Deanthropomorphized Pancomputationalism and the Concept of Computing

Abstract

Pancomputationalism is quite a wide-ranging concept, but most of its variants, either implicitly or explicitly, rely on Turing's conceptualizations of a computer and computing, which are obvious anthropomorphisms. This paper questions the concept of pancomputationalism based on Turing computing and asks what concept of computation can be used to avoid the constrains of anthropomorphisations.

Full text

Deanthropomorphized Pancomputationalism
and the Concept of Computing
Paweł Polak1, Roman Krzanowski2
Abstract. Pancomputationalism is quite a wide-ranging concept, but most of its variants,
either implicitly or explicitly, rely on Turing’s conceptualizations of a computer and
computing, which are obvious anthropomorphisms. This paper questions the concept of
pancomputationalism based on Turing computing and asks what concept of computation
can be used to avoid the constrains of anthropomorphisations.
Keywords: pancomputationalims, Turing machines, computer, antropomorphisation,
computing.
1. Introduction
Computer science since its birth use a concept of computing which is based on the concept
of Turing computation [35]3. The Turing computation model while very successful has its
limitations [36]. These limitations are seen specifically in the modeling of nature (see e.g.
[33]). This results in an interesting paradox: on one hand we claim that nature is
computational in Turing sense [2, 30], on the other hand certain natural processes are beyond
the computing capacity/capability of the Turing model (see e.g. [9, 34]). If we assume that
nature in some way computes better than Turing model does, abandoning the Turing
computational model as the sole model of computation would open to us a new
possibilities, closed by sticking to the Turing model4. Thus, investigating natural computing
denoted often interpretations as pancomputationalism may change for the better the way we
see and practice computing.
1 Pontifical University of John Paul II in Kraków, Faculty of Philosophy, Chair of History and Philosophy of
Science, 31-002 Kraków, ul. Kanonicza 9; e-mail: <atpolakp@cyf-kr.edu.pl>.
2Pontifical University of John Paul II in Kraków, Faculty of Philosophy, 31-002 Kraków, ul. Kanonicza 9.
3 We use the terms “Turing computation”, “Turing computing”, “Turing concept of computing”, “Turing
machine”, and “Turing model” as synonyms, which we admit is not entirely correct. But in the context of the
paper we judge this use of different terms to be acceptable and not leading to the misunderstandings and
obfuscation of the discussed issues.
4This view was inspired by Peter Denning claims that considering natural computational processes is the best
way to develop the computing as a discipline [11]. In his view “computing is evolving constantly.”
Consequently he stated: “Computing is no longer a science of just the artificial. It is the study of information
processes, natural and artificial.”
F O U N D A T I O N S O F C O M P U T I N G A N D D E C I S I O N S C I E N C E S
Vol. 44 (2019) No. 1
ISSN 0867-6356
e-ISSN 2300-3405DOI: 10.2478/fcds-2019-0004

Ontic pancomputationalism (later referred to as pancomputationalism) is the notion that
nature, or the universe, is a computer, and what nature does is essentially computation [4,
16, 38–40]. There are many versions of pancomputationalism depending on how one
attribute computational powers to nature- everything computes, some processes compute,
or only special processes compute (see e.g. [30]). Only digital pancomputationalism
explicitly claims that the universe is the Turing machine (TM) (see e.g. [4]). But other
versions of pancomputationalism (unlimited, limited, ontic, strong, weak, limited, causal to
name just a few, see. e.g. [2, 30]) implicitly rely on Turing’s concepts of computing
assuming algorithmic procedure [31]5. Thus, we may safely state that pancomputationalism
is therefore de facto Turing pancomputationalism [3, 27, 30].
Is it bad, indifferent or beneficial to use Turing concept of computing as the model of
the universe? Of course, there is no definite answer to this question. However, we may
speculate that with known limitations of Turing computing disclosed by the concept of
hypercoputation [6, 7, 36], with the postulate that the universe is Turing-like computing
system we import limitations of the Turing computing model (and our understanding of)
onto the Universe, which obviously would not be an advisable move. To restate the
argument, if the Turing model of computation brings important limitations (more on this
topic see [32]), imposing it on the Universe gives us the limited picture of the universe,
preventing us from seeing it as it may be.
We therefore pose the question of how pancomputationalism can, or should, be
reinterpreted if we were to free ourselves from our dependency on Turing’s
conceptualization of computing?
It is also important to realize that the idea of Turing computing and Turing machines are
strongly based on analogies to human kind. Indeed, the Turing machine is modeled after an
idealized mechanized clerk (see e.g. [5]). Thus, in pancomputaionalism, rejecting Turing’s
model of computation, we would be in fact deanthropomorphizing it. We all know (the
claim hardly needs further justification) that any antropomorphizations in science lead to
dead ends, if not to outright embarrassments [1, 10]. Thus, “deanthropomorphizing
pancomputationalism”, if we agree that Turing computing is the factor centered around
human person (like the Ptolemaic model of the universe), should open for us new
perspectives on computing, computer and the Universe, as the deanthropomorphizing of the
Universe did to the concept of Cosmos [22] (see section 4).
Thus, we ask again would this deanthropomorphization of pancomputationalism open
up a new perspective onto pancomputaionalizm? And may be, would the
deanthropomorphized concept of pancomputationalism lead to a redefinition of our concept
of computers and computing?
2. What does pancomputationalism assume?
We should focus on the three main assumptions behind pancomputationalism [4, 14, 39]:6
5There are of course some exceptions, e.g. Witold Marciszewski's view on pancomputationalism. It is founded
on the Turing concept of computing, however it is enriched by assumption of the possibility of infinite
computation in reality on data with infinite description [25].
6 While looking at different formulations of ontic pancomputationalism, we could note the striking fact that
almost all of them assume digital ontology and some form of digital computation (i.e., a Turing machine or
universal cellular automaton).
46
P. Polak, R. Krzanowski

1. Digitalism, the claim that the physical world is a digital structure;
2. Physical pancomputationalism, the claim that physical objects perform
computational processes; and
3. Zuse’s thesis, the claim that the physical world is a universal computer.
The first assumption of digitalism requires a very heavy ontological commitment. In its
stronger version, it claims that “the physical world is isomorphic to a digital structure.” In
its weaker version, it claims that “the physical world is isomorphic to a mathematical
structure.” These assumptions and their consequences have been subjected to detailed
analysis [4].
The most important argument in support of digitalism comes from fundamental physics.
According to current views, physical theories describing the fundamental levels of the
physical world are discrete (e.g., quantum mechanics), but this argument is weak because it
is not clear which properties of the universe that we currently assume to be fundamental
actually constitute the fundamental level of physical reality. We cannot assume digital
ontology for theories such as unification based on non-commutative geometries [18,
19].The above mentioned theory shows that the digital structures of QM and continuous
structures of general relativity are only the limiting cases of more general mathematical
structures.
The second argument in support of digitalism derives from the successes of discrete
mathematics in modern science. However, these successes only show that natural structures
can be approximated by digital structures in digital computers, and they do not support any
deeper metaphysical claims about nature.7 There is no proof that these digital and natural
structures are identical. Thus, while digitalism is frequently assumed by some physical
theories, it cannot be defended as the ontology of reality. (For another critique of digital
ontology, see [14] and [29].)
The second assumption of physical pancomputationalism, meanwhile, is very open to
interpretation (e.g. [29]), so in its generic form, without additional qualification about what
computation means, it does not contribute much to the overall concept of
pancomputationalism. It can therefore be simply omitted without any consequences for the
definition.
It is also difficult to support Zuse’s thesis rejecting digitalism. Zuse’s concept of
computation is strongly tied to Turing’s concept, making his thesis strongly
anthropomorphic. A thorough analysis of Zuse’s model can be found in the work of
Beraldo-de-Araújo and Baravalle [4].
3. More on Zuse’s thesis
Zuse claims that the universe is a universal computer, but an uncritical acceptance of this
thesis could lead to “a danger for very general ideologies that seem to explain everything”
[13 pp. 25–26]. According to these authors, it is safer to select a weaker thesis called realist
7 An argument concerning the possibility of assuming digital ontology was put forward by Zuse [39] and used
more recently by Beraldo-de-Araújo and Baravalle [4].
47
Deanthropomorphized Pancomputationalism ...

weak pancomputationalism: “All processes can be described as computational processes
because this happens to be a useful way of describing them in scientific theory.” However,
the statement “this happens to be a useful way” will not withstand a sustained critique,
because trying to explain metaphysics through pragmatic arguments never succeeds. Thus,
even the weaker version of Zuse’s thesis seems unacceptable.
However, replacing the concept of a computer in Zuse’s thesis with some non-
anthropomorphous concept of computation is a challenge. Some possible solutions have
been investigated [14, 15].
Zuse’s thesis also brings up the issues of computational language and symbolic
representation. It is not clear how we could interpret these in the context of
pancomputationalism. For example, some arguments assert that natural computation8 could
be better characterized by sub-symbolic computation [12]. If we need some form of
instruction, we need a finite number of instructions to develop an effective description.
Therefore, in some cases of non-Turing computation (see e.g. [7]), this could be satisfied.
In these cases, we could (although maybe not even theoretically) obtain a functional
analogy of a digital computer but without the constraints of Turing computability.
As we said above, interpretations of Zuse’s thesis are strongly anthropomorphic because
they rely on Turing’s concept of computation and computers. We may therefore question
how this aspect of Zuse’s thesis affects the overall concept of pancomputationalism.
Note: In conceptualizing natural computations, we cannot assume only the mechanical
(it mean algorithmic) procedure envisioned by Turing, so the classical concept of an
algorithm will not suffice here. Therefore, what concept of an effective procedure could be
used instead? The answer to this clearly depends on the understanding of computation.
As we stated earlier, we cannot be sure if any computational model is ultimate in the
sense that it cannot be developed further, because its mathematical structure already fits
perfectly to that of nature.9 It therefore seems that from a theoretical point of view
(assuming pancomputationalism), the concept of computation needs to be endlessly
developed, together with the concept of a computer (cf. [11]). One may ask why, though?
Computation existed prior to the creation of computing devices, which are artifacts that
were developed to facilitate this process. The precedence of computation over the computer
seems obvious, even from just reading Turing’s papers and not searching out other
historical records.
4. Anthropomorphic assumptions
Anthropomorphism is in some sense an attribution of human characteristics to nature or
animals. In the current discussion of anthropomorphization in computer sciences and
informatics, anthropomorphism refers specifically to the use of human analogues in models
for computation and computers.
Anthropomorphism in science has been slowly weeded out (see for example the
classical article [1]) because it strongly constrains scientific theories despite the occasional
8 “Natural computation” means here the interpretation of process in nature as in some sense computational
processes.
9 The close view was presented by Floridi [14], who uses a concept of “modes of presentation” of being.
48
P. Polak, R. Krzanowski

positive contributions (e.g. [23]). Modern science has found the rejection of
anthropomorphism in science to be a sound approach and not open to argument.
If pancomputationalism were to follow the lead of modern science, it would have to
exclude all the anthropomorphic assumptions behind pancomputationalism.10 Thus, we
would not be able to a priori assume that the (mathematical) structures of nature11 are
isomorphic to the mathematical structures described by our mathematics.
Drawing lines in the sand could provide inspiration for thinking about continuity, and
historically, it was one of the roots leading to the mathematical concept of continuity.
However, the mathematical concept is not a simple idealization or abstraction of human
action. On the other hand, the mechanical procedure described by Turing (see e.g. [37 p.
436]) is precisely an idealization of human calculation. Not only was it inspired by this—it
is in fact the very essence of this action (see e.g. [7 p. 697]). In the first case,
anthropomorphism plays an anticipatory role, but in the second, we are dealing with a
formalization of human behavior. While Turing's concept is extremely useful in science, we
need to remember that it is merely a successful approximation of nature’s mathematical
forms.12
5. Pancomputationalism without Turing
The approximate nature of scientific theories leads to a question: Could we use non
anthropomorphous computational processes for hypercomputation.13 It seems reasonable to
use the well-defined Turing model as the base for further developing concepts of
hypercomputation, but this strategy could be misleading because it includes some
anthropomorphic assumptions.
So, is the concept of non anthropomorphous computation even conceivable? Theoretical
physics may help here by suggesting some ways of thinking about not-anthropomorphic-
computing-nature. Regardless, we know how to use some non-computational values (in the
Turing sense) as an input for physical processes. For example, nature can give us a series of
random values that can be used in hardware random value generators.
We know that for non anthropomorphic computation we need to formulate a new
conceptual framework, because modern concepts of computing are based on
anthropomorphic assumptions. For example, excluding digital ontology for its
anthropomorphous basis leads to the concept of continuous representations, but simply
replacing a digital version with a continuous one may not suffice due to the need for a
10 Mycka [28 p. 257], for example, suggested that the newest physical theories show possibilities for
computational systems with infinite resources. For him, this could be an argument for rejecting the human
ideal of computing as a model of computation, so it is therefore an argument for replacing digital
anthropomorphous computation with a nonanthropomorphous analog version.
11 We are talking here about the mathematical structures of nature as conceptualized by Michael Heller [20] (see
also [24]. Similar concept see e.g. [17].
12 The anthropomorphisms in Turing's model could also play a heuristic role (anticipatory anthropomorphisms).
One attempt to change the role of the anthropomorphisms in the concept of computation is the attempt to
formulate the theory of Real Recursive Functions (which is described further in this section).
13 From the pancomputationalist point of view, we are using a tiny subset of the computational power of nature
by applying Turing’s model of computation to physical systems. All computer engineers know how to apply
Turing’s model to transistor-based electronic systems, but they also realize how redundant these computations
are. By imposing Turing's model on the nature we are also excluding the hypercomputations.
49
Deanthropomorphized Pancomputationalism ...

deeper reconstruction of the conceptual framework (which will be discussed in the
following section).
We should remember that the basis of non-anthropomorphous computation has already
been conceptualized [26]. Moore proposed a model of idealized computation in continuous
time that shows some possibilities for the construction of non anthropomorphous
computing systems. In the last decade, further theoretical development has taken place,
such as the generalized theory of Real Recursive Functions [8]. Such theories promise a
new paradigm in computing.
This is arguably not the final step in the development of new concepts for computation,
however. For ontic pancomputationalism, only nature itself can define the target, but for the
concept of a computer, there are some additional constraints.
6. The problem of encoding
One of the problems we face is a problem of decoding and encoding that is seen as the
essence of computing. Without encoding or decoding we would trivialize the concept of
computing. There could be no difference between physical process and computing.
To clarify this problem let us look at the formal definition of computing and computer
formulated by Beraldo-de-Araújio et al. [4]. Computation can be formally represented by
the following two definitions:
Definition 1: A process is a function P: I → O such that its domain I
is a set whose elements are called inputs and its co domain O is a set
whose elements are called outputs, while both I and O are subsets of a
physical world.
Definition 2: A computer is a function from a set of
input symbols I to a set of output symbols O such that is
generated from via a computation. A process is
computational if P is generated by a computer C, i.e.,
for all , where is a symbolic representation of .
According to Beraldo-de-Araújio et al, the essence of computation is “symbolic
manipulation” and the computer is mapping function between two sets of symbols. This
property of computing is also claimed by Horsman et al. [21 p. 15]. They claim that what is
most important for computing processes is the possibility of encoding and decoding:
Without encode and decode steps, there is no computation; there is simply a
physical system undergoing evolution. This, then, is one of the key ways in
which this framework distinguishes between a physical system ‘going about
its business’, and the same physical system undergoing the same physical
evolution, but this time being used to compute. This is how we can escape
from falling into the trap of “everything is information” or “the universe is a
computer”: a system may potentially be a computer, but without an encoding
and a decoding step it is just a physical system.
50
P. Polak, R. Krzanowski

Horsman et al. show that the relationship between encoded structures is essential to the
encoding/decoding process. This then translates into the notion that our theoretical
structures describing the world should be isomorphic to nature’s own structures.
Considering ontology, our theoretical descriptions will always be approximations if
ontology is limited only to the ontological commitment of accepted theories. Thus, the
encoding/decoding process depends on having an appropriate theoretical description. For
non anthropomorphic computations, we would not know how to achieve this.14
7. Conclusions
We argue in this paper that to overcome limitations of the concept of computing based on
the Turing model we need to look somewhere else. One source of potential inspiration for
the new conceptualization of computing is nature or natural computations.
Our current conceptualization of natural computations is denoted as
pancomputationalism. Most of its variants, either implicitly or explicitly, rely on Turing’s
conceptualizations of a computer and computing, bringing with it, its the limitations and
conceptual framework. One of the obvious problems with the Turing model (we argue) is
its anthropomorphic roots. And antropomorphisation, while seductive, always puts limits on
the associated concepts.
Thus, considering the current conceptualizations of pancomputationalism as strongly
informed by the Turing model, we argue that a new concept of computers and computing
while inspired by nature, should not be based on the anthropomorphic framework of the
Turing model. That is what we call as “deanthropomorphisation of pancomputationalism”.
We do not have or propose the solution to the problem of computation without Turing.
However, we hope that the discussion of the limitation of the Turing-based models of
pancomputationalism will encourage the search for the new, nature-inspired concepts of
computing.
Acknowledgments
We would like to thank Paweł Stacewicz for discussions and some bibliographical help.
14 A pessimistic vision of the prospects of modeling of natural computations was described by Mycka [28 pp.
257–258], who stated that in practice, the problem of the boundaries of decidability in analogous computation
is relative to accepted physical theories. He also suggested that there probably were some aspects of nature that
could not be analyzed by humans, so the process could only be simulated rather than modeled.
51
Deanthropomorphized Pancomputationalism ...

References
[1] Agassi, J., Anthropomorphism in Science. In: Dictionary of the History
of Ideas: Studies of Selected Pivotal Ideas (Editor: P. P. Wiener). New
York: Scribner, 1968, 97–91.
[2] Anderson, N.G., Piccinini, G., Pancomputationalism and the
Computational Description of Physical Systems [preprint]. 2017.
[3] Barrow, J.D., A New Mathematics for a New Era [Matematyka nowej
ery]. Philosophical Problems in Science (Zagadnienia Filozoficzne w
Nauce), (16), 1994, 87–99.
[4] Beraldo-de-Araújo, A., Baravalle, L., The Ontology of Digital Physics.
Erkenn, 82 (6), 2017, 1211–1231.
[5] Blass, A., Gurevich, Y., Algorithms: A quest for absolute definitions.
Bulletin of the European Association for Theoretical Computer
Science, 81, 2003, 195–225.
[6] Copeland, B.J., What Is Computation? Synthese, 108 (3), 1996, 335–
359.
[7] Copeland, B.J., The Broad Conception of Computation. American
Behavioral Scientist, 40 (6), 1997, 690–716.
[8] Costa, J.F. et al., A foundation for real recursive function theory.
Annals of Pure and Applied Logic, 160 (3), 2009, 255–288.
[9] Cubitt, T. et al., Undecidability of the Spectral Gap. Nature, 528
(7581), 2015, 207–211.
[10] Davies, J., Anthropomorphism in science. EMBO reports, 11 (10),
2010, 721–721.
[11] Denning, P.J., Computing is a Natural Science. Communications of the
ACM, 50 (7), 2007, 13–18.
[12] Dodig-Crnkovic, G., The Development of Models of Computation with
Advances in Technology and Natural Sciences. In: Proceedings of The
6th AISB Symposium on Computing and Philosophy: The Scandal of
Computation - What is Computation? (Editors: M. Bishop and Y. J.
Erden). 2013, 1–8.
[13] Dodig-Crnkovic, G., Müller, V.C., A Dialogue Concerning Two World
Systems: Info-Computational vs. Mechanistic. In: Information and
Computation (Editors: M. Burgin and G. Dodig-Crnkovic). Singapore:
World Scientific Publishing Co., 2011, 149–184.
[14] Floridi, L., Against digital ontology. Synthese, 168 (1), 2009, 151–178.
52
P. Polak, R. Krzanowski

[15] Floridi, L., A defence of informational structural realism. Synthese, 161
(2), 2008, 219–253.
[16] Fredkin, E., An introduction to digital philosophy. International
Journal of Theoretical Physics, 42 (2), 2003, 189–247.
[17] French, S., Ladyman, J., In Defence of Ontic Structural Realism. In:
Scientific Structuralism (Editors: A. Bokulich and P. Bokulich).
Dordrecht: Springer Netherlands, 2011, 25–42.
[18] Heller, M. et al., Noncommutative Unification of General Relativity
and Quantum Mechanics. Journal of Mathematical Physics, 46 (12),
2005, 122501.
[19] Heller, M., Sasin, W., Noncommutative Unification of General
Relativity and Quantum Mechanics. International Journal of
Theoretical Physics, 38 (6), 1999, 1619–1642.
[20] Heller, M., Dispute around sructural realism [Spór o realizm
strukturalistyczny]. In: Filozofia i wszechświat: wybór pism. Kraków:
TAiWPN UNIVERSITAS, 2006, 215–234.
[21] Horsman, C. et al., When does a physical system compute?
Proceedings of the Royal Society A: Mathematical, Physical and
Engineering Sciences, 470 (2169), 2014, 20140182–20140182.
[22] Koyré, A., From the closed world to the infinite universe. Charleston,
S.C.: Forgotten Books, 2008.
[23] Kracher, A., Imposing Order—The Varieties of Anthropomorphism.
Studies in Science and Theology, 8, 2002, 239–261.
[24] Krzanowski, R., Minimal Information Structural Realism.
Philosophical Problems in Science (Zagadnienia Filozoficzne w
Nauce), (63), 2017, 59–75.
[25] Marciszewski, W., Universe as a computer and eschata [Wszechświat
jako komputer i sprawy ostateczne]. Computerworld, (9), 1999.
[26] Moore, C., Recursion theory on the reals and continuous-time
computation. Theoretical Computer Science, 162 (1), 1996, 23–44.
[27] Müller, V.C., Pancomputationalism: Theory or Metaphor?. In:
Philosophy, computing and information science (Editors: R.
Hagengruber and U. Riss). London: Pickering & Chattoo, 2014, 213–
221.
[28] Mycka, J., Continuous and discrete computation as an
anthropomorphous and a physical concept of effective computation
[Obliczenia dyskretne i ciągłe jako realizacja antropomorficznej i
fizycznej koncepcji efektywnej obliczalności]. In: Światy matematyki:
tworzenie czy odkrywanie? Księga pamiątkowa ofiarowana
53
Deanthropomorphized Pancomputationalism ...

profesorowi Romanowi Murawskiemu (Editors: I. Bondecka-
Krzykowska and J. Pogonowski). Poznań: Wydawnictwo Naukowe
Uniwersytetu im. Adama Mickiewicza, 2010, 247–260.
[29] Pexton, M., Emergence and interacting hierarchies in shock physics.
Euro Jnl Phil Sci, 6 (1), 2015, 91–122.
[30] Piccinini, G., Computation in Physical Systems. In: The Stanford
Encyclopedia of Philosophy (Editor: E. N. Zalta). Metaphysics
Research Lab, Stanford University, 2017.
[31] Piccinini, G., Physical computation: a mechanistic account. 2015.
[32] Piesko, M., Uncalculable calculability [Nieobliczalna obliczalność].
Kraków: Copernicus Center Press, 2011.
[33] Pour-El, M.B., Richards, J.I., Computability in Analysis and Physics.
Cambridge: Cambridge University Press, 2016.
[34] Ringel, Z., Kovrizhin, D.L., Quantized gravitational responses, the
sign problem, and quantum complexity. Science Advances, 3 (9), 2017,
e1701758.
[35] Turing, A.M., On Computable Numbers, with an Application to the
Entscheidungsproblem. Proceedings of the London Mathematical
Society, s2-42 (1), 1937, 230–265.
[36] Turing, A.M., Systems of Logic Based on Ordinals. Proceedings of the
London Mathematical Society, s2-45 (1), 1939, 161–228.
[37] Turing, A.M., Computing Machinery and Intelligence. Mind, 59 (236),
1950, 433–460.
[38] Wolfram, S., A new kind of science. Champaign, IL: Wolfram Media,
2002.
[39] Zuse, K., Calculating Space [Rechnender Raum]. Elektronische
Datenverarbeitung, 8, 1967, 336–344.
[40] Zenil, H. ed., A computable universe: understanding and exploring
nature as computation. Singapore: World Scientific, 2013.
Received 20.06.2018, Accepted 11.12.2018
54
P. Polak, R. Krzanowski