis and Gaia To stay alive, an organism has needs – it must maintain various values such as internal temperature, chemical concentrations, etc, within bounds of viability and comfort. Not too hot, not too cold. Homeostasis is the ability to react to perturbations (e.g. of temperature) so as to maintain it within viability limits. The Gaia Hypothesis (Lovelock 1972) is that the Earth with its biota acts homeostatically, similarly to an organism, to maintain global properties within ranges appropriate for the biota. Teleology?? Non-Symbolic AI lecture 12 Summer 2004

Non-Symbolic AI lecture 12 For example … ... The sun has steadily increased in heat output over the lifetime of the Earth (say 30% less luminous, 3.8bn years ago). One should normally expect the Earth’s temperature to have started off far far colder, and increased until now it was far too hot for us (say currently around 2900 C) But it seems the Earth’s surface temperature has been maintained at around 200 C for aeons. A nice temperature! How? Non-Symbolic AI lecture 12 Summer 2004

Non-Symbolic AI lecture 12 Daisyworld In 1983 the ‘parable of Daisyworld’ was presented as a possible mechanism, explaining how feedback loops between living organisms and the environment can produce regulation, homeostasis. Suppose 2 species of Daisies, Black and White, live on a Grey planet. They each have a preferred temp, and will die if temp is too different. Non-Symbolic AI lecture 12 Summer 2004

Non-Symbolic AI lecture 12 What Happens? Temperature Solar output The average temp of Earth remains in the viability zone far more than one might expect. Somehow the Earth+biota is self-regulating. How? Non-Symbolic AI lecture 12 Summer 2004

Non-Symbolic AI lecture 12 Mystery ? The Earth+biota seems to self-regulate at a temp appropriate for biota. Teleology? Planning? Surely not… Only other available explanation for purposive behaviour seems to be Darwinian evolution -- but there hasn’t been a population of evolving Earths. Daisyworld model shows how you can get homeostasis without evolution or any teleology. Non-Symbolic AI lecture 12 Summer 2004

New simplified model of Daisyworld Combination of a Hat-function and any feedback gives regulation Hat-function = viability zone Simplest version = Witch’s Hat Non-Symbolic AI lecture 12 Summer 2004

Simplified Feedbacks in Daisyworld Non-Symbolic AI lecture 12 Summer 2004

Initially consider just one Daisybed Rate of change of Temp = albedo * (Suntemp – Temp) –Temp + Feedback-term * Hat-function(Temp) Non-Symbolic AI lecture 12 Summer 2004

Non-Symbolic AI lecture 12 Equilibrium when … LHS = 0 when a Linear function in T intersects a Hat-function of T Non-Symbolic AI lecture 12 Summer 2004

Extends the viability zone B unstable A, C are stable equilibria Without or With Feedback Non-Symbolic AI lecture 12 Summer 2004

Non-Symbolic AI lecture 12 Rein Control So a Hat-function plus Positive feedback extends the viability zone to the left. And a Hat-function plus Negative feedback will extend the viabilty zone to the right. Different types of feedback needed for each direction – to regulate in both directions you need 2 “Reins”. A rein can pull but cannot push – “Rein Control” (Manfred Clynes 1969) Non-Symbolic AI lecture 12 Summer 2004

Non-Symbolic AI lecture 12 Both together ? What interactions are there between Black and White Daisies – what transfer of heat, or leakage? Non-Symbolic AI lecture 12 Summer 2004

Fully connected is a bad idea If B and W daisies are at same temperature, there will be same number of each. 1 B + 1 W = 1 Grey, same as a lifeless planet Non-Symbolic AI lecture 12 Summer 2004

Loosely Coupled is what you want Non-Symbolic AI lecture 12 Summer 2004

Non-Symbolic AI lecture 12 Simple Daisyworld This simplified Daisyworld exposes the underlying mechanism much more clearly, Virtually any combination of a Hat-function with any (monotonic) feedback in effect extends the range of the Hat-function, or viability zone. This true for any organism, not just Gaia-as-an-organism. Some forms of homeostasis, of an organism reacting “appropriately” to environmental disturbances, are cheap. Proto-perception ? Non-Symbolic AI lecture 12 Summer 2004

Dynamical Systems viewpoint Note that the “organism” does not regulate by comparing observed temperature with an internal representation of the desired temperature. Rather, the regulating-behaviour is the outcome of the dynamics of feedbacks both internal and external, between organism and environment, settling into an attractor. Non-Symbolic AI lecture 12 Summer 2004

Non-Symbolic AI lecture 12 Lecture Plan Lec 12: Homeostasis – a new and simplified daisyworld model, with Rein Control. Lec 12: Perception -- extending the underlying mathematical insight to a novel form of (en-)active perception. Lec 13: Dynamics of Robots. Lec 13: Representation – the rights and wrongs of Representation-language in the context of understanding perception. Non-Symbolic AI lecture 12 Summer 2004

Extending to (En-)active Perception Rein control = Hat-function + Feedback Hat-function = “viability zone” But also Hat-function = directional response of a photoreceptor Non-Symbolic AI lecture 12 Summer 2004

Non-Symbolic AI lecture 12 Let’s build a 2-D Dalek Photoreceptor A at end of moving tentacle, Hat-function receptive field. Sensor response generates torque D, against Spring B to nose C. Dalek is only able to rotate around its centre. Non-Symbolic AI lecture 12 Summer 2004

Let’s put it together randomly Now let’s have 100 such tentacles, all independent bar springs connecting to nose Let’s have directions of D (anti- or clockwise) at random. Angles of acceptance of A at random Strengths of springs B at random Non-Symbolic AI lecture 12 Summer 2004

Comparison with Daisyworld Angle of Acceptance A = Hat-function = “viability zone” Direction of D = +ve or –ve Feedback = B or W Daisies Springs B = loose semi-coupling between different feedbacks Non-Symbolic AI lecture 12 Summer 2004

What behaviour to expect? Underlying maths is the same. Instead of regulating temperature for daisies to stay in their viability zone, this will regulate tentacle-angles to stay within their acceptor-zones. Collectively, they will pull the nose around to point towards a light -- Phototaxis, despite the random wiring up of tentacles (… and actually, you could get rid of the nose altogether …) Non-Symbolic AI lecture 12 Summer 2004

Non-Symbolic AI lecture 12 Phototaxis Waggle point-source of light in front – it’ll pick it up and pursue it. Robust to 3 orders of magnitude on angles of acceptance, to 2+ orders of magnitude on spring constants/torque parameters. Non-Symbolic AI lecture 12 Summer 2004

Non-Symbolic AI lecture 12 Incidentally … … Samples from data from all the “left-moving” tentacles (normalised) – (almost) all on the RHS of the Witch’s Hat … … … … all at positions like C. And vice versa for all the “right-moving” tentacles. Non-Symbolic AI lecture 12 Summer 2004

Dynamical Systems viewpoint Note that the “organism” does not regulate by comparing observed temperature with an internal representation of the desired temperature. light-direction light-direction Rather, the regulating-behaviour is the outcome of the dynamics of feedbacks both internal and external, between organism and environment, settling into an attractor. Non-Symbolic AI lecture 12 Summer 2004

Non-Symbolic AI lecture 12 Lecture Plan Lec 12: Homeostasis – a new and simplified daisyworld model, with Rein Control. Lec 12: Perception -- extending the underlying mathematical insight to a novel form of (en-)active perception. Lec 13: Dynamics of Robots. Lec 13: Representation – the rights and wrongs of Representation-language in the context of understanding perception. Non-Symbolic AI lecture 12 Summer 2004