CCNR Talks

Artificial Ecosystem Selection: Simulation, experiment and some interesting questions

Alex Penn: SENSe, University of Southampton (formally CCNR Sussex)

15.11.2006 Alergic Seminar

Play or download talk + discussion here

In this talk I will discuss my work on artificial ecosystem selection, that is, selection on the level of 'ecosystems' or multi-species communities in order to alter 'ecosystem-level properties'. First shown to be effective in artificial selection experiments on microbial ecosystems (1,2), a response to ecosystem selection has been demonstrated, but not the mechanisms by which it operates. Ecosystem selection can be considered using the specific requirements that must be met by a population of units in order for a response to selection to occur (3). I will discuss the concepts of variation, heredity, and phenotype in the ecosystem context, along with possible mechanisms via which ecological systems could satisfy these criteria. It has been suggested that possible novel sources of higher-level heritable variation, qualitatively different to those which exist at the individual level, could exist in ecological systems, and I will present some simple models which demonstrate these dynamics (4,5). These possibilities bear interesting comparison to the origin of new levels of inheritance during major transitions in evolution (6).

I will also present the results of the first field-trial of this technique, practical experiments on using ecosystem selection to improve the growth of L. culinaris in a semi-arid environment with degraded soil. It is hoped that the ecosystem selection technique can be developed as a tool to 'design' bespoke ecosystems for practical applications in bioremediation, agriculture and waste treatment.

1. Swenson, W. and Wilson, D.S. and Elias, R. (2000) Artificial Ecosystem Selection. PNAS, 97: 9110
2. Swenson, W. and Arendt, J. and Wilson, D.S. (2000), Artificial selection of microbial ecosystems for 3-chloroaniline biodegradation, Environ. Mirobiol., 2: 9365
3. Lewontin, R. (1970). The Units of Selection, Annual. Rev. Ecol. Syst.,17:83
4. Penn, A.S. (2003) Modelling Artificial Ecosystem Selection: a preliminary investigation, In proc. Advances in Artificial Life, 7th European Conference, ECAL 2003.
5. Penn, A. and Harvey, I. (2004), The role of non-genetic change in the heritability, variation, and response to selection of artificially selected ecosystems, In proc. Artificial Life IX: Proceedings of the Ninth International Conference on the Simulation and Synthesis of Life.
6. Maynard Smith J. and Szathmary, E. (1995), Major Transitions in Evolution, Spektrum.

The role of AI in the paradigm shift towards an embodied-embedded cognitive science

Tom Froese (CCNR)

15.11.2006 Life and Mind Seminar Series

Play or download talk + discussion here

In this talk I will look at the role of AI and robotics in the ongoing paradigm shift of the cognitive sciences. I will start by analyzing the shift from cognitivism towards a more embodied-embedded cognitive science, and conclude that this move was mostly due to empirical developments in AI which allowed purely philosophical arguments to be resolved experimentally.
More recently, there has been a lot of discussion of a further shift towards enactivism. I will try to give a brief overview of the central aspects of an enactive theory of cognition. In particlar, I will highlight enactivism's central concern with the notion of subjectivity and then analyze AI and robotic's role (if any) in facilitating this second shift. It might turn out that this shift in understanding needs something qualitatively different from the resolution of disputes in either the philosophical or empirical domain of the cognitive sciences.

Information Flows in the Perception-Action Loop

Daniel Polani (University of Hertfordshire)

08.11.2006 Alergic Seminar

Sorry, no video.

Information is an essential and omnipresent resource and has long been suspected as a major factor shaping the emergence of intelligence in animals and as a guideline to construct artificial intelligent systems. In search for fundamental principles guiding the self-organization of neural networks, as has long been suspected by Barlow (1959, 1989, 2001), Linsker (1988) formulated a number of information-theoretic hypotheses. His model (and most of its successors) was purely passive. However, recent work by Touchette and Lloyd (2000) extending early work by Ashby (1953), as well as some work by Polani et al. (2001) has shown that actions can be incorporated into the information-theoretical analysis.
As was found by Klyubin et al. (2004), incorporating actions into an information-theoretic formalization of the perception action-loop of agents has dramatic consequences in terms of self-organization capabilities of the processing system. As opposed to Linsker's model which required some significant pre-structuring of its neural network, this new model makes only minimal assumptions about the information processing architecture.
The results demonstrate how the agent's embodiment, i.e. the coupling of its sensors and actuators to the environment, is entirely sufficient to give rise to structured pattern detectors driven by optimization principles applied to the information flow in the system. Since for agents acting in a world fundamental embodiment comes for free, above observations make the impressive success of morphological computation (Pfeifer 2000, 2006) in its interaction with the agent's control conceptually more transparent.
In addition, the presented formalism ties in naturally with other approaches utilizing Shannon information to shape behaviour (Sporns and Lungarella 2006; Prokopenko et al. 2006). In particular, Klyubin et al. (2005) showed that the formalism can be naturally used to implement information-based taskless utilities, closely related to other self-motivated learning paradigms (Der 2001; Bovet and Pfeifer 2005; Sporns and Lungarella 2006; Prokopenko et al. 2006).
In the present talk, we will motivate Shannon information as a primary resource of information processing, introduce a model which allows to consider agents purely in terms of information and show how this model gives rise to the aforementioned observations. If there is time, the talk will discuss the use of information-theoretic methods to structure the information processing also in real robot systems.

Play as re-creation: an enactive route to human sense-making

Ezequiel Di Paolo (CCNR)

08.11.2006 The Mind and Life Seminars

Play or download talk + discussion here

A new philosophy seminar/reading/discussion group organized by members of the CCNR but open to everyone will start meeting regularly on Wednesday afternoons to discuss topics related to cognitive science, philosophy of mind, phenomenology, psychology, robotics and philosophy of biology in a relaxed and friendly atmosphere. The aim is to provide a constructive and collective learning environment for the discussion of complex ideas at the forefront of modern perspectives in the sciences and philosophies mind and life. This will be done in a seminar-style meeting featuring presentations, demonstrations, open discussion (and perhaps even practical sessions). We also aim at facilitating dialogue and exploring possible collaborations between people engaged in theoretical, modelling and empirical work.

Early A-life Art at the Slade School of Fine Art

Paul Brown (CCNR)

01.11.2006 Alergic Seminar

Sorry, no video.

The Experimental and Computing Department of the Slade School of Fine Art, University College London, was founded by Malcolm Hughes in 1972 and lasted until 1982. In 1974 it became the first art dept. in the world to have a formal computer program when it acquired a Data General Nova II minicomputer. It became a major European focus for artists interested in the potential of computer systems. Major influences included System (or procedural) art, deterministic chaos and cellular automata. The work done at the Slade can now be recognised as an early root of the current interest in computational and generative art as well as the scientific pursuit of a-life.
This talk will describe the work undertaken at the Slade and some of the key figures involved like Edward Ihnatowicz (SAM - the Senster), Harold Cohen, Chris Briscoe, Julian Sullivan and Darrel Viner.
Paul Brown is an artist and writer who has been specialising in art & technology and specifically the computational arts since the late 1960's. He is currently a Visiting Professor in Informatics where he is working on the DrawBots project.

Minimal developmental systems

Rachel Wood (CCNR)

25.10.2006 Alergic Seminar

Play or download talk + discussion here

This talk presents work on a new approach to modelling developmental processes in animals and infants. The work is founded on developmental systems theory (DST), a broad perspective on ontogenetic adaptation in which it is proposed that developing agents and their contexts should be treated as systems in which neither genetic material nor environmental factors can be given causal primacy. In this approach we use evolutionary robotics, specifically Beer's `minimal systems' methodology (Beer, 1996; Slocum, Downey and Beer 2000) to build `minimal developmental systems'.
The talk will introduce principles of DST and present two models constructed using the proposed minimal developmental systems approach. In the first model, an extension of Di Paolo's acoustic approach paradigm, simple simulated agents learn coordinated patterns of signalling and movement in order to solve the experimental task. The results suggest that it is possible to obtain a distributed form of developmental system in which two agents undergo coupled processes of ontogenetic adaptation.
The second model to be presented is a minimal implementation of Piaget's delayed manual search task (Piaget 1952: 1980), also known as the A not B error. In this model we demonstrate a homeostatic mechanism for mediating synaptic plasticity and present results in which simulated agents produce similar performance errors to those observed in human infants. These results suggest an empirical hypothesis about the role of similar organizing principles for developmental plasticity underpinning the balance between change and conservation. The talk concludes with a brief discussion of future directions for the work.

Evolving Reaction-diffusion Controllers for Minimally Cognitive Animats

Kyran Dale (CCNR)

18.10.2006 Alergic Seminar

Play or download talk + discussion here

This talk describes work carried out to investigate whether a classic reaction-diffusion (RD) system could be used to control a 'minimally cognitive' animat. The reaction-diffusion system chosen was that first described by Gray and Scott (Gray-Scott) and the minimally cognitive behaviors those used by Beer et. al involving the fixation and discrimination of diamond and circle shapes by a whiskered animat. The parameters of this RD-controller were evolved using an evolutionary, or genetic, algorithm (GA).

Variadic Neural Networks: Networks for Unordered Variable-Length Data

Simon McGregor (CCNR)

11.10.2006 Alergic Seminar

Play or download talk + discussion here

A number of interesting statistical and geometric functions operate on unordered data of variable dimensionality. It is difficult to operate on this class of data using standard neural network architectures, which presuppose fixed-dimensional input data. In this talk, I will describe a novel network architecture specifically designed for variadic (variable arity) data; I will discuss the computational power of this class of networks, and also present some experimental results.

From Homeostasis to Sensory motor Couplings

Takashi Ikegami (University of Tokyo)

04.10.2006 Alergic Seminar

Play or download talk here

Play or download discussion here

We know how sensory-motor couplings can simulate primitive but a variety of cognitive behaviors. But the origin of sensory-motor coupling itself is remained as an open question. We tackle this problem by re-visiting the problem of the origin of life forms both theoretically and experimentally. I insist that dynamic homeostasis is the key notion that bridges the gap between self-preservation and sensory-motor coupling.
Autopoiesis theory tells us that life form emerges when a spatial region generates a self-boundary and starts to maintain by itself. My first talk is to introduce a simple chemical experiment of an oil droplet that generates a surfactant boundary and showing self-motility with it.
In order to see how homeostatic behavior leads to sensory-motor couplings theoretically, we generalized Inman Harvey's simple Daisy World agent to explore the world. The agent's surface is covered with two kinds of daisies that cooperatively adjust the surface temperature. We showed how the agent can maintain the surface temperature constant irrespective of the environmental temperature and how explorative behavior can be generated at the same time.
My simulation experiment showed that the homeostatic state is maintained by "Homeochaos" and a partially ordered chaotic state can generate a coherent and explorative movement in the inhomogeneous environment. This is going to be my second talk.

Programmable Springs: Developing Programmable Compliance Actuators for Autonomous Robots

Bill Bigge (CCNR)

03.10.2006 COGS Research Seminar

Play or download here (unfortunately only about a minute of the talk was recorded)

Conventional approaches to actuation and motion control are designed to eliminate any perturbations from the system and provide smooth precise control of speed or position and a high level of stiffness. By contrast, emerging approaches to autonomous robotics rely on exploiting the environment to aid motion. In passive dynamic systems motion is modulated by interactions between the mechanism and the environment; instead of forcing the actuators to follow pre-planned trajectories the environment is used to guide motion. Developing real robots that can exploit these dynamics requires the use of actuators that can react to the environment, exhibiting behaviour that varies from high stiffness to complete compliance or zero impedance. I will outline our design for an electric actuator, called a programmable spring, which can be configured to emulate complex spring, damping and zero impedance systems within its range of movement and mechanical limits. This design forms the basis for a prototype actuator intended as a cost effective 'off the shelf' component for robotics development. Our design includes a sophisticated high level control architecture that allows the actuator to exhibit complex multimodal behaviour whilst offering the user a high degree of control.