University of Birmingham AISB/IACAP World Congress 2012 - Alan Turing 2012

2-6 July 2012

Home

Symposia

Invited Speakers

Schedule

Venue

Registration Accommodation

FAQ

Submission

Organization

Important Dates

Poster

Campus Pictures

Contact

(Call for Proposals for Symposia)

Keynote Talk by
AARON SLOMAN

Affiliation

Honorary Professor, School of Computer Science, University of Birmingham, UK

Title

Varieties of Meta-Morphogenesis in the Bootstrapping of Biological Minds

Abstract

Evolution has produced mechanisms for producing new forms of information processing in biological entities. One such mechanism is natural selection itself, which depends on structures and processes available in the physical universe to generate new opportunities. Others are learning mechanisms in individuals. Some are social mechanisms by which information is collected, shared and used, e.g. using pheromone trails, swarming mechanisms, and audio warnings.

Some changes are transient -- depending on immediately available information, while others are long term changes in how information is processed. Long-term changes can be produced by ontological extensions, stored generalisations, new procedures, values and goals, or new mechanisms, forms of representation and architectures. Some mechanisms produce fundamental changes in the underlying competences, including changes in the ability to learn, such as differences between species with very different learning capabilities, and the differences between the learning capabilities of a newborn human infant and that same individual as a post-doctoral researcher in mathematics.

Turing's 1952 paper on morphogenesis proposed mechanisms for producing new physical structure. We can extend the label 'morphogenesis' to refer to mechanisms for producing new information processing, of which current AI learning systems are examples. In some cases, the changes produced are changes in types of morphogenesis. We can use "meta-morphogenesis" to refer to processes that alter the processes of morphogenesis of information processing. Familiar special cases are evolution of evolvability, and learning to learn. In some cases architectural change is required, e.g. new mechanisms for self-monitoring and self-modification.

Useful boot-strapping competences must have been acquired piece-meal over millions of years, with earlier versions enabling evolution of new forms, by providing new "platforms" on which evolution could build, under changing pressures from the environment (physical, biological, and social): producing more new forms of morphogenesis.

A key feature of complex human-created information processing systems has been increasing reliance on stacked layers of virtual machinery and coexisting interacting virtual machines, including self-monitoring and self-modulating virtual machines. Some of these could never have been implemented purely AS physical machines, even though they are implemented IN physical machines (and partly also in the environments of the machines). I suggest that evolution got there first.

Some of the information contents of self-monitoring virtual machines have important analogies with contents of minds (including the impossibility of physical identification from outside the machine, and other sorts of 'privacy'). This may provide clues to meta-morphogenetic steps required for evolution of biological minds.

The study of meta-morphogenesis may reveal how, over generations, interactions between changing environments, changing animal morphologies, and previously evolved information-processing capabilities, might combine to produce increasingly complex forms of "informed control", from microbes to internet-based learning communities. In particular, human mathematical capabilities could be both a product and a source of meta-morphogenesis.

Developing these ideas is a worthy Turing-inspired multi-disciplinary project for the 21st Century and beyond, unifying science, engineering and philosophy.

These ideas are expanded here.

Date

Thursday, 5 July 2012, 6pm