This Frontier Science session
FS4: Evolution's Use of Construction Kits
Thursday 3-5pm in Poynting Physics Room S06
Small lecture rom top floor
This will be an interactive discussion of questions partly inspired by Alan Turing's work on Morphogenesis, published two years before he died. What might Alan Turing have done if he had lived 30-40 years longer?
What sort of school/university education would enable future graduates to contribute to the research project?
What sorts of construction kits produced by biological evolution
made possible evolution of minds, especially mathematical minds?
(Such as the minds of Euclid and his predecessors.)
Could similar artificial construction kits help us build robots with human intelligence?
If we understood more about such construction kits could that help us produce better educated humans, better able to live together and to solve hard problems?
What sorts of construction-kits in brains, or minds, enable a child to assemble new information? Does a child also have to create new construction kits for new kinds of learning? How can a brain do that?
Is there a way to introduce some of these ideas and questions, and later on relevant experience in building computer models, into schools without waiting for top down syllabus changes -- giving future leaders in science, medicine, technology and education a head start, a decade or more before a lumbering "official" educational system wakes up to the opportunity?
Expanded below, in The educational aims of the session.
Presenter: Aaron Sloman
Honorary professor of Artificial Intelligence and Cognitive Science
School of Computer Science, University of Birmingham
This is part of the Meta-Morphogenesis project, summarised here: goo.gl/9eN8Ks
Our school system probably does not provide them any opportunity to build the new ideas formally into their teaching, but it may be relevant to informal discussions about evolution, life, mind, computing, and the future of AI, and in some cases discussions about careers. At the very least this will help to counter the mistaken impression, now being circulated widely, that learning about computing is relevant mainly for people who want to go into industry, possibly via Computing Science degree courses, or who just want to have fun building new apps.
There is a very much deeper reason for teaching computing: namely, our planet contains a very wide variety of information processing systems, including microbes and other organisms, ecosystems, social and economic systems, and perhaps most importantly for us, brains of humans and other animals. If teachers of relevant disciplines, e.g. biology, neuroscience, psychology and philosophy know nothing about forms of computation required for life, thought, perception, reproduction, and development, their pupils may not be properly prepared for research and teaching in these disciplines, just as most professionals in these disciplines have previously been unprepared, through no fault of their own. Likewise future researchers in those fields ignorant of a wide variety of forms of computation and modelling.
Increasingly, people who wish to go into areas of science that study those subjects will need to have personal experience of designing, building, testing, analysing, debugging and comparing working information processing systems, even if only fairly simple ones. Otherwise they will not know how to formulate or to evaluate and challenge new explanatory theories.
Meeting that need will, eventually, require broadening the variety of types of programming taught in schools, and integrating programming with education in several non-computing disciplines for the most able students -- the future leaders in science, industry, education, philosophy and other fields. At present the system is not ready for that, so I am merely trying to plant some seeds.
One aim of the talk will be to draw attention to capabilities of non-mathematicians that have mathematical content, insofar as they involve understanding and use of mathematical features of the environment, including partly built bird's nests, shoe-laces and children's construction kits. Even pre-verbal toddlers seem to acquire and use such mathematical knowledge, along with nest-building birds, elephants, and other intelligent animals. These competences make use of mathematical knowledge about arrangements of matter unreflectively -- not explicitly thought about or communicated. That comes later, but only (so far) in humans. What changes when a child becomes able to understand why it is that if you look past the left edge of a doorway you'll see more of the left side room if you move right, less if you move left, and can later tell someone else which way to move to see more of a hidden object. This uses mathematical reasoning about straight lines, that for most people is unconscious.
Progress in replicating such human and animal capabilities in AI and robotics, and explaining them in psychology and neuroscience has been very slow. That's partly, I think, because we don't yet understand what needs to be explained. More examples will be presented in the talk.
In 2011 I was asked to contribute to a book celebrating Turing's centenary the following year. One of the things I read was his amazing paper on The Chemical basis of Morphogenesis published in 1952, two years before he died. That led me to ask what he would have done if he had lived another 30 or 40 years. My tentative answer was the Meta-Morphogenesis (M-M) project: an attempt to understand all the important transitions in information-processing capabilities and mechanisms between the very earliest living or pre-biotic entities several billion years ago and the huge variety now on the planet, including many intelligent species, such as crows, elephants, dolphins, squirrels, orangutans and humans.
It's possible that attempting to identify intermediate forms of information processing in organisms that were distant ancestors of organisms now on earth may provide new clues about the types of information processing in current systems, that can't be found by more direct study. I'll try to explain how.
All the ideas coming out of this research are gradually being assembled on
a rather messy web site dedicated to the M-M project.
I hope to interest some broad minded teachers and others interested in
education, and discuss possible ways in which ideas like these could be
informally made available for teachers and learners who want to explore beyond
the standard disciplinary boundaries -- perhaps eventually producing some of the
leading thinkers, including scientists and engineers, of the future, who will
not get the required foundation from current syllabus structures, but who have
the abilities and motivation to go further in non-school time. I see no reason
why this should not begin in primary schools alongside the forms of programming
already being taught to some very young children. See
Studying evolved biological construction kits and information-processing mechanisms they build may help us understand evolution of minds and extend our ability to design intelligent, human-like robots.
More on the Turing-Inspired Meta-Morphogenesis project
Since Turing discussed the possibility of intelligent machines in 1950 there have been many outstanding achievements in artificial intelligence, robotics and computational cognitive science -- including logical and algebraic theorem provers and proof checkers. Yet we don't know how to give a machine visual perception and spatial reasoning abilities found in pre-verbal human toddlers, and many other animals, e.g. weaver birds, shown here: https://www.youtube.com/watch?v=6svAIgEnFvw
Although computers can perform logical and algebraic reasoning and can even discover new mathematical theorems and prove them, current robots and mathematical reasoning systems cannot match the forms of spatial intelligence found in many non-human animals (e.g. crows, squirrels, elephants, and many more).
These include the cognitive abilities apparently required for the discovery of the truths and proofs in Euclidean geometry leading up to Euclid's Elements, published about 2.5 thousand years ago.
Can we understand what's missing in current AI but present in human toddlers and
other animals and how it evolved? Perhaps those forms of mathematical spatial
reasoning require new forms of computation?
First steps towards a general theory of evolved construction kits:
Background: The Turing-Inspired Meta-Morphogenesis project
Personal web site: http://www.cs.bham.ac.uk/~axs
Installed: 22 Sep 2015
Last updated: 25 Oct 2015; 26 Dec 2015