A general-purpose list of conclusions for
talks on the Meta-Morphogenesis project!
Version for Sussex University COGS talk on 27 Jan 2015:
Construction-kits for evolvable types of minds
Many of the conclusions merely identify questions we need to answer
that we have not yet answered -- some of which have not even been
We must abandon ideological wars (e.g. about enactive, embodied,
dynamical, neural, situated, behaviour-based... etc. approaches) in AI/Cognitive
Science/Robotics and cooperate more, especially in the hard task of identifying
long term goals -- including scientific, engineering and philosophical goals.
Instead of slanging one another about what forms answers must take, we should
try to cooperate on finding out what we don't already know.
We need deeper and more comprehensive knowledge of what the problems are:
what needs to be explained, what we cannot yet model, replicate, explain, or
even list, e.g. huge, unnoticed gaps in our knowledge about:
-- functions of vision (e.g. in online and offline intelligence).
-- roles of proto-mathematical capabilities in children and other animals
e.g. geometry, topology, ...
-- which abilities develop when in children, and why, and how.
-- what different forms of life can and cannot do and why?
steps, big and small
-- -- E.g. what do we and what don't we understand about the competences of
What varieties of affective states and processes can occur, and how they are
related to information-processing architectures and control mechanisms
pleasures, pains, hopes, fears, ambitions, regret, dismay, triumph,
grief, jealousy, anger, compassion, loyalty to family, to friends, to school, to
neighbours, to country, to ancestors, ... , being torn by conflict,
delighted anticipation, apprehensive anticipation, anticipation full of
-- what it is to find something funny,
-- why and how the ability to find things funny evolved,
-- what sorts of mechanisms make artistic enjoyment, artistic creativity,
How forms of cognition can depend on physical environment and culture
-- how and why human language evolved, and what the pre-cursors were,
-- -- Don't assume the only languages are external: rich internal languages
-- -- Probably came first.
We need to try more systematically to find out which animal abilities current
AI/robotic/neural systems and models lack or fail to explain, and why: e.g. are
the failures due to limitations of our concepts and tools, or limitations of
empirical data, or limitations of our educational system...?
We may be missing some important intermediate biological phenomena
- perhaps a systematic survey of evolutionary and developmental trajectories
involving information and
information-processing would help. (HARD - missing data, invisible data, etc.)
We may have to go back to the earliest forms of life on the planet, and
precursors to find clues.
Could a universe composed of interacting Newtonian "point masses" support
construction kits needed for life? If not, why not?
What relevant powers do chemistry and quantum mechanisms add?
We need a theory of construction kits for (... many things...) and need to
understand better the kind of fundamental construction kit (FCK) provided by the
physical universe - especially chemistry (Turing, 1952).
It's clear that evolution produced a wide variety of construction kits in a wide
variety of branching lineages. These must all ultimately be built from the FCK,
so they are Derived Construction Kits (DCKs). We need to try to find out out or
work out what sorts of DCKs have been or must have been produced by evolution,
and how some of then changed the mechanisms of evolution. (Hence
There are both Concrete and Abstract construction kits -- the latter not
produced by evolution though representations of them and instances of many
of them were.
Because products of Concrete construction kits occupy space time they can
interact with one another and even form new composite construction kits that are
often more tan the sum of their parts (qualitatively new phenomena).
The existence of Construction Kits seems to be a far more realistic foundation
for our semantic concepts of modality (what is and is not possible, or
necessarily true) than
possible world semantics. (Not discussed today.)
Construction kits have mathematical properties, whose consequences determine
what sorts of things can be built from them and what sorts cannot.
We need to understand What sorts of construction kit support construction
processes that create new sorts of construction kits that can build new machines
and new kits with very different properties ... e.g. through natural selection
and its products, in an on-going process: Meta-Morphogenesis.
What patterns of individual development of information-processing can occur in
different species, including humans, elephants, cetaceans, crows, squirrels,
octopuses, ... e.g. discovery of "toddler theorems" by young humans
If, as some claim, sensory-motor morphology completely specifies and explains
cognitive functions (information-processing capabilities?) how does that square
with what humans born blind, deaf, blind and deaf, missing arms, missing legs,
conjoined as twins, afflicted by thalidomide, with cerebral palsy, or seriously
deprived by illness or injury while very young are able to achieve, e.g. Helen
Keller (who lost sight and hearing very young), Alison Lapper, the artist and
writer without arms, Esref Armagan, the painter born blind, Abigail and Brittany
Hensel, the conjoined twins, and many others.
A vast amount of behaviour involving information processing in humans,
especially in early months and years happens out of sight, in mouths, initially
controlling sucking, later used in many forms of play, exploration, manipulation
of foods of various kinds, requiring different sorts of manipulation in the
mouth, and increasingly after the first year or two in production of pre-cursors
to speech, then speech itself, taking different forms in different cultures.
(Similar points are relevant to many other animals with mouths, beaks, etc.)
Theories of cognition emphasising embodiment and sensory motor morphology seem
to imply that we'll never give robots human intelligence without giving them
human-like mouths, lips, cheeks, tongues, mouth musculature,
swallowing and breathing mechanisms, vocal mechanisms, etc.
"Information" here is not Shannon's information, but Jane Austen's information,
involving semantic contents of different kinds, especially control information.
How many varieties of information, uses of information, modes of representation
of information, modes of transition of information, occur in nature; and which
of them are required in intelligent machines?
Mathematical structures underlie everything:
Why is it more useful to regard natural selection as a blind theorem prover
than a blind watch-maker?
What have we learnt about information processing systems in the last century or
so that Darwin could not have known? Which of those discoveries or inventions is
relevant to answering hard biological questions, e.g. questions about the nature
of qualia, or the varieties of types of consciousness?
Is it possible to discover how very complex information-processing systems work
by systematically investigating and modelling the physical machinery used?
Or will be first need a deeper understanding of what they are used for in living
Are there related things we have not yet discovered that are relevant to
answering the questions we have formulated about life, mind, intelligence,
evolution, what human-made machines can do, etc.?
What else would Alan Turing have done, if he had not died so young? Why was he
so interested in chemistry?
OUR EDUCATIONAL SYSTEMS NEED RADICAL REVISION....
School of Computer Science
The University of Birmingham