ChapterPDF Available

A Dialogue Concerning Two World Systems: Info-Computational vs. Mechanistic

Authors:

Abstract

The dialogue develops arguments for and against adopting a new world system, info-computationalist naturalism, that is poised to replace the traditional mechanistic world system. We try to figure out what the info-computational paradigm would mean, in particular its pancomputationalism. We make some steps towards developing the notion of computing that is necessary here, especially in relation to traditional notions. We investigate whether pancomputationalism can possibly provide the basic causal structure to the world, whether the overall research programme appears productive and whether it can revigorate computationalism in the philosophy of mind.
xxxi
CONTENTS
Preface v
1. Cybersemiotics and the Question of Knowledge 1
Søren Brier
2. Information Dynamics in a Categorical Setting 35
Mark Burgin
3. Mathematics as a Biological Process 79
G. J. Chaitin
4. Information, Causation and Computation 89
John Collier
5. From Descartes to Turing: The computational Content 107
of Supervenience
S. Barry Cooper
6. A Dialogue Concerning Two World Systems, 149
Info-Computational Vs. Mechanistic
Gordana Dodig-Crnkovic & Vincent C. Müller
7. Does Computing Embrace Self-Organization? 185
Wolfgang Hofkirchner
8. Analysis of Information and Computation in Physics Explains 203
Cognitive Paradigms: From Full Cognition to Laplace
D
eterminism to Statistical Determinism to Modern Approach
Vladik Kreinovich, Roberto Araiza, and Juan Ferret
http://www.sophia.de - http://orcid.org/0000-0002-4144-4957
Dodig-Crnkovic, Gordana and ller, Vincent C. (2011), ʻA dialogue c oncerning t wo world
systems: Info-computational vs. mechanisticʼ, in Gordana Dodig-Crnkovic and Mark Burgin
(eds.), Information and computation: Essays on scientific and philosophical understanding
of foundations of information and computation (Boston: World Scientific), 149-84.
xxxii Contents
9. Bodies—Both Informed and Transformed Embodied 225
Computation and Information Processing
Bruce J. MacLennan
10. Computation on Information, Meaning and Representations: 255
An Evolutionary Approach
Christophe Menant
11. Interior Grounding, Reflection, and Self-Consciousness 287
Marvin Minsky
12. A Molecular Dynamic Network: Minimal Properties and 307
Evolutionary Implications
Walter Riofrio
13. Super-recursive Features of Evolutionary Processes and the 331
Models for Computational Evolution
Darko Roglic
14. Towards a Modeling View of Computing 381
Oron Shagrir
15. What’'s Information, for an Organism or Intelligent Machine? 393
How can a Machine or Organism Mean?
Aaron Sloman
16. Inconsistent Knowledge as a Natural Phenomenon: 439
The Ranking of Reasonable Inferences as a Computational
Approach to Naturally Inconsistent (Legal) Theories
Kees (C.N.J.) de Vey Mestdagh & Jaap Henk (J.H.)
Hoepman
17. On the Algorithmic Nature of the World 477
Hector Zenil and Jean-Paul Delahaye
Index 497
149
CHAPTER 6
A DIALOGUE CONCERNING TWO WORLD SYSTEMS:
INFO-COMPUTATIONAL VS. MECHANISTIC
Gordana Dodig-Crnkovic & Vincent C. Müller
Mälardalen University & Anatolia College/ACT
www.idt.mdh.se/personal/gdc & www.typos.de
The dialogue develops arguments for and against adopting a new world system info-
computationalist naturalism that is poised to replace the traditional mechanistic view.
We try to figure out what the info-computational paradigm would mean, in particular its
pancomputationalism. We discuss steps towards developing the new generalized notion
of computing that is necessary here which includes both symbolic and sub-symbolic in-
formation processing, and its relation to traditional notions. We investigate whether pan-
computationalism can possibly provide the basic causal structure to the world, whether
the overall research programme of info-computationalist naturalism appears productive,
especially when it comes to new rigorous approaches to the living world and whether it
can revigorate computationalism in the philosophy of mind.
It is important to point out that info-computational naturalism does not invalidate
mechanistic approach within the domains where it has shown its soundness, but it extends
the domain of research to the classes of phenomena which were not possible to adequate-
ly address in terms of mechanistic approaches. The relationship is like the one between
relativistic and classical physics or between non-Euclidean and Euclidian geometry, i.e.
the new paradigm includes and generalizes the old one, making the older mechanistic
paradigm a special case of the more general framework.
1. Introduction
1.1. Galileo, Ptolemy, Mechanicism and Systèmes du Monde
In his 1632 Dialogue Concerning the Two Chief World Systems (Dialogo
sopra i due massimi sistemi del mondo), Galileo contrasts two different
world views: the traditional Ptolemaic geocentric system where every-
150 G. Dodig-Crnkovic & V. C. Müller
thing in the Universe circles around the Earth, vs. the emerging Coperni-
can system, where the Earth and other planets orbit the Sun. Even though
the question whether the Earth was the center of the Universe or not was
important in itself, the real scientific revolution was going on in the
background; the transition from qualitative Aristotelian physics to the
Galileo-Newtonian quantitative mechanistic physics necessary to support
the new worldview. The new model with equations of motion for celes-
tial bodies following Newton’s laws set the standard for all of physics to
come. This mechanistic paradigm accommodates even for Quantum Me-
chanics and Theory of Relativity, two theories that are both part of the
classical mechanical “Clockwork Universe” and question its basic intui-
tions about a perfectly intuitive, regular and predictable “World-
Machine(Machina Mundi).
The mechanistic world view is based on the following principles:
(M1) The ontologically fundamental entities of the physical reality
are physical structures (space-time & matter) and change of
physical structures (motion).
(M2) All the properties of any complex physical system can be de-
rived from the properties of its components.
(M3) Change of physical structures is governed by laws.
(M4) The observer is outside of the system observed.
Mechanistic models assume that the system is closed, isolated from the
environment, and laws of conservation (energy, mass, momentum, etc.)
thus hold. Environment, if modelled at all, is treated only as a perturba-
tion for the steady state of the system.
1.2. Info-Computational Naturalism !"#$%&
What we begin to see at present is a fundamentally new paradigm of not
only sciences but even a more general paradigm of the universe, compa-
rable in its radically novel approach with its historical predecessors the
Mytho-poetical Universe and the Mechanistic Universe. We identify this
new paradigm as Info-Computational Universe.
A Dialogue Concerning Two World Systems 151
According to info-computational naturalism (ICON) the physical uni-
verse is fundamentally an informational structure whose dynamics are
identified as computational processes [Dodig-Crnkovic, 2006; 2008].
This computation process is Natural computing; see Bruce MacLennan’s
article in this volume. Mark Burgin’s article, “Information Dynamics in a
Categorical Setting”, presents a common framework for information and
computation, building a mathematical stratum of the general theory of
information based on category theory.
A remarkable feature of info-computationalism is its ability to unify
living and nonliving physical world and to provide clues to mental ca-
pacities in humans and animals. Of all grand unifications or systèmes du
monde as Greg Chaitin says in his Epistemology as Information Theory:
From Leibniz to ! [Chaitin, 2007a] this is the first one holding promise
to be able to explain and simulate in addition to non-living universe
even the structure and behavior of living organisms, including the hu-
man mind.
Complexity is important for many physical phenomena, and is an es-
sential characteristic of life, the domain in which the info-computational
approach best shows its full explanatory power. Living organisms are
complex, goal-oriented autonomous information-processing systems with
ability of self-organization, self-reproduction (based on genetic infor-
mation) and adaptation. They (we) evolved through pre-biotic and bio-
logical evolution from inanimate matter. The understanding of basic in-
fo-computational features of living beings has consequences for many
fields, especially information sciences, cognitive science, psychology,
neuroscience, theory of computing, artificial intelligence and robotics but
also biology, sociology, economics and other fields where informational
complexity is essential.
Being on the edge of a brand new era we have a good enough reason
to follow Galileo’s example and try to contrast two world systems the
existing and well established mechanistic framework with the new
emerging unfinished but promising info-computational one.
The info-computational world view is based on the following
principles:
(IC1) The ontologically fundamental entities of the physical reality
are information (structure) and computation (change).
152 G. Dodig-Crnkovic & V. C. Müller
(IC2) Properties of a complex physical system cannot be derived
solely from the properties of its components. Emergent prop-
erties must be taken into account.
(IC3) Change of informational structures is governed by laws.
(IC4) The observer is a part of the system observed.
Info-computational models comprise open systems in communication
with the environment. The environment is a constitutive element for an
open complex info-computational system. A network of interconnected
parts is a typical configuration, where understanding is sought on the me-
ta-level with respect to constituent parts. Info-computational models in-
clude mechanistic ones as a special case when interaction of the system
with the environment may be neglected.
In what follows we will try to contrast the mechanistic and info-
computational positions. This dialogue between Müller (VCM) and
Dodig-Crnkovic (GDC) is the result of a series of discussions on the top-
ic we had on different occasions over the last couple of years.
2. Pancomputationalist Claims
VCM
When both authors were invited to contribute a debate to a conference in
2008
1
, we jointly submitted an abstract that included the following char-
acterization:
Info-computationalism is the view that the physical universe can be
best understood as computational processes operating on informa-
tional structure. Classical matter/energy in this model is replaced by
information, while the dynamics are identified as computational pro-
cesses. In this view the universe is one gigantic computer that con-
tinuously computes its next states, following physical laws.
Info-computationalism here appears as a conjunction of two theses: one
1
“Philosophy’s Relevance in Information Science” at the University of Paderborn, 3.-
4.10.08, organized by Ruth Hagengruber. http://groups.uni-
paderborn.de/hagengruber/pris08
A Dialogue Concerning Two World Systems 153
about processes (computation) and one about structure (information). In
this dialogue, I want to focus on the first one, that all processes are com-
putational, which I shall call “pancomputationalism”.
2
In any case, if this
pancomputationalism fails, the stronger thesis of info-computationalism
fails with it.
Our first task is to gain a better understanding of the thesis involved. I
shall propose some alternative readings that will be further elucidated
and evaluated in this dialogue. I will start with the strongest thesis and
move to weaker ones; so if one agrees with a particular thesis on this list,
one will probably agree with all that follow further down.
One reading of the basic thesis in the above quote is:
P1: The universe is a computer
This seems to be the strongest version, so I shall call it “strong pancom-
putationalism”. Perhaps I should mention that this view normally in-
cludes the thesis that the universe is physical, something that we shall
just assume in the following. A bit more restricted is the reduction to
changes or ‘processes’ in the universe:
P2: All processes are computational processes
This thesis is the main target of our discussion here, so I shall just call it
“pancomputationalism”. Very often, the point of the theory in question,
however, is not what processes are (whatever that may mean, exactly),
but how they can be described, so this suggests another formulation:
P3: All processes can be described as computational processes
Weak pancomputationalism. This formulation, however is ambiguous as
it stands, and I shall thus avoid it. Its ambiguity stems from the fact that –
as I will explain presently - there are very different reasons to claim that
processes can be described in this way; reasons concerning the theory of
computing and reasons concerning the nature of the universe. Whether
one wants to take a realist or an anti-realist view of computing will be
decisive here. Reasons concerning the nature of the universe (realist rea-
sons) might be formulated as follows:
2
The term “pancomputationalism” was probably coined by [Floridi, 2004, 566].
154 G. Dodig-Crnkovic & V. C. Müller
P4: All processes can be described as computational processes be-
cause we discovered that they are computational
This theory can justly bear the title of a pancomputationalism because it
claims to have discovered a fundamental fact about the world. It is, in
fact, just the pancomputationalism of P2 plus the claim that this feature
has been discovered, so I think P4 can be disregarded; it is just ‘pancom-
putationalism’. Another possible explanation of P3 that relies on a par-
ticular (anti-realist) notion of computation is:
P5: All processes can be described as computational processes be-
cause there is nothing more to being a computational process
than being described as such
Anti-realist weak pancomputationalism. This theory does not claim that
the universe has a particular structure; in fact it is often used to argue
against a theory that a part of the world (the mind) is computational in
any substantial sense. Instead, it stems from the anti-realist view that it is
our description as such that makes a process into a computational one.
Versions of this tradition are represented, for example, by David
Chalmers [Chalmers, 1993; 1994; 1996] John Searle [Searle, 1980; 1990;
1992, 207f] and Oron Shagrir [Shagrir, 2006]. While P5 thus has a lot of
support in the literature, I would suggest that it is too weak for a substan-
tial pancomputationalism in the sense envisaged by Dodig-Crnkovic.
What we need in our development of P3 is a realist formulation, like
this one:
P6: All processes can be described as computational processes be-
cause this happens to be a useful way of describing them in sci-
entific theory
Realist weak pancomputationalism. This thesis takes a realist view of
computation and then claims that all the actual processes of the universe
are such that they can be described as computational in a scientific theory
of the universe (while some processes in other possible worlds might not
have this feature). It is ‘weak’ because it only talks about ways of de-
scription, not about realist ontology – unlike P2.
If we wanted to regard the issue just in terms of how things may be
described without claiming that this description is or should be part of a
A Dialogue Concerning Two World Systems 155
scientific theory, then we would descend to what is really just a meta-
phorical remark, and thus the end of our sequence:
P7: All processes can be described as if they were computational pro-
cesses
Metaphorical pancomputationalism. This thesis is probably true and it
looks like it might be extremely useful in areas as different as economics
and microbiology. It does not say anything about the world, however, but
that things may be described as if they were computational – be that sci-
entific or not.
A further question is whether the claim of pancomputationalism, in
one of the versions above, is meant to be a claim about ‘everything’ or
‘everything deep down’; i.e. is computation a fundamental property of
the universe (‘deep down’) and other properties relate to it systematical-
ly, e.g. they reduce to it? Or is literally everything in the universe compu-
tational? Against the latter, stronger, claim there are many areas of the
universe (social, aesthetic, mental) that do not seem computational at all.
3
Are they, or are they ‘deep down’?
GDC
P1: The universe is a computer
As all subsequent theses P2-P6 are just the weaker versions of P1 let me
focus on the strongest claim, P1 in the first place. The pancomputational-
ist original claim is exactly P1. Of that which is universe we say that it is
a computer. What pancomputationalists
4
actually aim for is not only giv-
ing the universe just another name (“computer”) but they suggest that
universe computes.
It is pretty obvious that universe computer is not of the same sort as my
PC, as it contains stars, rocks, oceans, living organisms and all the rest (in-
cluding PC’s). So we talk about a more general idea of computing and a
computer. This question of computing in real world, the nature of compu-
3
Here are some examples of social, aesthetic or mental facts that do not seem computa-
tional: “The struggle over copyright in the digital age is really a power struggle,Her
hair curled beautifully”, “My breath was taken away by the sight”.
4
http://en.wikipedia.org/wiki/Pancomputationalism
156 G. Dodig-Crnkovic & V. C. Müller
ting as implemented in physics, is addressed in this volume by MacLennan
and Shagrir. Cooper and Sloman discuss questions of the relationship be-
tween computing, information and mind. Needless to say, in the pancom-
putational universe, mind is a result of natural computation that our brains
supported by bodily sensors and actuators constantly perform.
In sum, I would say that all of proposed pancomputational claims P1-
P6 are correct. The universe is a computer, but an unconventional (natu-
ral) one; its dynamics (temporal evolution) best can be understood as
computation and it can be described as such. There is nothing more to
being computational process but being used as computation.
5
For the last claim P7, however, I propose it to be modified. Metaphor
is a figure of speech while pancomputationalism concerns the physical
world. I would say that pancomputationalism is a metaphor in the same
sense as Niels Bohr’s liquid drop model is a metaphor of atomic nucleus.
In sciences we are used to talk about models and for a good reason. We
use model as a tool to interact with the world. So if you agree to call it
model instead of a metaphor, I would agree even with P7. I propose the
following:
P7’: All physical processes can be modeled as computational processes
which I recognize as pancomputationalist claim. We will have several
occasions in this dialogue to return to the question of what is computing
and to discuss unconventional (natural) computing that is going on in
computational universe. In principle, there seems to be no ontological
hindrance to our including the system or process we try to compute
among the models we use. We then get what I take to be the fundamental
idea of pancomputationalism: The function governing a process is calcu-
lated by the process itself
6
. The following remark by Richard Feynman
lucidly explains the idea:
It always bothers me that according to the laws as we understand
them today, it takes a computing machine an infinite number of logi-
5
As Kaj Børge Hansen puts it, “A computation is a thought experiment. We often in-
crease our power and ability to do thought experiments by aiding our limited memory and
imagination by symbolic representations, real and virtual models, and computers.(Per-
sonal communication.)
6
For this formulation I thank KB Hansen.
A Dialogue Concerning Two World Systems 157
cal operations to figure out what goes on in no matter how tiny a re-
gion of space and no matter how tiny a region of time… I have often
made the hypothesis that ultimately physics will not require a math-
ematical statement, that in the end the machinery will be revealed
and the laws will turn out to be simple.
[Feynman, 1965, 57]
7
3. Are There any Arguments for Pancomputationalism?
VCM
As far as I can tell the arguments in favor of pancomputationalism have
been largely intuitive, indicating that this view is useful and offers an ele-
gant all-encompassing view of the world in terms that are well understood.
This feature it shares, however, with any number of all-encompassing ide-
ologies (like pantheism or vulgar liberalism). These intuitive arguments
apply to all of P1-P7 without offering any particular support for the
stronger versions. What is missing is a positive argument that out of the
many overall theories this one is true and in one version of it.
GDC
Pancomputationalism apart from being universal has nothing to do with
pantheism which is not based on scientific methods. I dont think that
vulgar liberalism either should be mixed in here as it is not a theory
about the universe in its entirety, so I would suggest comparison with
atomism as a good example of universal theory. In natural sciences the
most general theories are the best ones. Universal laws are the best laws.
Being universal is nothing bad, just on the contrary! It is expected of a
theory of nature to be universal.
The central question is how epistemologically productive this para-
digm is, as it really is a research programme (on this I share the view
presented by Wolfgang Hofkirchner in this volume) whose role is to mo-
bilize researchers to work in the same direction, within the same global
7
Used as the motto for the 2008 Midwest NKS Conference,
http://www.cs.indiana.edu/~dgerman/2008midwestNKSconference/index.html
158 G. Dodig-Crnkovic & V. C. Müller
framework. The majority of natural sciences, formal sciences, technical
sciences and engineering are already based on computational thinking,
computational tools and computational modeling [Wing, 2008].
Allow me to list some arguments for paradigm change, since it was
said that these are missing. Following are some of the promises of info-
computationalism:
The synthesis of the (presently alarmingly disconnected) knowledge
from different fields within the common info-computational framework
which will enrich our understanding of the world. Present day narrow
specialization into different isolated research fields has gradually led into
impoverishment of the common world view.
Integration of scientific understanding of the phenomena of life
(structures, processes) with the rest of natural world helping to achieve
“the unreasonable effectiveness of mathematics” such as in physics
(Wigner) even for complex phenomena like biology that today lack
mathematical effectiveness (Gelfand).
8
Understanding of the semantics of information as a part of data-
information-knowledge-wisdom sequence, in which more and more
complex relational structures are created by computational processing of
information. An evolutionary naturalist view of semantics of information
in living organisms is based on interaction (information exchange) of an
organism with its environment.
A unified picture of fundamental dual-aspect information/compu-
tation phenomenon applicable in natural sciences, information science,
cognitive science, philosophy, sociology, economy and number of others.
Relating phenomena of information and computation understood in
interactive paradigm makes it possible for investigations in logical plu-
ralism of information produced as a result of interactive computation.
9
Of
special interest are open systems in communication with the environment
and related logical pluralism including paraconsistent logic.
Advancement of our computing methods beyond the Turing-Church
paradigm, computation in the next step of development becoming able to
8
See Chaitin, “Mathematics, Biology and Metabiology” (Foils, July 2009)
http://www.umcs.maine.edu/~chaitin/jack.html
9
This logical pluralism is closely related to phenomena of consistency and truth; see also
de Vey Mestdagh & Hoepman in this volume.
A Dialogue Concerning Two World Systems 159
handle complex phenomena such as living organisms and processes of
life, knowledge, social dynamics, communication and control of large
interacting networks (as addressed in organic computing and other kinds
of unconventional computing), etc.
Of all manifestations of life, mind seems to be information-
theoretically and philosophically the most interesting one. Info-
computationalism (pancomputationalism + paninformationalism) has a
potential to support (by means of models and simulations) our effort in
learning about mind.
10
4. What is Computing? (I) The Fragile Unity of Pancomputational-
ism
VCM
There are several theories about what constitutes acomputation, the
classical one being Turing’s, which identifies computation with a digital
algorithmic process (or “effective procedure”). If, however, pancomputa-
tionalism requires a larger notion of computing that includes analog
computing and perhaps other forms, it would seem necessary to specify
what that notion is while making sure that Turing’s notion is included.
It is far from clear that there is a unifying notion that can cover all that
the pancomputationalist wants, and therefore there is a danger that the
advertised elegance of a single all-encompassing theory dissolves under
closer inspection into a sea of various related notions.
It should be noted for fairness, however that while it is clearly a de-
sideratum to specify the central notions of one’s theory, pancomputation-
alism can hardly be faulted for failing to achieve what is generally re-
garded as a highly demanding task, namely a general specification of
computation.
10
On the practical side, understanding and learning to simulate and control functions and
structures of living organisms will bring completely new medical treatments for all sorts
of diseases including mental ones which to this day are poorly understood. Understanding
of information-processing features of human brain will bring new insights into such fields
as education, media, entertainment, cognition etc.
160 G. Dodig-Crnkovic & V. C. Müller
One possible response to this challenge deserves a mention here, name-
ly the response that relies on the traditional use of mathematical or formal
tools in science. This could be condensed into the following thesis:
F: All physical processes can be described formally
I suspect that sympathies for this view stand behind much of the support
for pancomputationalism. However, F is not identical to strong pancom-
putationalism and not even easy to reconcile with it, for three reasons: a)
F talks about the possibility to describe things, i.e. it does not make any
claim to a realist reading [unlike P1 and P2], b) its use would identify
computing with formal description and c) it explicitly talks about pro-
cesses and thus forbids any swift moves from P2 to P1 in case that P1
is the desired thesis. What is the logical relation between pancomputa-
tionalism and thesis F?
GDC
Actually the lack of understanding for what computing is may be a good
argument for starting this whole research programme. At the moment,
the closest to common acceptance is the view of computing as infor-
mation processing, found in Neuroscience, Cognitive science and num-
ber of mathematical accounts of computing; see [Burgin, 2005] for expo-
sition. For a process to be a computation a model must exist such as al-
gorithm, network topology, physical process or in general any mecha-
nism which ensures predictability of its behavior.
The three-dimensional characterization of computing can be made by
classification into orthogonal types: digital/analog, interactive/batch and
sequential/parallel computation. Nowadays digital computers are used to
simulate all sorts of natural processes, including those that in physics are
described as continuous. In this case, it is important to distinguish be-
tween the mechanism of computation and the simulation model. It is in-
teresting to see how computing is addressed in the present volume, espe-
cially Barry Coopers account of definability and Bruce MacLennans em-
bodied computing. We will mention symbolic vs. sub-symbolic compu-
ting as important in this context. So symbolic part is what is easily rec-
A Dialogue Concerning Two World Systems 161
ognized as thesis F
11
. In a sense we may say that F applies even to sub-
symbolic computing on a meta-level. What is described formally is not
the computational process step by step, but the mechanism that will pro-
duce that process.
Information processing
12
is the most general characterization of com-
puting and common understanding of computing in several fields. In the
info-computational approach information is a structure and computation is
a process of change of that structure (Dodig Crnkovic, 2006). I have used
the expression dynamics of information” for computation. No matter if
your data form any symbols; computation is a process of change of the
data/structure. On a fundamental quantum-mechanical level, the universe
performs computation on its own (Lloyd, 2006). Symbols appear on a
much higher level of organization, and always in relation with living or-
ganisms. Symbols represent something for a living organism, have a func-
tion as carriers of meaning.
13
(See Christophe Menant in this volume).
5. What is Computing? (ii) discrete Vs. Continuum or Digital vs. An-
alog
VCM
I used to believe that what a computational process is was nicely defined
by Church and Turing in the 1930ies, namely that these are the “effec-
tive” procedures”, just the algorithms that can be computed by some Tu-
ring machine. This does, at least, provide something like a ‘core’ notion.
11
As already pointed out, we have two different types of computation: physical substrate
sub-symbolic and symbolic which also is based on physical computation. In Mark
Burgin’s words: “However, we can see that large physical computations can give a futile
symbolic result, while extended and sophisticated symbolic computations sometimes
result in meager physical changes. Pancomputationalism actually cannot exist without
accepting the concept of physical computation. Discovery of quantum and molecular
computing shows that the same symbolic computation may result from different kinds of
physical computation.” (From e-mail exchange with Mark Burgin, 26.07.2009 )
12
A popular account in Wikipedia http://en.wikipedia.org/wiki/Computation
13
Douglas Hofstadter has already addressed the question of a symbol formed by other
symbols in his del, Escher, Bach. [Taddeo and Floridi, 2005] present a critical review
of the symbol grounding problem with a suggestion that symbols must be anchored in
sub-symbolic level.
162 G. Dodig-Crnkovic & V. C. Müller
It can be expanded in several ways, but any notion of computing should
include this ‘core’. One expansion is ‘hypercomputing’; the idea that
there can be algorithmic procedures that compute what no Turing ma-
chine could compute (typically by carrying out infinitely many compu-
ting steps). Now, I do not think that a machine can carry out infinitely
many steps in finite time and come up with an output [Müller, 2008a]
but I would grant, of course, that hypercomputing is computing, if only it
were physically possible in this world, or indeed in any possible world.
There might be a set of computing procedures that is larger than the
one defined by Church-Turing – and there is certainly a mathematical set
of computable functions larger than that computable by Turing machine
(e.g. that computable by Turing’s idea of his machine plus “oracle”).
This is still quite far from saying that the universe is a computer (P1
above), however. So pancomputationalism probably has to add ‘analog
computers’ as well, machines who’s processing is not digital steps and
who’s output thus requires measurement to a degree of accuracy (if there
is any ‘output’ at all).
14
My understanding of ‘computer’, as suggested by
[Turing, 1936], is that such machines characteristically go beyond mere
calculators (like those already invented by Leibniz and Pascal) in that
they are universal; they can, in principle, compute any algorithm, be-
cause they are programmable in this sense, Zuse’s Z3 was the first
computer (1941). If this feature of universality is a criterion for being a
computer, then analog machines do not qualify because they can only be
programmed in a very limited sense. This is a question of conventional
terminology, however, so if we want to call such analog devices ‘com-
puters’, we can. What is not clear to me is how this relates to the notion
of ‘symbol’, traditionally a central one for computing. Presumably, ana-
log computers do not use symbols, or digital states that are interpreted as
symbols.
So, if we grant that computing includes digital hypercomputing and
analog computing, this raises two questions: First, how can you guaran-
tee that the notion of ‘computing’ you are using here is in any sense uni-
14
This extension to ‘analog computers’ is not necessary if pancomputationalism adds the
thesis of ‘digital physics’, that the world is fundamentally digital; something that I find
rather implausible - though there are arguments about this issue [Müller, 2008b] [Floridi,
2009].
A Dialogue Concerning Two World Systems 163
fied, i.e. one notion? (The question raised above.) And second, how does
this broadening of the notion help to support the notion that everything is
computing, beyond providing an ad hoc answer to the obvious challenge
that not everything is a digital computer?
GDC
In the above, you identify hypercomputing as one way to expand the no-
tion of computing, by carrying on infinitely many computational steps in
a finite time, so we can focus our discussion for a while on the analysis
of that statement. We can translate this question in its turn into Chaitin’s
question about the existence of real numbers; see Chaitin How real are
real numbers? [in Chaitin, 2007b, 276]. For Chaitin real numbers are
chimeras of our own minds, they just simply do not exist! He is in good
company. Georg Leopold Kronecker’s view was that, while everything
else was made by man, the natural numbers were given by God. The log-
icists believed that the natural numbers were sufficient for deriving all of
mathematics. In the above, you seem to suggest this view.
Even though pragmatic minded people would say that discrete set can
always be made dense enough to mimic continuum for all practical pur-
poses, I think on purely principal grounds that one cannot dispense with
only one part in a dyadic pair and that continuum and discrete are mutu-
ally defining.
15
Here I would just like to point out that the discrete continuum prob-
lem lies in the underpinning of calculus and Bishop George Berkeley in
his book The analyst: or a discourse addressed to an infidel mathemati-
cian, argued that, although calculus led to correct results, its foundations
were logically problematic. Of derivatives (which Newton called flux-
ions) Berkley wrote:
15
I suppose that this dyadic function comes from our cognitive apparatus which makes
the difference in perception of discrete and continuous. It is indirectly given by the world,
in a sense that we as a species being alive in the world have developed those dyad-
ic/binary systems for discrete (number) and continuous (magnitude) phenomena as the
most effective way to relate to that physical world.
Much of our cognitive capacities seem to have developed based on vision, which has on
its elementary level the difference between: signal/no signal.
164 G. Dodig-Crnkovic & V. C. Müller
And what are these fluxions? The velocities of evanescent incre-
ments. And what are these same evanescent increments? They are
neither finite quantities, nor quantities infinitely small, nor yet noth-
ing. May we not call them ghosts of departed quantities?
16
Philosophical problems closely attached to the idea of infinity in mathe-
matics are classical ones.
From physics on the other hand, there are persistent voices, such as
[Lesne, 2007] witnessing for the necessity of continuum in physical
modeling of the world. Here is the summary:
This paper presents a sample of the deep and multiple interplay be-
tween discrete and continuous behaviours and the corresponding
modellings in physics. The aim of this overview is to show that dis-
crete and continuous features coexist in any natural phenomenon,
depending on the scales of observation. Accordingly, different mod-
els, either discrete or continuous in time, space, phase space or con-
jugate space can be considered. [Lesne, 2007]
[Floridi, 2009] proposes the Alexandrian solution to the above Gordian
knot by cutting apart information from computation, and expressing eve-
rything in terms of information. This would be analog to describing a
verb with a noun; it is possible but some information gets lost. It is nev-
ertheless true that informational structure of the universe is richer than
what Turing Machines as a typical mechanical/mechanistic model can
produce.
17
digital ontology (the ultimate nature of reality is digital, and the
universe is a computational system equivalent to a Turing Machine)
should be carefully distinguished from informational ontology (the
ultimate nature of reality is structural), in order to abandon the for-
16
http://www.maths.tcd.ie/pub/HistMath/People/Berkeley/Analyst/Analyst.html Berkeley
talks about the relationship between the model and the world, not about the inner struc-
ture of the model itself. Worth noticing is KB Hansen’s remark that “problems observed
by Berkeley have been solved by Bolzano, Cauchy, Riemann, Weierstrass, and Robinson.
Modern mathematical analysis rests on solid foundations.”
17
For we talk about computational processes that not only calculate functions but are able
to interact with the world, posses context-awareness, ability of self-organization, self-
optimization and similar.
A Dialogue Concerning Two World Systems 165
mer and retain only the latter as a promising line of research. Digital
vs. analogue is a Boolean dichotomy typical of our computational
paradigm, but digital and analogue are only “modes of presentation”
of Being (to paraphrase Kant), that is, ways in which reality is expe-
rienced or conceptualised by an epistemic agent at a given level of
abstraction. A preferable alternative is provided by an informational
approach to structural realism, according to which knowledge of the
world is knowledge of its structures. The most reasonable ontological
commitment turns out to be in favour of an interpretation of reality as
the totality of structures dynamically interacting with each other.
[Floridi, 2009, 151]
What info-computationalist naturalism wants is to understand that dy-
namical interaction of informational structures as a computational pro-
cess. It includes digital and analogue, continuous and discrete as phe-
nomena existing in physical world on different levels of description and
digital computing is a subset of a more general natural computing.
The question of continuum vs. discrete nature of the world is ages old
and it is not limited to the existing technology. Digital philosophy as well
as Turing machine has been epistemologically remarkably productive
(see Stephen Wolframs work, e.g. [Wolfram, 2002] along with Ed Fred-
kin and number of people who focused on the digital aspects of the
world). Digital is undoubtedly one of the levels we can use for the de-
scription, but from physics it seems to be necessary to be able to handle
continuum too (as we do in Quantum Mechanics). For a very good ac-
count, see [Lloyd, 2006].
6. What is computing? (III) Natural Computing as a Generalization
of the Traditional Notion of Computing
GDC
We have already discussed hypercomputing as the possibility of carrying
on infinitely many (computational) steps in a finite time as a question of
our understanding of the nature of the world (continuous, discrete) and
our idea of infinity. There is however yet another possibility to approach
the question of computing beyond the Turing model which goes under
166 G. Dodig-Crnkovic & V. C. Müller
different names and has different content: natural computing, uncon-
ventional computing, analog computing, organic computing, sub-
symbolic computing, etc.
In order to expound the present understanding of computing, and its
possible paths of development we study the development of the compu-
ting field in the past half a century, driven by the process of miniaturiza-
tion with dramatically increased performance, efficiency and ubiquity of
computing devices. However, this approach based on the understanding
of computation as symbol manipulation performed by a Turing Machine
is rapidly approaching its physical and conceptual limits.
Ever since Turing proposed his machine model identifying computa-
tion with the execution of an algorithm, there have been questions about
how widely the Turing Machine model is applicable. Church-Turing
Thesis establishes the equivalence between a Turing Machine and an al-
gorithm, interpreted as to imply that all of computation must be algo-
rithmic. Hector Zenil and Jean-Paul Delahaye in this volume investigate
the question of the evidence of the algorithmic computational nature of
the universe.
With the advent of computer networks, the model of a computer in
isolation, represented by a Turing Machine, has become insufficient; for
an overview see (Dodig Crnkovic 2006). Today’s software-intensive and
intelligent computer systems have become huge, consisting of massive
numbers of autonomous and parallel elements across multiple scales. At
the nano-scale they approach programmable matter; at the macro scale,
multitude of cores compute in clusters, grids or clouds, while at the plan-
etary scale, sensor networks connect environmental and satellite data.
The common for these modern computing systems is that they are en-
semble-like (as they form one whole in which the parts act in concert to
achieve a common goal like an organism is an ensemble of its cells) and
physical (as ensembles act in the physical world and interact with their
environment through sensors and actuators).
A promising new approach to the complex world of modern autono-
mous, intelligent, adaptive, networked computing has successively
emerged. Natural computing is a new paradigm of computing (MacLen-
nan, Rozenberg, Calude, Bäck, Bath, Müller-Schloer, de Castro, Paun)
which deals with computability in the physical world such as biological
A Dialogue Concerning Two World Systems 167
computing/organic computing, computing on continuous data, quantum
computing, swarm intelligence, the immune systems, and membrane
computing, which has brought a fundamentally new understanding of
computation.
Natural computing has different criteria for success of a computation.
The halting problem is not a central issue,
18
but instead the adequacy of
the computational response. Organic computing system e. g. adapts dy-
namically to the current conditions of its environment by self-
organization, self-configuration, self-optimization, self-healing, self-
protection and context-awareness. In many areas, we have to computa-
tionally model emergence not being algorithmic (Aaron Sloman, Barry
Cooper) which makes it interesting to investigate computational charac-
teristics of non-algorithmic natural computation (sub-symbolic, analog).
Interesting to observe is epistemic productiveness of natural computing
as it leads to a significantly bidirectional research (Rozen); as natural
sciences are rapidly absorbing ideas of information processing, field of
computing concurrently assimilates ideas from natural sciences.
VCM
P!NP. Or, to be a bit more explicit: I really suspect that Turing was right
about his set of digitally computable functions, no matter how long it
might take to compute them. All of the fashionable ‘beyond Turing’
computing (small, networked, natural, adaptive, etc. etc.) is either just
doing what a Turing machine does or it is not digital computing at all. If
it is not digital computing, then my question (notorious by now) is: Why
call it computing? In what sense of that word?
GDC
Let me remind that for a process to be a computation a model must exist
such as algorithm, network topology, physical process or in general any
mechanism which ensures predictability of its behavior. So we distin-
guish computation models and physical implementations of computation.
18
In the Turing model a computation must halt when execution of an algorithm has fin-
ished.
168 G. Dodig-Crnkovic & V. C. Müller
Talking about models of computation beyond Turing model, super-
recursive algorithms are an instructive example. They represent compu-
tation which can give a result after a finite number of steps, does not use
infinite objects, such as real numbers, and nevertheless is more powerful
than any Turing machine. Inductive Turing machines described in
[Burgin, 2005] have all these properties. Besides, their mode of computa-
tion is a kind of a natural computation, as demonstrated with respect to
evolutionary computations.
When it comes to physical implementations, natural computing pre-
sents the best example of the more general computational process than
that used in our present days computers. In what sense of the word is that
computing? In the sense of computation as a physical process, see Feyn-
mans remark about physical computing from 2.2 above.
Physical processes can be used for digital and analog computation. It
is true that historical attempts to build analog computers did not continue
because of the problem with noise. In a new generation of natural com-
puters we will use features organic computing possess in order to control
complexity. Organic systems are very good at discerning information
from noise.
This leads us to the next important characteristics of natural comput-
ers. They will not be searching for a perfect (context free) solution, but
for a good enough (context dependent) one. This will also imply that not
all computational mechanisms will be equivalent, but we will have clas-
ses of equivalence of computational devices in the same sense as we
have different types of computational processes going on at different lev-
els of organization.
Why call it computation?
Simply because it is a generalization of present day computation from
discrete symbol manipulation to any sort of (discrete, continuous) ma-
nipulation of symbols or physical objects (discrete, continuous) , which
follow physical or logical laws.
A Dialogue Concerning Two World Systems 169
7. Computation and Causation
VCM
As we already saw, pancomputationalism seems to rely on an under-
standing of computation that is rather unconventional. Conventional un-
derstanding investigates a physical process and then says about that pro-
cess that it computes (a function). Which processes in the world are the
ones that are computing is a thorny question that hinges on the criteria;
on whether one regards computing as a matter of discovery or a matter of
perception; etc. So, even if one says things like that the universe is in-
formation processing [Wiener, 1961, 132 etc.], this is still meant in the
sense that there is some ‘stuff’ in the world that is undergoing processes
which are information processing – not that the universe is a computer.
No matter which processes are regarded as computational ones (i.e.
how narrow or wide the notion of computing is taken to be), a usual as-
sumption is that the same computation can be carried out by different
physical processes one example of this is the remark that the same
software can ‘run’ on different hardware, even on hardware that is struc-
turally quite different. What this underlines is that the output of a compu-
tation, e.g. “0”, is a different entity from the outcome of the physical
process, e.g. a switch being in “position A” (which stands for0). The
computation is not the cause of the position of the switch, but the physi-
cal process is. The same computational process on different hardware
would have resulted in “0”, but quite possibly not in a switch in “position
A”. In fact, a computation cannot cause anything, it is just a syntactic
event, or perhaps the syntactic description of an event (out of the massive
literature on this issue, see [Piccinini, 2008]). If this is right, then compu-
tation cannot be used as an overall empirical theory, as I indicated in my
paper for the Paderborn meeting [Müller, 2009].
GDC
Pancomputationalism indeed often (but not necessarily) relies on an un-
derstanding of computation that is unconventional. Exactly that un-
conventional computation is one of the most exciting innovations that
pancomputationalism supports. Not all adherents of unconventional
170 G. Dodig-Crnkovic & V. C. Müller
computation (computation beyond Turing limit) are pancomputational-
ists. Unconventional computation will be found all over this volume
(Cooper, MacLennan, Shagrir, …) based on different arguments and
approaches. There are conferences and journals on unconventional
computing, organic computing, and natural computing. I see it as a
good sign of coherence coming from different, often completely unre-
lated fields. A good overview on non-classical computation may be
found in [Stepney, 2005].
When it comes to the issue of causal inefficacy of computation,
that is really not a problem for control systems or robotics where you
indeed see computation causing an artifact to interact with the envi-
ronment. Info-computationalism has no problem with computation not
being causally connected with the physical world. As the world com-
putes its own next state, it means that computation has causal power.
Not only spontaneous computation of the universe in form of natural
computation is causally effective, even human-designed (constructed)
devices controlled by computational processes show that computation
is what directly connects to the world.
In the same way as there is no information without (physical) rep-
resentation [Karnani, et al., 2009], there is no computation without
information (which must have physical representation). So any output
of a computation performed by a computer (say “0” from your exam-
ple) can in principle be used as an input for a control system that
launches a rocket or starts any sort of machine controlled by a com-
puter. Today we have numerous examples of embedded computers and
even embodied ones (see MacLennan in this volume) where computa-
tional processes control or in other ways impact physical world.
8. Is Pancomputationalism Vacuous or Epistemically Productive?
VCM
Presumably, pancomputationalism is an empirical theory, so it should
indicate which empirical evidence it will count as supportive and make
predictions about empirical findings that if they do not materialize
would count as evidence against the theory, perhaps even as falsification.
A Dialogue Concerning Two World Systems 171
The absence of such links to empirical findings would increase suspicion
that the theory is actually devoid of content. Has the theory produced any
new testable hypotheses?
GDC
“Theories are nets: only he who casts will catch.” Novalis
Novalis is quoted by Karl Popper in the introduction to The Logic of Sci-
entific Discovery. In the third chapter, Popper elaborates:
Theories are nets cast to catch what we call “the world’’: to rational-
ize, to explain, and to master it. We endeavor to make the mesh ever
finer and finer.
19
[Popper, 1959]
Not only so that no theory, however general, can capture all aspects of
the universe simultaneously (and thus we have a multitude of different
general scientific theories valid in their specific domains, on specific lev-
el of abstraction), but even more importantly: pancomputationalism is
not a single monolithic theory but a research programme. We talk about
système du monde. This volume provides examples and shows how
things happen to develop more in the spirit of Let a Thousand Flowers
Bloom. The process of consolidation, purification, formalization is the
next step. We are still in a discovery phase.
In order to understand the development of a research programme let
us return to the analogy of pancomputationalism with atomism
20
, the be-
lief that all physical objects consist of atoms and void (Leucippus and
Democritus). We can equally ask how atomism could have possibly been
falsified. I don’t think it could. Because atomism (and in a similar way
pancomputationalism) is not to be understood as a single hypothesis but
as a research programme. In the strict sense atomism has already been
19
This example can be paraphrased to say that not only that our nets are getting finer, but
maybe altogether different methods of fishing and not only the finer-grained mesh nets
can be devised.
20
I have used this analogy with atomism for many years, only recently to see in a Docu-
mentary/Drama “Victim of the brain (on Hofstadter/Dennett’s “The Mind’s I” [Hof-
stadter and Dennett, 1981] featuring Daniel Dennett and Marvin Minsky), that Douglas
Hofstadter uses exactly that argument in a very elegant way, see
http://www.mathrix.org/liquid/#/archives/victim-of-the-brain
172 G. Dodig-Crnkovic & V. C. Müller
falsified because atoms are not indivisible (but made of nucleus made of
nucleons made of quarks made of …) and void (vacuum) is not empty
(but full of virtual particles that pop up into being and disappear again).
The process of the development of atomist research project has both
changed the original idea about atoms and void and what we identify as
their counterparts in the physical world.
The fundamental question thus does not seem to be about the “truth”
of a singular statement such as: Is it true that there are only atoms and
void? But as we use atoms and void as a net to catch the structure of the
physical reality, those ideas are instrumental to our understanding, and in
the interaction with the world, both our concepts and what we are able to
reach to in the world change concurrently. What is fundamental is con-
struction of meaning, or epistemological productiveness of a paradigm,
or how much we can learn from the research programme.
You [Müller, 2008c, 38] rightly use Kant to suggest the way to ad-
dress the question of how to define Computing and Philosophy, namely
by answering the following questions:
What can we hope for (from Computing and Philosophy)?
What should we do (with Computing and Philosophy)?
What can we know (about Computing and Philosophy)?
Equivalent questions can be asked about info-computationalist pro-
gramme. A theory (or a paradigm) is an epistemic tool, that very tool No-
valis and Popper use to catch (or extract as Cooper in this volume says)
what for us is of interest in the world. Compared to mytho-poetic and
mechanistic frameworks the emerging info-computational paradigm is the
most general one and the richest in expressive repertoire developed
through our interaction with the world.
21
When the dominating interaction
with the world was mechanistic, the most general paradigm was mechanis-
tic. The world in itself/for itself is simply a reservoir/resource [Floridi,
2008] of possible interactions for a human. We know as much of the world
as we explore and “digest” (as a species or as a community of praxis).
21
Our nets are global computer networks of connected computational, information pro-
cessing devices. The classic era of mechanism was focused on matter and energy. Our
own info-computational paradigm focuses on information and computation.
A Dialogue Concerning Two World Systems 173
Since we wish to devise an intelligible conceptual environment for
ourselves, we do so not by trying to picture or photocopy whatever is
in the room (mimetic epistemology), but by interacting with it as a
resource for our semantic tasks, interrogating it through experience,
tests and experiments. Reality in itself is not a source but a resource
for knowledge. Structural objects (clusters of data as relational enti-
ties) work epistemologically like constraining affordances: they al-
low or invite certain constructs (they are affordances for the infor-
mation system that elaborates them) and resist or impede some others
(they are constraints for the same system), depending on the interac-
tion with, and the nature of, the information system that processes
them. They are exploitable by a theory, at a given Level of Abstrac-
tion, as input of adequate queries to produce information (the model)
as output. [Floridi, 2008]
All we have are constructs made for a purpose, and so is even the case
with pancomputationalism: let’s say world is a computer, what sort of
computing is it then? It is not a vacuous tautology but a proposal for ex-
ploration, a research programme. It presupposes a dynamical reflexive
relationship between our understanding of the physical world and our
theoretical understanding of computation or what a computer might be.
The worst thing which can happen is that some of the world is impos-
sible to use for learning of any new principles or building any new smart
machines. That may happen if physical processes are irreducible and if
we want to know the result of computation we have to use the replica of
a system, which is not very useful. But that is not the major issue. First of
all even in case or randomness [Chaitin, 2007a] when no information
compression is possible the physical world shows remarkable stability
and we can expect it to repeat the same behavior under same circum-
stances, so we don’t have to actually repeat all computations, but re-
member recurring behaviors.
174 G. Dodig-Crnkovic & V. C. Müller
9. Pancomputationalism and the Mind
VCM
The view that the human mind is a computer has been a cornerstone of the
cognitive sciences from their beginning, supported by the philosophical
position of ‘machine functionalism’. It has come under increasing pressure
in recent years, and under the impression of the main arguments many
have been lead to abandon it.
22
Is there any substantial sense in which info-
computationalism relieves this pressure and blows some life into the no-
tion that the mind is a computer beyond saying that everything is?
GDC
Yes, I would say so. I would like to claim that info-computationalism
(info-computationalist naturalism) has something essentially new to offer
and that is natural computation/organic computation, which applies to
our brains too.
The classical critique of old computationalism based on abstract, syn-
tactic notion of computation represented by Turing Machine model does
not apply to the dynamic embodied physical view of computing that new
natural computational models support. [Scheutz, 2002] has the right di-
agnosis:
Instead of abandoning computationalism altogether, however, some
researchers are reconsidering it, recognizing that real-world comput-
ers, like minds, must deal with issues of embodiment, interaction,
physical implementation, and semantics.
Scheutz similarly to Shagrir in this volume concludes that according to
all we know brain computes, but the computation performed is not of in
the first place a Turing Machine type. Several papers in this issue con-
tribute to the elucidation of earlier misunderstandings; from Marvin Min-
22
A quick indication of some main points: The problem of meaning in a computational
system (Chinese room and symbol grounding), the critique of encodingism (Bickhard),
the stress on non-symbolic or sub-symbolic cognition, the integration of cognition with
emotion and volition, the move away from a centralized notion of cognition and towards
‘embodiment’, etc.
A Dialogue Concerning Two World Systems 175
sky’s analysis of the hard problem of consciousness to Aaron Slomans
approach to mind as virtual machine. This book presents an effort to
build the grounds for understanding of computing in its most general
form and to use it in addressing real world phenomena, including life and
mind, those topics mechanistic models are not suitable to deal with.
Part of our previous discussion about discrete vs. continuum is rele-
vant for the argument about computational nature of mind. If computa-
tion is allowed to be continuous, then the mind can be computational:
Brains and computers are both dynamical systems that manipulate
symbols, but they differ fundamentally in their architectures and op-
erations. Human brains do mathematics; computers do not. Comput-
ers manipulate symbols that humans put into them without grounding
them in what they represent. Human brains intentionally direct the
body to make symbols, and they use the symbols to represent internal
states. The symbols are outside the brain. Inside the brains, the con-
struction is effected by spatiotemporal patterns of neural activity that
are operators, not symbols. The operations include formation of se-
quences of neural activity patterns that we observe by their electrical
signs. The process is by neurodynamics, not by logical rule-driven
symbol manipulation. The aim of simulating human natural compu-
ting should be to simulate the operators. In its simplest form natural
computing serves for communication of meaning. Neural operators
implement non-symbolic communication of internal states by all
mammals, including humans, through intentional actions. (…) I pro-
pose that symbol-making operators evolved from neural mechanisms
of intentional action by modification of non-symbolic operators.
[Freeman, 2009]
The above shows nicely the relationship between symbolic and non-
symbolic computing. All that happens inside our heads is non-symbolic
computing. (Freeman claims it is non-symbolic while Shagrir with neu-
roscientists claims it is computing.) Our brains use non-symbolic compu-
ting internally to manipulate relevant external symbols!
If we learn to interpret life as a network of information processing
structures and if we learn how our brains (and bodies) perform all that in-
176 G. Dodig-Crnkovic & V. C. Müller
formation processing then we will be able to make new computers which
will smoothly connect to our information processing cognitive apparatus.
To summarize, we can choose digital description but then we will be
able to see the world in that “digital light”. If we choose continuum, we
will capture different phenomena. Pancomputationalism does not exclude
any of the (discrete, continuum, digital or analog) computing (infor-
mation processing). Info-computational naturalism, being a general uni-
fying approach connects natural information processes with correspond-
ing informational structures.
10. Concluding remarks
VCM
What I think this exchange shows is that a lot of work remains to be done
before we can say that pancomputationalism is a well-understood and
evaluated position (not to mention info-computationalism, which in-
volves further claims). I am therefore not of the view that the position is
refuted, but that we need to clarify its claims, its fruitfulness and its pos-
sible problems it is to this program that we hoped that our discussion
would contribute (and to my mind it did).
It is in this intention that I suggested a list of possible theses at the
outset. It might be useful to list them here again (where P4-P6 are possi-
ble readings of P3):
P1: The universe is a computer (strong pancomputationalism)
P2: All processes are computational processes (pancomputational-
ism)
P3: All processes can be described as computational processes (weak
pancomputationalism)
P4: All processes can be described as computational processes be-
cause we discovered that they are computational (= P2)
P5: All processes can be described as computational processes be-
cause there is nothing more to being a computational process
A Dialogue Concerning Two World Systems 177
than being described as such (anti-realist weak pancomputation-
alism)
P6: All processes can be described as computational processes be-
cause this happens to be a useful way of describing them in sci-
entific theory (realist weak pancomputationalism)
P7: All processes can be described as if they were computational
processes (metaphorical pancomputationalism)
In response, we were told in no uncertain terms that, out of the various
theses, strong pancomputationalism (P1) is the intended reading. Fine, in
this strong realist reading the answer to my first question becomes even
more urgent: What would be the case if the theory were false, i.e. what
would a counterexample look like? My suggestion (in good Popperian
tradition, since his name was invoked), is that there is a danger for very
general ideologies that seem to explain everything, but really are empty
and explain nothing. If classical atomism is still a useful or true theory
(unlike pantheism), there must be a sense in which it can be interpreted
as such.
In the defense of the theory, it was stressed that pancomputationalism
should be viewed as a ‘research program’, a ‘paradigm that it is epis-
temologically productive’, and that in any case theories should not be
viewed as statements but as nets. All of this looks like P6, rather than P1.
As long as it is granted that these two are different theses, this strategy
might be accused of claiming the stronger thesis but defending the weak-
er one. I see several instances of this problem here.
One example is the response to the problem of the apparent causal in-
efficacy of computing, countered by examples of computers that do
things. As an indication to show how this is not the same, let us look at
Sloman’s suggestion that there are ‘virtual machines’ active in the human
brain that causally generate aspects of conscious experience [Sloman,
2009]. This looks like he is saying that computing has causal power – but
not quite, since he says that virtual machines are mathematical objects
that do nothing, only running virtual machines have causal powers (since
they run on physical hardware, I would add). He makes the crucial dis-
tinction. My problem did not consist in the strange suggestion that com-
178 G. Dodig-Crnkovic & V. C. Müller
puters do not have causal powers (mine certainly has) but in the question
whether the computational processes qua computational processes have
these powers (since their output is only the “0” in my example, not the
position of a physical switch). In pancomputationalism, the stronger the-
sis about computing processes per se is claimed, and the weaker about
running/actual/realized computing processes is defended.
A further example of this strategy is the defense of the stronger P1 via
the weaker claim that the universe is processing information. It may well
be true that information processing is an elementary feature of the uni-
verse, but information processing is information processing; computing
is computing. Perhaps computing is one species of information pro-
cessing among others (in some sense of ‘informationit is), but why ex-
pand the one notion into the other? If we really want to say that all in-
formation processing is computational, is that a definitional remark or is
this a discovery about information processing? If it is definitional, I
might adopt my understanding of the thesis proposed here but I would
then note that we now identify one unclear notion (computing) with an-
other even less clear one (information processing); which does not look
like a good strategy. In any case, all problems that beset the pancomputa-
tionalist approach also beset that of info-computationalism, plus the new
ones associated with ‘information’. If the remark is expressing a discov-
ery, I would like to see the evidence for the claim that there is non-
computational information processing does not exist. (In other words, I
would come back to my first remark and wonder what a non-
computational process would be, on the pancomputationalist account.)
Last but not least, the claim that the universe is computational looked
quite strong when that term was understood as Turing computation, but
then computation was dissolved into a much wider notion, the borders of
which I cannot quite discern (I keep coming back to this issue). My wor-
ry about apparently non-computational processes in the world could not
be countered because “everything is computing” is a priori, and we do
not even want to take it as a reductionist claim (“everything is compu-
ting, deep down”).
One example of my confusion is the interpretation of the remark by
Feynman quoted above to support pancomputationalism. In what sense
of the word is that computing? In the sense of computation as a physical
A Dialogue Concerning Two World Systems 179
process, see Feynman’s remark about physical computing from 2.2.”
This sounds circular and Feynman’s remark does not help. I think it can
well be read as opposing the idea that the universe computes (disregard-
ing any context). He could be taken to say: Since it would take a compu-
ting machine infinite time, the “machinery” that is revealed in the end is
not computational.
It truly is not clear to me how much can be explained with a wide no-
tion of computing that somehow incorporates digital and analog, formal
and physical, Turing and dynamic systems, etc. etc. I have nothing
against these proposals, indeed my feeling is that some processes can
usefully be described as computational (though even P6 in its generality
is false) and many more as if they were. I also suspect that the metaphor-
ical power of info-computationalism is strong enough to support an en-
tire research program which will generate many interesting insights.
Having said that, we have seen what happened to fruitful and successful
research programs like classical AI or computationalism in the philoso-
phy of mind that rested on weak foundations they eventually hit the
wall. I suspect that this will be the fate of info-computationalism also.
GDC
First let us go back to Feynman. It is not a coincidence that this quote
was used as the motto for the 2008 Midwest NKS Conference which
gathered most prominent pancomputationalists. They interpret Feynman
as saying that nature computes much more effective than any of our pre-
sent machines. Moreover Feynman seems to imply that our going via
mathematical models of physical phenomena might be the reason for that
ineffectiveness.
Now the question of what is reasonable to understand as computation.
For the nascent field of natural computation, we can apply the well
known truth that our knowledge is in a constant state of evolution. Ray
Kurzweil would even warn: Singularity is near, singularity where
knowledge production exceeds our ability to learn [Kurzweil, 2005].
Moreover, by integrating/assimilating new pieces of knowledge, the
whole existing knowledge structure changes. Atomism has changed sub-
stantially from the Democritus’ original view. And yet it has not been
180 G. Dodig-Crnkovic & V. C. Müller
refuted but only modified. Why? Because it was epistemically produc-
tive! Simply put, atomism has helped us to think, helped us to build new
knowledge and to interact in different way with the physical world.
That is exactly what we expect from info-computationalism to pro-
vide us with a good framework which will help our understanding of the
world, including life and ourselves and our acting in that world. It is ba-
sically about learning and making sense of the world. Having history of
several major paradigm shifts behind us, we have no reason to believe
that info-computational framework is the absolutely perfect answer to all
questions we may ask about the life, universe and everything but it seems
to be the best research framework we have right now.
If pancomputationalism claims that the entire universe computes, a
discovery of a process in the world which is impossible to understand as
computational would falsify the pancomputational claim. Something
changes, but we have no way to identify that process as computation. A
stereotypical claim would be: writing a poem. That cannot possibly be a
computational process! On which level of organization? I want to ask.
On a level of neuroscience all that happens in the world while someone
writes a poem is just a sequence of computational processes. Poets might
find that level uninteresting, as well as they might find uninteresting the
fact that the beautiful lady they sing of is made of atoms and void. But
there are cases where we really want to know about how things work on
a very basic level.
As a research program info-computationalism will either show to be
productive or else it will die out. The only criterion for survival is how
good it will be compared with other approaches. That development of a
research programme is a slow but observable process. Following the
number of articles, journals, conferences etc. dealing with unconvention-
al computing, organic computing, or natural computing we can assess
how active the field is. Subsequently we will also be able to follow its
results.
Here is again a summary of what makes info-computationalist natu-
ralism a promising research programme:
- Unlike mechanicism, info-computationalist naturalism has the ability
to tackle as well fundamental physical structures as life phenomena
A Dialogue Concerning Two World Systems 181
within the same conceptual framework. The observer is an integral
part of the info-computational universe.
- Integration of scientific understanding of the structures and processes
of life with the rest of natural world will help to achievethe unrea-
sonable effectiveness of mathematics” (or computing in general)
even for complex phenomena of biology that today lack mathemati-
cal effectiveness (Gelfand) – in sharp contrast to physics (Wigner).
- Info-computationalism (which presupposes pancomputationalism
and paninformationalism) presents a unifying framework for com-
mon knowledge production in many up to know unrelated research
fields. Present day narrow specialization into various isolated re-
search fields has led to the alarming impoverishment of the common
world view.
- Our existing computing devices are a subset of a set of possible
physical computing machines, and Turing Machine model is a subset
of envisaged more general natural computational models. Advance-
ment of our computing methods beyond the Turing-Church paradigm
will result in computing capable of handling complex phenomena
such as living organisms and processes of life, social dynamics,
communication and control of large interacting networks as ad-
dressed in organic computing and other kinds of unconventional
computing.
- Understanding of the semantics of information as a part of the data-
information-knowledge-wisdom sequence, in which more and more
complex relational structures are created by computational pro-
cessing of information. An evolutionary naturalist view of semantics
of information in living organisms is given based on interac-
tion/information exchange of an organism with its environment.
- Discrete and analogue are both needed in physics and so in physical
computing which can help us to deeper understanding of their rela-
tionship.
- Relating phenomena of information and computation understood in
interactive paradigm will enable investigations into logical pluralism
of information produced as a result of interactive computation. Of
special interest are open systems in communication with the envi-
ronment and related logical pluralism including paraconsistent logic.
182 G. Dodig-Crnkovic & V. C. Müller
- Of all manifestations of life, mind seems to be information-
theoretically and philosophically the most interesting one. Info-
computationalist naturalism (pancomputationalism + paninforma-
tionalism) has a potential to support, by means of models and simula-
tions, our effort in learning about mind and developing artifactual
(artificial) intelligence in the direction of organic computing.
The spirit of the research programme is excellently summarized in the
following:
“In these times brimming with excitement, our task is nothing less
than to discover a new, broader, notion of computation, and to under-
stand the world around us in terms of information processing.” [Ro-
zenberg and Kari, 2008]
Acknowledgements
The authors want to thank Ruth Hagengruber, Kaj Børge Hansen, Luci-
ano Floridi and Mark Burgin for useful comments on earlier versions of
this paper.
References
Burgin, M. (2005). Super-recursive algorithms. Berlin: Springer.
Chaitin, G. (2007a). ‘Epistemology as information theory: From Leibniz to "’. In Dodig-
Crnkovic, G. & Stuart S., (Eds.), Computation, information, cognition: The nexus and
the liminal. Newcastle: Cambridge Scholars Publishing pp. 27–51.
Chaitin, G. (2007b). Thinking about Gödel and Turing: Essays on Complexity, 1970-
2007. Boston: World Scientific.
Chalmers, D.J. (1993). ‘A computational foundation for the study of cognition’, in.
Chalmers, D.J. (1994). ‘On implementing a computation’. Minds and Machines, 4, pp.
391–402.
Chalmers, D.J. (1996). ‘Does a rock implement every finite-state automaton?’ Synthese,
108, pp. 309–333.
Dodig-Crnkovic, G. (2006). Investigations into information semantics and ethics of com-
puting. Västerås: Mälardalen University Press.
Dodig-Crnkovic, G. (2008), ‘Semantics of Information as Interactive Computation’, in
Moeller, M., Neuser W. & Roth-Berghofer T. (Eds.), Fifth international workshop on
philosophy and informatics, Kaiserslautern 2008. Berlin: Springer.
Feynman, R.P. (1965). The character of physical law. Boston: MIT Press 2001.
A Dialogue Concerning Two World Systems 183
Floridi, L. (2004). ‘Open problems in the philosophy of information. Metaphilosophy,
35, pp. 554–582.
Floridi, L. (2008). ‘The method of levels of abstraction’. Minds and Machines, 18, pp.
303–329.
Floridi, L. (2009). ‘Against digital ontology’. Synthese, 168, pp. 151–178.
Freeman, W. (2009). ‘The Neurobiological Infrastructure of Natural Computing: Inten-
tionality’’. New Mathematics and Natural Computation, 5, pp. 19–29.
Hofstadter, D.R. & Dennett D.C. (Eds.). (1981). The mind's I: Fantasies and reflections
on self and soul. New York: Basic Books.
Karnani, M., Pääkkönen K. & Annila A. (2009). ‘The physical character of information’.
Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sci-
ence, 465, pp. 2155–2175.
Kurzweil, R. (2005). The singularity is near: When humans transcend biology. London:
Viking.
Lesne, A. (2007). ‘The discrete versus continuous controversy in physics’. Mathematical
Structures in Computer Science, 17, pp. 185–223.
Lloyd, S. (2006). Programming the Universe: A Quantum Computer Scientist Takes on
the Cosmos. New York: Alfred A Knopf.
Müller, V.C. (2008a). ‘On the possibility of hypercomputing supertasks’. Minds and
Machines, resubmission under review.
Müller, V.C. (2008b). ‘Representation in digital systems’. In Briggle, A., Waelbers K. &
Brey P., (Eds.), Current issues in computing and philosophy. Amsterdam: IOS Press
pp. 116–121.
Müller, V.C. (2008c). ‘What a course on philosophy of computing is not’. APA Newslet-
ter on Philosophy and Computers, 8, pp. 36–38.
Müller, V.C. (2009). ‘Pancomputationalism: Theory or metaphor?’ In Hagengruber, R.,
(Ed.), The relevance of philosophy for information science. Berlin: Springer p. forth-
coming.
Piccinini, G. (2008). ‘Computation without representation’. Philosophical Studies, 134,
pp. 205–241.
Popper, K. (1959). The logic of scientific discovery. New York: Basic Books.
Rozenberg, G. & Kari L. (2008). ‘The many facets of natural computing’. Communica-
tions of the ACM, 51.
Scheutz, M. (2002). ‘Computationalism: the next generation’. In Scheutz, M., (Ed.),
Computationalism: new directions. Cambridge: Cambridge University Press.
Searle, J.R. (1980). ‘Minds, brains and programs’. Behavioral and Brain Sciences, 3, pp.
417–457.
Searle, J.R. (1990). ‘Is the brain a digital computer?’ Proceedings and Addresses of the
American Philosophical Association, pp. 21–37.
Searle, J.R. (1992). The rediscovery of mind. Cambridge, Mass.: MIT Press.
Shagrir, O. (2006). ‘Why we view the brain as a computer’. Synthese, 153, pp. 393–416.
Sloman, A. (2009). ‘What Cognitive Scientists Need to Know about Virtual Machines’.
Proceedings of CogSci 2009, The Annual Meeting of the Cognitive Science Society,
31, p. forthcoming.
184 G. Dodig-Crnkovic & V. C. Müller
Stepney, S.e.a. (2005). ‘Journeys in non-classical computation I: A grand challenge for
computing research’’. The International Journal of Parallel, Emergent and Distribut-
ed Systems, 20, pp. 5–19.
Taddeo, M. & Floridi L. (2005). ‘Solving the symbol grounding problem: A critical re-
view of fifteen years of research’. Journal of Experimental and Theoretical Artificial
Intelligence, 17, pp. 419–445.
Turing, A. (1936). ‘On computable numbers, with an application to the Entschei-
dungsproblem’. Proceedings of the London Mathematical Society, pp. 230–256.
Wiener, N. (1961). Cybernetics: or control and communication in the animal and the
machine. Cambridge, Mass.: MIT Press.
Wing, J. (2008). ‘Five deep questions in computing’. CACM, 51, pp. 58–60.
Wolfram, S. (2002). ‘A new kind of science’, in, Champaign, Il.: Wolfram Media.
... Computational naturalism (pancomputationalism, naturalist computationalism, computing nature) [46][47][48], see even [3,4], is the view that the entirety of nature is a huge network of computational processes, which, according to physical laws, computes (dynamically develops) its own next state from the current one. Among prominent representatives of this approach are Zuse, Fredkin, Wolfram, Chaitin, and Lloyd, who proposed different varieties of computational naturalism. ...
... With its layered computational architecture, natural computation provides a basis for a unified understanding of phenomena of embodied cognition, intelligence, and learning (knowledge generation), including meta-learning (learning to learn) [47,80]. Natural computation can be modelled as a process of exchange of information in a network of informational agents [76], i.e., entities capable of acting on their own behalf, which is Hewitt's actor model applied to natural agents. ...
Article
Full-text available
The emerging contemporary natural philosophy provides a common ground for the integrative view of the natural, the artificial, and the human-social knowledge and practices. Learning process is central for acquiring, maintaining, and managing knowledge, both theoretical and practical. This paper explores the relationships between the present advances in understanding of learning in the sciences of the artificial (deep learning, robotics), natural sciences (neuroscience, cognitive science, biology), and philosophy (philosophy of computing, philosophy of mind, natural philosophy). The question is, what at this stage of the development the inspiration from nature, specifically its computational models such as info-computation through morphological computing, can contribute to machine learning and artificial intelligence, and how much on the other hand models and experiments in machine learning and robotics can motivate, justify, and inform research in computational cognitive science, neurosciences, and computing nature. We propose that one contribution can be understanding of the mechanisms of ‘learning to learn’, as a step towards deep learning with symbolic layer of computation/information processing in a framework linking connectionism with symbolism. As all natural systems possessing intelligence are cognitive systems, we describe the evolutionary arguments for the necessity of learning to learn for a system to reach humanlevel intelligence through evolution and development. The paper thus presents a contribution to the epistemology of the contemporary philosophy of nature.
... Computational naturalism (pancomputationalism, naturalist computationalism, computing nature) [46][47][48], see even [3,4], is the view that the entirety of nature is a huge network of computational processes, which, according to physical laws, computes (dynamically develops) its own next state from the current one. Among prominent representatives of this approach are Zuse, Fredkin, Wolfram, Chaitin, and Lloyd, who proposed different varieties of computational naturalism. ...
... With its layered computational architecture, natural computation provides a basis for a unified understanding of phenomena of embodied cognition, intelligence, and learning (knowledge generation), including meta-learning (learning to learn) [47,80]. Natural computation can be modelled as a process of exchange of information in a network of informational agents [76], i.e., entities capable of acting on their own behalf, which is Hewitt's actor model applied to natural agents. ...
Article
Full-text available
The emerging contemporary natural philosophy provides a common ground for the integrative view of the natural, the artificial, and the human-social knowledge and practices. Learning process is central for acquiring, maintaining, and managing knowledge, both theoretical and practical. This paper explores the relationships between the present advances in understanding of learning in the sciences of the artificial (deep learning, robotics), natural sciences (neuroscience, cognitive science, biology), and philosophy (philosophy of computing, philosophy of mind, natural philosophy). The question is, what at this stage of the development the inspiration from nature, specifically its computational models such as info-computation through morphological computing, can contribute to machine learning and artificial intelligence, and how much on the other hand models and experiments in machine learning and robotics can motivate, justify, and inform research in computational cognitive science, neurosciences, and computing nature. We propose that one contribution can be understanding of the mechanisms of ‘learning to learn’, as a step towards deep learning with symbolic layer of computation/information processing in a framework linking connectionism with symbolism. As all natural systems possessing intelligence are cognitive systems, we describe the evolutionary arguments for the necessity of learning to learn for a system to reach human-level intelligence through evolution and development. The paper thus presents a contribution to the epistemology of the contemporary philosophy of nature.
... 9 InformationC exists wherever physical reality exists, just like other physical phenomena. Information as a physical phenomenon that exists objectively has been recognized in the studies of, for example, Mynarski [9], Rovelli [11], Devlin [34], Heller [42,43], Collier [44], Stonier [45], Polkinghorne [46], Seife [47], Caroll [52], Dodig Crnkovic [50,60], and Dodig Crnkovic, Muller and Burgin [61]. Furthermore, Hidalgo [10], Stonier [45], and Seife [47], among others, posited that this information can be manipulated. ...
... Do we even care? (Suggested by the writings of Seife [47], Dodig Crnkovic [50,60], and Dodig Crnkovic and Mueller [61], for example.) Question 8: Can information be equated to some kind of structure, and what would this mean for the concept of structure? ...
Article
Full-text available
Abstract: This paper presents the concept of physical information, and it discusses what physical information is, and how it can be defined. The existence of physical information has been discussed in several studies which recognize that properties of information are characteristic of physical phenomena. That is, information has an objective existence, a lack of meaning, and can be quantified. In addition, these studies recognize how a phenomenon that is denoted as physical information can be expressed as an organization of natural or artificial entities. This paper argues that concepts of (abstract) information that are associated with meaning also depend (to a substantial degree) on physical information.