ANewKindofPhilosophy

ManifestoforaDigitalOntology
JacopoTagliabue1

Preprintdraft. Thefinalversionispublishedin'IrreducibilityandComputationalEquivalence',H.Zenil(ed.),SpringerVerlag

Abstract: Stephen Wolfram's ambitious 'A New Kind of Science' (NKS) rethinks and rebuilds almost every scientific field in the light of the study of cellular automata and theiremergentbehavior.Alittleknownfactamongprofessionalphilosophersisthatthere is plenty of room for philosophy too in NKS: on the one hand, Wolfram's core arguments require sophisticated conceptual analysis to be properly assessed and evaluated; on the other, it is pretty clear that Wolfram himself regards 'A New Kind of Science' as the obvious premise for a 'A New Kind of Philosophy'. In this contribution, we shall focus mainly on the second part of NKS philosophical import: in particular, we take Wolfram's own analysis as a starting point to explore the answer to the following question: do philosophersneeddigitalphilosophy?Ouranswerwillbeyes,forWolframsownreasons and more. First, we argue that philosophy as a whole may benefit from the unorthodox intuitions delivered by the systematic study of CA; second, we outline three promising areas of research for this new kind of philosophy, highlighting that a digital approach to substantial and methodological issues may bring very interesting consequences for many contemporary debates in the discipline. Finally, we place digital philosophy into the wider context of contemporary sciences (computer science, Artificial Intelligence and cognitive sciences), arguing that a genuine interdisciplinary approach would help tackling in new waysthegreatestchallengesofthesefields.

Introduction Stephen Wolfram's ambitious 'A New Kind of Science' (hence NKS) explicitly calls for a radical revision of prettymucheveryscientificfield2:thedetailsofthestory(andarguablythestrengthoftheargument)vary from field to field, but the discoveries made by Wolfram by studying cellular automata (hence CA) are the coreofthewholebook. While CA are wellknown among scientists since the Eighties (invented by John Von Neumann in the Fifties3, though it is fair to acknowledge Wolfram himself4 for much of the popularity of the field thereafter),theyarenotaspopularamongprofessionalphilosophers,evenifmanycomputabilityconcepts are now part of any serious practitioner's toolkit. This is very unfortunate, since there is plenty of room for philosophyinNKS:ontheonehand,NKS'sboldestconclusionsandclaimsrestonfairlysophisticatenotions from epistemology (what can we know?), philosophy of science (how should we judge a scientific paradigm?) and even metaphysics (what is the nature of computational phenomena?); on the other, it is prettyclearthatWolframhimself5regards'ANewKindofScience'astheobviouspremisefora'ANewKind ofPhilosophy': Amongthem[thefundamentalissuesphilosophersaddress]arequestionsabouttheultimatelimits of knowledge, free will, the uniqueness of the human condition and the inevitability of mathematics.Much hasbeensaidoverthecourseof philosophicalhistoryabouteachofthese.Yet

1 2

SanRaffaeleUniversity(Milan),iLabs(Milan). Wolfram2002,pp.116. 3 SeeVonNeumann(1951). 4 SeeWolfram(1983),Wolfram(1984). 5 SeeforexampletheremarksinWolfram2002,p.10.

inevitably it has been informed only by current intuitions about how things are supposed to work. Butmydiscoveriesinthisbookleadtoradicallynewintuitions.6

In this contribution, we shall focus mainly on the second aspect of NKS' philosophical import7. In particular, we take Wolfram's own analysis as a starting point to explore the answer to the following question: do philosophers need digital, CAinspired philosophy? (Spoiler alert: yes, they do, for Wolfram's ownreasonsandsomemore). The paper is organized as follows: in Section I we sketch a theory of systematic philosophy and in its relations with other forms of rational enquiry (mathematics and empirical science). In Section II we introduce NKS discoveries ('the new intuitions') and see how they can be fruitfully applied to existing debates in contemporary philosophy. Finally, in Section III we shall argue that NKS and CA provide much morethanjustnewintuitions:theyprovidephilosopherswithnewtoolsthat mayhelpshapethediscipline andchangeitsplaceintheworld. I.Whatisphilosophy? Tooutsidersandeventoinsiders,fromtimetotimephilosophicaldisputesseempuzzling:sometimes what is at issue is not that clear (what is a possible world, anyway?8), sometimes it is hard to understand why we should favor one among many competing explanations (after all, there are supposedly crucial "thought experiments" in philosophy but they are not experiments in the scientific sense, right?). To further complicate the matter, philosophers do not agree on a definition, since the scope and method of the discipline are themselves matter of dispute within the discipline. Without the ambition of settling the issue once and for all, I shall propose a working definition (which I take to be relatively uncontroversial) that will be used in what follows to better understand the importance of Wolfram's remarks and CA for philosophy. Asphilosophers9,wearecommittedtocomeupwithalistofbasicentitiesandsimplerulesoutofwhich everything we see atoms, people, galaxies, mathematical objects, moral values, mental states can be built. Nonphilosophers10 can easily imagine the work as some sort of reverseLEGO: you start with the wholemodelinfrontofyouandthetaskistocompilethelistofitemsthatwereintheLEGOkitinthefirst place,togetherwiththeassemblymanual(theworldweliveinisahellofaLEGOmodel,soitisnowonder that philosophers are fighting all the time about which items should make the final list). As it is stated, our game raises an immediate objection: isn't it sciences (and especially physics) that tell us what the fundamental features of reality are? if so, why bother with philosophy at all? To understand why we need philosophy and sciences alike, it is important to understand what kind of LEGO bricks philosophers are after. While it is physics that tells us, say, that particle X's behavior is causing event E, it is philosophy that tells us what 'causing' means: sure, the fact that E is caused by X is grounded in some physical features of

Wolfram2002,p.10. The first together with a philosophically friendly introduction to CA is addressed at length in Berto, Tagliabue (2012). 8 Actually, the seemingly scifi notion of 'possible world' is a fascinating theoretical concept, useful in many branches oflogicandphilosophy.ThelocusclassicusisLewis(1986);foranuptodatereview,seeDivers(2002). 9 Caveat: this is especially true for ontology, which is in some sense the most fundamental branch of philosophy (a viewthatcanbetracedbacktoAristotle)andtheoneasweshallseewhichmaybenefitthemostfromCA(notthat Iampartial,butincidentally,itisalsomyareaofspecialization). 10 Nonphilosophers may wish to consult Conee, Sider (2007), which is a gentle introduction to the field written for nonspecialists.
7

our world, but unfortunately this does not account for the concept of causation in its full generality11. In other words, sciences presuppose a concept causation whose fundamental structure is not explained by science: upon reflection, this is not strange at all, since science and physics also presuppose summation, whose fundamental structure is explained by mathematics, and implications, whose fundamental structure isexplainedbylogic. Philosopher's LEGO model is thus made by the most important qualitative features of our world (like causation,identity,morality,rationality,andsoon),theconceptsweuseinscienceandeverydayreasoning to make sense of what happens around us12. So the question becomes: among these concepts, can we single out a minimal set of fundamental ones to which we can reduce all the others? To further drive this pointhome,considerthefollowingscreenshotsfromConway'sfamousCA13,theGameofLife:

t1 t12 CA practitioners will recognize a socalled glider floating in the space. A software may surely help us calculate where the glider will be after n timesteps in the universe evolution; but what is the correct descriptionofthesituationfromanontologicalperspective?Takethetwofollowingcharacters: A) Philosopher A agrees with a commonsensical account: we are seeing one object moving in the space. B) Philosopher B says that no object is moving: she insists that there are no composite objects, just atomiccells.Sincecellsdonotmove,movementisimpossible. Of course, both philosophers can agree with the software calculation, since the LEGO model they see is exactly the same; however, the qualitative notions they employ in their descriptions differ a lot14: only for (A) cells can be combined to form further "emergent" objects, so that only in (A)'s universe things may be

t0

Asasmallillustration,consider:'Inflationcausedunemployment'.Howwouldyouevenbegintoexplainthisrelation usingphysics? 12 Sayingthatanaccountofcausationisnotthatimportantwillnotdoforobviousreasonsforexample,thefactthat causationisattheheartofresponsibility,afoundingnotionofoursociety. 13 See Berkelamp, Conway, Guy (1982). For a demonstration of Life amazing capabilities, a quick Internet search will givethedesiredresult. 14 Ofcourse,anotherphilosopher,C,mayarguethatthisisjusta"semanticaldispute",aproblemwithwords,notwith theworld'sfundamentalstructure.Thisisarespectableposition,drawingoncomplexbackgroundinthemethodology of ontology and metaphysics, but we shall not pursue it here (see Chalmers, Manley, Wasserman (2009) for a collection of works in metametaphysics). It is worth noting that metaphysical skepticism is itself a philosophical positionandonewhosepractitionersusuallyendupexplaininginphilosophyjournals.

11

created and destroyed. If this seems a rather unimportant fact concerning Life, just substitute 'gliders in a CA'with'personsinouruniverse':arewereal(or,whichisthesame,canwedie)15? Now that we hopefully have some idea of the scope of philosophy, it is time to address two crucial questions regarding its method: how did Philosopher A and Philosopher B reach their conclusion? How shouldwedecidebetweencompetingtheories? To answer these questions, imagine playing the reverseLEGO game again (say, with a LEGO cathedral). You have the model in front of you you cannot break it, you cannot touch it and all you can observe are the model's macroscopic features: you see the church's spires and gothic gargoyles, but you can't see how they have been built. There are countless LEGO kits compatible with the model (i.e. they are such that you canhavecountlessindistinguishablecathedralsstartingwithdifferentkits),sohowdoyouchoose16? A first, obvious desideratum is conceptual economy. 'Less is more', 'Occam's razor', 'Lex parsimoniae', 'Simplex sigillum veri'17 are all effective slogans for this principle, stating that we should favor simpler theories(theorieswithfewerprimitiveorundefinednotions)overmorecomplexoneswhentheexplicative poweristhesame.Inourcase,wecouldtrytosolvetheLEGOmodelbyreproducingthecathedralwiththe fewestpossiblebricktypes:westartwithjustonetype(say,thesmallerLEGObrick)andthenweintroduce others only if we cannot continue with what we have already included. If we do the job with rigor, we shouldendupwiththesmallestpossiblelistoffundamentalbricktypes. Unfortunately,whilethismakesmuchsenseinourexample,itoverlookstwocrucialpointsinsolvingthe game: first, we do not know the "smaller LEGO brick" to begin with; second, it may be very hard to judge when we cannot continue with what we already have in the list anymore. To appreciate the former difficulty, it is enough to note that our evidence is compatible with a very simple assumption: the model is builtfromauniquegiantbrickcathedralshaped;ifwedidnotknowthatLEGOdoesnotproducecathedral shaped bricks, we could not rule out this hypothesis. To appreciate the latter problem, we point out that even in the presence of a complete kit, we may still be unable to assemble the cathedral: the pieces are all there, but without enough ingenuity, we simply cannot make them work so we may be tempted to introduceadhocpiecestocompletethemodel. In both cases, our search for a solution heavily relies on intuitions18: we have a strong intuition that the cathedral is a composite object and we have a strong intuition that the smaller LEGO brick, by itself, is not enough to produce the church's spires and gothic gargoyles19. But where do intuitions come from? For the most part from common sense and science. This is why philosophy may strongly benefit from a detour in other fundamental disciplines: physics and mathematics, in particular, provide us with new intuitions that often correct commonsensical arguments. To take a famous example, for centuries the infinite was considered to be without any interesting structure: thanks to Cantor, many of us have now changed our intuitions into a more robust conception of infinite, allowing a very rich and interesting structure (some may still believe that there are strictly more integers than even numbers; but we know better); in other words, things we believed possible turn out to be impossible, and vice versa. Of course this is not to say
In the example several philosophical topics are intertwined: for an introduction to the themes of composition see Casati,Varzi(1999);forthemetaphysicsofchangeseeSider(2005). 16 As we shall remark, this is also a problem for different scientific theories explaining the same set of facts (the so called"underdeterminationproblem",seeStanford(2009)). 17 Literally,'SimplicityisthemarkofTruth'. 18 SeealsothewisewordsinLewis1983,pp.ixxi,andHudson2006,pp.1315. 19 In many respects, this image is an oversimplification: in particular, no two theories in general explain (equally well) exactly the same set of phenomena, so the final judgment is a tradeoff between theoretical simplicity and explanatorypowerabalancewhichmaybeitselfmatterofphilosophicaldebate(thatiswhywewillneverrunoutof work). It is worth emphasizing that almost any "fundamental" discipline shares the problem: for example, competing "theoriesofeverything"inphysicsarenotjudgedoverexperiments,butovertheoreticalconsiderationsofconceptual economyandintuitionsofwhatmaybea"possible"fundamentallayerofreality.
15

that physics and mathematics may solve philosophical puzzles by themselves, since, as we have seen, the structures investigated by philosophy are somewhat presupposed by both. However, while philosophers sometimesdeclarefromthearmchairtheimpossibilityofXjustbecauseourimaginationdoesnotyetcover X, good science has always been doing a great job in broadening our horizons. To the extent that CA are goodscience,itshouldnotbesurprisingwehavesomuchtolearnfromthem. II.Anewkindofintuitions As we have seen philosophy is, in a strong sense, dependent upon intuitions (together with theoretical considerations of simplicity, explanatory power, etc.). Therefore, when science challenges previously held intuitions, we should expect our philosophical perspectives to be challenged as well. In a nutshell, this is also Stephen Wolfram's idea when he claims that the 'discoveries in this book [NKS] lead to radically new intuitions'20. What these discoveries and these intuitions are is obvious to any reader of NKS, however a quickreviewofthebook'smainthemesmaybeusefulasareference21: NKS1) Simple rules may produce very complex behavior: contrary to what we think, complex behaviordoesnotrequirearichontologytobeginwith(prettymuchaswithLEGO,itisastonishing thecomplexityyoucanachievewithatinysetofprimitives). NKS2) Computation is everywhere: computational phenomena are not limited to computers; in some sense, computation is the most fundamental process in reality, one to which everything else istobereduced. NKS3) Universal computation is much more widespread than what we think: many systems, when investigated through the lens of computability, turn out to be equivalent in a precise mathematical sense (i.e. they all can compute the same set of functions and this set comprises every possible computablefunction22). NKS4) The ability to predict the world is severely constrained by (NKS3): once a system is proved equivalenttoaUniversalTuringMachine(UTM),itcanbeshownthatnoalgorithmicprocedurecan accurately predict its behavior we can only run a simulation and wait for the result to be computed step by step. This, in turn, highlights the importance of computerassisted discovery in anyareaofrationalenquiry. NKS5) Digital universes behavior is phenomenologically rich: in particular, digital universes may be indistinguishable from continuous counterparts due to their ability of hosting an incredible variety ofpatternsatdifferentlevelsofgranularity. Taken together, (NKS1) (NKS5) are a strong challenge to our intuitions. Granted, some of these claims are not new to philosophers around the globe: see, for example, the survey in Piccinini (2010) regarding (NKS2),orconsiderthefollowing,strikingdeclarationfromDanDennett:
20 21

Wolfram2002,p.10. Caveat: the list is not meant to be exhaustive nor to be the list Wolfram himself would have written; however it nicely summarizes the most important points for what follows. Moreover, while some claims can be held independentlyfromtheother,theyaremostcompellingwhenpresentedasasingleworldview. 22 They are all equivalent to a Universal Turing Machine; according to the ChurchTuring thesis, each and every computablefunctioncanbecomputedusingthisdevice.

Every philosophy student should be held responsible for an intimate acquaintance with the Game ofLife.Itshouldbeconsideredanessentialtoolineverythoughtexperimenter'skit,aprodigiously versatile generator of philosophical important examples and thought experiments of admirably clarityandvividness.23

However, it is fair to say that NKS as a whole constitutes an exceptionally strong case for what Ed FredkincalledtheFiniteNatureHypothesis: FiniteNatureisahypothesisthatultimatelyeveryquantityofphysics,includingspaceandtime,will turnouttobediscreteandfinite;thattheamountofinformationinanysmallvolumeofspacetime willbefiniteandequaltooneofasmallnumberofpossibilities.()WetakethepositionthatFinite Nature implies that the basic substrate of physics operates in a manner similar to the workings of certainspecializedcomputerscalledcellularautomata.24 To appreciate an original contribution of NKS to philosophy, let us analyze Wolframs approach to the problem of free will. What is the riddle of free will? We would like all of the following statements to be true,but(hereisthetrick)itseemstheycannotbepossiblyalltrue: 1)Thephysicalworldisdeterministic25andcausallyclosed26(i.e.:onlyphysicaleventsmaycausephysical events). 2)Theworldphysicalstateattntogetherwiththelawsofnatureissufficient,foreachm,foranyphysical stateattn+m. 3) If we have free will, for each action A we have freely deliberated and performed, we could have done otherwise. 4)Wehavefreewill. (1) is an empirical statement: if tomorrow we discover that biology is not deterministic or that non physical entities (like ghosts or Greek goddesses) may influence physical events, we will find that (1) is indeed false. However, for scientists and philosophers holding (1) is tantamount as holding a naturalistic perspectiveontheworldso,itwillbetterbetrue.(2)isaconceptualtruth,thedefinitionofdeterminism: if the world is deterministic, the big bang and the physical laws necessitate the evolution of the entire universe. (3) is basic conceptual analysis: part of what it means to have free will is to not be necessitated when one is acting. Finally, (4) is an intuitive truth: we do feel we are free moreover, we arguably assign moral responsibility based on this assumption (i.e., youre a good person because you made a good action wheninsteadyoucouldhavemadeabadone).

23 24

Dennett1991,p.37. Fredkin1993,p.116. 25 At the level relevant to explain our actions say, the level of biological processes in the brain. Of course, physics may not be deterministic at its bottom level (say, quantum reality), but it is obscure how this fact would vindicate freedomofthewill. 26 SeeKim(1993)forthisclaimanditsrelationtothemind/matterdebate.

Thecrucialquestionistherefore:howisthetensionbetween(2)and(3)tobeexplained?Somebelieves that the intuition is correct: there is a tension and it cannot be resolved: a deterministic world is not compatiblewithusbeingfree27;so,if(1)stays,itturnsoutthatfreewillislikeunicornsandchimeras,i.e.a fictional entity. Many philosophers, however, contend that the tension can be explained away, giving us all the free will we desire; interestingly enough, Daniel C. Dennett, one of the most important contemporary thinkersinthistradition,employsCAtoarguethatyoucannotinferthatyouractionsareunavoidablefrom the fact that the basic physics is deterministic28. Stephen Wolfram attempts his own analysis starting from (NKS1)(NKS5): From the discoveries in this book it finally now seems possible to give an explanation for this [free will].Andthekey,Ibelieve,isthephenomenonofcomputationalirreducibility.29 Computational irreducibility is precisely captured by (NKS4): although a system follows definite underlying laws, its overall behavior can still have aspects that fundamentally cannot be described by reasonable laws30. Pretty much as atoms in a CA, our neurons follow simple rules; however, whenever a system reaches a threshold in its computational complexity, we simply cannot readily make predictions about the behavior of the system31: we attribute free will to humans because there is no way to accurate predict their behavior, so that the concept of freedom fills the explanatory gap between the microlevel and the macrophenomena32. According to Wolfram, CA play a leading role in providing a new framework to understand the phenomenon: while explanations from chaos theory and quantum randomness have been recently proposed, nothing like this is actually needed33: thanks to the new kind of science, we have now a precise model of how the explanatory gap works. Good science has shown us once more how to crackariddle(howunpredictabilityarisesfrommicrodeterminism)wepreviouslydidntunderstand34. III.Philosophyinadigitalworld NotwithstandingtheimportanceofotherNKSthemesforexistingphilosophicaldiscussions35,wewishto further pursue Wolframs claim about the global philosophical import of CA. As Andrew Ilachinski points out, CA can be fruitfully used as conceptual vehicles for studying pattern formation and complexity36: if philosophy can be thought of some kind of reverseengineering of realitys structure, CA are a most precious conceptual lab to explore emergent behavior in its purest form. In particular, we shall explore three different ways in which CA may improve the method of philosophy as it is actually practiced all of themareexpansionsofWolframscoreideaontheimportofCAbasedintuitionsforknowledgeingeneral.
SeeVanInwagen(1983)foranexcellentstartingpointtothecontemporarydiscussion. SeeDennett2003,pp.4044.ForaphilosophicaldiscussioninthecontextofCA,seeBerto,Tagliabue(2012). 29 Wolfram2002,p.750. 30 Wolfram2002,p.750. 31 Wolfram2002,p.751. 32 Fellow philosophers may indeed argue that Wolframs claim is that there is no real free will, just an epistemic limitation we call free will. We leave to the informed reader to decide whether this position is a version of compatibilismorsomeformofeliminativism. 33 Wolfram2002,p.752. 34 Of course, Wolframs argument hardly settles the issue once and for all. However, its very important to recognize howthecreative,yetrigorousapplicationofascientificdiscoverytoaconceptualproblemmayhelpusseepreviously unnoticedconnections. 35 For example, the nature of space and time (Wolfram 2002, pp. 481496) and the metaphysics of computation (Wolfram2002,pp.637714). 36 Ilachinski2001,p.7.
28 27

III.AReverseengineeringreverseLEGO A first obvious point in the 'New kind of philosophy' Manifesto is the use of digital universes (say, Rule 110 or The Game of Life) to model existing philosophical theories: what happens, in other words, if we let ourselves apply philosophical reasoning in a world whose starting kit is known a priori? As we have seen before37, CA may provide the purest environment in which different intuitions may be vividly tested and theories may be compared: while CA per se (just as mathematics and physics) cannot solve, say, the metaphysicalproblemofchange,observingtheargumentsinvolvedinthedisputeinthecontextofadigital universemayhelpbringsomeclaritytothewholedebate. Moreover,philosophicalproposalshavebeendevelopedinconnectionwithoureverydayreality,butitis not always clear how to assess them due to the intricate network of concepts and intuitions that come in play when we observe the complex world we live in maybe, simpler yet rich worlds, may help. As an example,considerthefollowingtwosentences: L1)Allgoldspheresarelessthanamileindiameter. L2)Alluraniumspheresarelessthanamileindiameter. Theyarebothtruegeneralizationsaboutourphysicalworld,yettheydifferinanimportantrespect:(L2), but not (L1), is what we would regard (as scientists) as a true law of nature, something connected to the deep structure of the world and not the result of a sum of accidental facts. What is this difference groundedupon?Orwhichisprettythesamewhatis,generallyspeaking,alawofnature?Obviously,itis not the logical form of a sentence that makes it a law: (L1) and (L2) are both universally quantified statementsinvolvingphysicalfeaturesofourworld.Sowhatisit38? Under a popular account, natural laws are the true generalizations contained in the best deductive systemdescribingourworld39.Whatdoesthatmean?TakethelongandsomewhatconfusingBookOfThe World, containing all the facts about the evolution of our Universe (say, things like the Big Bang, the Halley comet, the Solar System, exhaustively described in the appropriate scientific language). We can hope to systematize this vast knowledge of particular facts with several possible axiomatic systems, competing for the best tradeoff between simplicity, strength, predictive power, etc.: some are conceptually simple, but not enough powerful to deliver the content of the BookOfTheWorld, some are powerful, but at the cost of many axioms and primitive notions40. Suppose the system with the best tradeoff is S: then, the natural laws of the world are all and only the true generalizations contained in S's axioms or theorems. Going back to our earlier case, we can see how the account elegantly manages to solve the problem: since, presumably, S contains axioms concerning elements' stability and decay but not axioms concerning an upperboundtogoldspheres'diameter,just(L2)canbederivedfromSandthatiswhyitisalawofnature. Apartfrompurelyconceptualconsiderations41,astrikingdifficultywithsuchanaccountisthat,textbook examples aside, no one has the slightest idea of how this system S may look like. Actually, the situation is even worse than that: no one has the slightest idea of how competing systems may look like, how the alleged tradeoff may present itself, what strategies we may use to judge them. So, apart from purely
ThereadermayalsowishtorecallDanielDennett'sremarksintheprevioussection. SeeCarroll(2010)forageneralintroduction. 39 SeeLewis(1973),Lewis(1994),drawingfromideaspresentedinRamsey(1978). 40 To avoid problems of language specificity, it is usually added the constraint that the predicates in the system stand for"naturalproperties"(whichisanothertechnicaltermthatmaybenefitfromaCAmodeling).Forthepurposeofthe presentdiscussion,itisjustimportanttonotethatsystemsareinsomewaynormalizedbeforethecomparison. 41 Suchasthefactthat,generally,theworldissupposedtobethewayitisbecauseofthelawsgoverningitsevolution, nottheotherwayaround.
38 37

conceptual considerations, we may still have the doubt that the theory is not (as we philosophers love to say) extensionally adequate, i.e., sometimes it fails to recognize a law in a lawful generalization and/or sometimesitsaysthatanunlawfulgeneralizationisalaw. DigitaluniversessuchasLifemaybeveryhelpful:wecaneasily(sotospeak)compiletherelevantBook OfTheWorld and then ask ourselves: What is the best system that can generate this amazingly rich universe? Arguably, if we take as axioms the initial conditions of a Life's run plus Life's fundamental dynamics, we should obtain a simple and strong system capable of generating all the particular facts contained in the BookOfTheWorld. However, the system as it is will not generate any easily recognizable laws about, say, the movement of gliders under several conditions. Does this mean there are no laws concerning gliders? Even if we acknowledge their "derivative", emergent nature, this does not mean we cannot exploit solid regularities to understand their behavior: why should not these be bona fide laws? Again, if this seems a rather unimportant fact concerning a CA, just consider the analogous question about ourworld:aretherelawsinthespecialsciences(frombiology,topsychology,toeconomics)orarephysical laws the only laws of nature, while anything else is, at best, a bunch of ceteris paribus conditions, or plain accidentalgeneralizations?42 Of course, there may well be theoretical merits in the view that Life's laws are just the basic facts about its microdynamics: this is not the place to further discuss the issue, nor it is CA job to settle it. However, it is crucial to see how rephrasing a theory in terms of CA may be helpful in seeing clearly the bold consequencesofsomemovesandthepotentialproblemsinvolved. III.BDevelopinganddebuggingphilosophicaltheories As a second point in our Manifesto, we want to stress the importance of building a rigorous philosophy ofCA.ConsideragainaclassicalLifedynamics:

t0

t2

t4

In this case we have an eater (southwest) devouring a glider (traversing the region in the lattice from northeast). It may seem perfectly legitimate to say that this is an instance of causation, i.e. to assert that 'Theeatercausedtheglidertodisappear'.Thecorrespondingcounterfactual,infact,seemstrue:hadn'tthe eaterbeingpresent,thegliderwouldhavenotdisappeared(butwouldhavereachedthesouthwestcorner in a few timesteps). According to many philosophers, this is good evidence that we have witnessed a causalinteraction. The concept of causation is one among the most crucial in understanding our world think for example about the importance of causal processes within science. Of course, many philosophical accounts are availableinthemarket toexplainthisphenomenon:some43takethecausalrelationassomehowprimitive, a brute, basic ingredient of our world. But here a rigorous ontology of the digital universe may be very
42 43

Foramoreextendeddiscussionofthispoint,seeBerto,Tagliabue(2012). Seeforexamplethenowverypopular"manipulabilityaccount"defendedinWoodward(2003).

helpful, highlighting that the above instance of causation is produced by the simple elements defining the Life universe nothing more than that, no strange "causal powers", no magical relations (how do we know that? Well, we built the very thing)44. All the concepts we have when we describe Life dynamics are somehow sufficient to deterministically generate causal phenomena: investigating how exactly this is possible is a challenging task that goes beyond the scope of this contribution, yet this is a clear, paradigmatic example of how an important philosophical concept may benefit from a systematic investigationofCAphilosophicalfeatures45. Closelyrelatedtotheideaofa"CAphilosophy"istheideaof"philosophicaldebugging"."Debugging"isa familiar activity to practitioners in Computer Science: in a nutshell, the idea is that, when an algorithm produces unexpected results, we test it step by step to understand where (and why) things start to go wrong.Considerasimplealgorithmtocalculatethefactorialofsomenumber,say1046: for(i=1;i=<10;i++) { intn=1; n=n*i; } returnn; If we run the algorithm, we would get 10 as the result of 10!, which is clearly wrong. So where is the mistake? If we follow step by step the algorithm, we notice that each time in our loop the variable n (the onestoringthepartialresultofthemultiplications)isrestartedto1;obviously,weshoulddeclareitoutside thelooptogettheintendedcomputation: intn=1; for(i=1;i=<10;i++) { n=n*i; } returnn; This is very nice, but, as philosophy is usually practiced, none of these beautiful practices of error checking is possible: we start with some assumptions, we somehow get weird outcomes, but there is not alwaysaneasywaytotellwhenwemadeamistakeinbetween. CA are very precise mathematical objects: as such, they can support a very precise philosophical structure, one which will be easy enough to translate into your favorite programming language so that it may run on universe's simulations. In that way, we can use simulations as benchmark of our philosophical theory, by letting the computer compute all the consequences of our assumptions (by teaching it, so to speak,toseethatrealitythrough thelensofourtheory).So,if thesimulation producesan eaterdevouring a glider and our theory does not detect causation, we may stop the universe and check the reasoning as appliedtothesituationtounderstandwhatiswrongandfixit.
Recall one ofour earlier lesson: we should not declare the impossibilityof X just because our imagination does not yetcoverX. 45 Asapioneeringexampleofthisattitude,seeBerto,Rossi,Tagliabue2010,Ch.3. 46 The syntax is admittedly C# like, but it is pretty much basic pseudocode (and yes, a recursion would be a more naturalchoiceforcomputingafactorial).
44

Probably, not all debates in contemporary metaphysics may be adequately represented in a digital universe, but surely for many of them CA would provide a new, exciting way to develop and test original accounts:somuchforthosewhothinkwecannotmakeexperimentsinphilosophy. III.CImprovingontologybasedAI As a last area of interest, we mention briefly the potential benefit of a CA philosophy to computer science, in particular to Artificial Intelligence. While, of course, it is not known how exactly we can make a computer truly intelligent, we have a pretty good idea of why it is, sometime, truly stupid as anyone who uses a smartphone to send messages well knows, the software sometimes suggests weird words since it doesnotunderstandthecontextoftheword,noritsmeaning(i.e.itsrelationtotheactualworldandother words).Information,perse,isquiteuseless:thekeytoknowledgerepresentationisstructured("semantic") information47. One way to improve computers' understanding is to teach them concepts and relations, modeled in some formal language they can easily manipulate: that is why over recent years formal ontology has been increasingly useful to solve real business challenges48. Back to our philosophical debugging, we see that by teaching lessons in ontology to our laptop we are arguably making it capable of assigning (proto)meaning to the words it uses to describe the digital universe. If this is true, any progress we can make in the small, yetqualitativelyrichworldofCAshouldbeapotentialcontributiontotheongoingdevelopmentofapplied ontologies. Moreover, since CA phenomena are particularly abstract, a thorough understand of their philosophicalsignificancemayturnouttobeofgreatinteresttoawideraudience,comprisingpractitioners incomputerscience,ArtificialIntelligenceandcognitivescience. Ifartificialmindsarethenextbigthing49,aCAinspiredphilosophymaybringthefieldcloserthaneverto agreatscientificrevolution. Conclusion NKS published ten years ago was the exciting, groundbreaking exposition of a new scientific paradigm, whose scope and success it is still too early to judge. However, we argued at length that, independentlyofthespecificclaimsmadeinthatbook,theconceptualframeworkputforwardbyWolfram isalively,engaging,challengingproposaltochange(sometimes,radically)ourhabitsofthought. Unfortunately, many philosophical issues directly connected with NKS could not be discussed first and foremost, a comparison between the philosophy of emergent properties and the science of emergent computation50, which strikes me as the most promising area for a new synthesis of tools and ideas. Hopefully, the presented material would help philosophers understand the importance of a direct acquaintance with NKS and digital universes in general; indirectly, we also suggested that science would benefit from a better acquaintance with philosophy, but a solid defense of this latter point need to be the focusofanotherwork.

47 48

SeeDavies,Fensel,VanHarmeleneds.(2003). SeeforexampletheworkscollectedinBorgo,Lesmo(2008). 49 AsarguedforexampleinKurzweil(2005). 50 SuchasCrutchfield(1994).

Up until the beginning of the 20th century, philosophers were often at the very frontier of their times science, something which became less and less common in the course of the century51. A New Kind of Philosophymayhelpthemregainthatplaceinthiscentury52.

While the facts are uncontroversial, it is harder to find a satisfying explanation. The increasing specialization of scienceandtheriseofanalyticphilosophysociallyorganizedasascienceforsurecontributedtotheunfortunate separation. 52 Many thanks to Francesco Berto (who, incidentally, taught me half the philosophy I know) and Giulia Livio for helpfulcommentsandsuggestionsonearlierdraftsofthiswork.

51

References Berlekamp E., Conway J., Guy R., 1982, Winning Ways for Your Mathematical Plays (Vol. II), Academic Press. BertoF.,RossiG.,TagliabueJ.,2010,TheMathematicsofModelsofReference,CollegePublications. Berto F., Tagliabue J., 2012, 'Cellular Automata', The Stanford Encyclopedia of Philosophy, Edward N. Zalta(ed.),<http://plato.stanford.edu/archives/sum2012/entries/cellularautomata/> BorgoS.,LesmoL.,2008,eds.,FormalOntologiesMeetIndustry,IOSPress. Carroll J. W., 2010, 'Lawsof Nature', The Stanford Encyclopedia of Philosophy, Edward N. Zalta(ed.), URL = <http://plato.stanford.edu/archives/spr2012/entries/lawsofnature/> CasatiR.,VarziA.,1999,PartsandPlaces,MITPress. Chalmers D., Manley D., Wasserman R., 2009, eds., Metametaphysics: New Essays on the Foundations of Ontology,OxfordUniversityPress. ConeeE.,SiderT.,2007,RiddlesofExistence:AGuidedTourofMetaphysics,ClarendonPress. CrutchfieldJ.P.,1994,'TheCalculiofEmergence',PhysicaD,75:1154. Davies J., Fensel D., Van Harmelen F., 2003, eds., Towards the Semantic Web. OntologyDriven Knowledge Management,JohnWiley&Sons. Dennett,D.,1991,'RealPatterns',JournalofPhilosophy,88:2751. DennetD.,2003,FreedomEvolves,VikingPenguin. DiversJ.,2002,PossibleWorlds,Routledge. Fredkin E., 1993, 'A New Cosmogony', in PhysComp '92: Proceedings of the Workshop on Physics and Computation,IEEEComputerSocietyPress,pp.116121. HudsonH.,2006,TheMetaphysicsofHyperspace,OxfordUniversityPress. IlachinskiA.,2001,CellularAutomata,WorldScientificPublishing. KimJ.,1993,SupervenienceandMind,CambridgeUniversityPress. KurzweilR.,2005,TheSingularityisNear,VikingPenguin. LewisD.,1973,Counterfactuals,HarvardUniversityPress. LewisD.,1983,PhilosophicalPapers(Vol.I),OxfordUniversityPress. LewisD.,1986,OnthePluralityofWorlds,BlackwellPublishers. LewisD.,1994,'HumeanSupervenienceDebugged',Mind,103:473490. RamseyF.,1978,Foundations,RoutledgeandKeganPaul. SiderT.,2005,FourDimensionalism:AnOntologyofPersistenceandTime,OxfordUniversityPress. Stanford K., 2009, 'Underdetermination of Scientific Theory', The Stanford Encyclopedia of Philosophy, Edward N. Zalta(ed.), URL = <http://plato.stanford.edu/archives/spr2012/entries/scientific underdetermination/>. VanInwagenP.,1983,AnEssayonFreeWill,OxfordUniversityPress. Von Neumann, J., 1951, 'The General and Logical Theory of Automata', in Cerebral Mechanisms in Behavior:TheHixonSymposium,JohnWiley&Sons. WoodwardJ.,2003,MakingThingsHappen,OxfordUniversityPress. WolframS.,1983,'StatisticalMechanicsofCellularAutomata',ReviewsofModernPhysics,55:601644. WolframS.,1984,'UniversalityandComplexityinCellularAutomata',PhysicaD,10:135. WolframS.,2002,ANewKindOfScience,WolframMedia.