Anthropomorphization of Artificial Intelligence

by Tomáš Kopeček

Brno 2015

Annotation

Artifical intelligence is becoming topic in scientific and more and more in public space as well. From the beginning it is followed by questions which outreach ordinary scientific discussion in terms of Kuhn's normal science. Is creating of human-like artificial intelligence moral? Will it replace mankind? Is existence of such being even possible if it has no soul? Should it have right to vote? Most of this discussion is influenced by many effects from position of science, religious stances or political goals and not last in the line materiality of engineering work. On the second side it has no impact only in scientific world, but also in its neighbourhood and whole society in forms of law or in formulation of public fears or power struggle of science and religion.

Topic is slightly researched by sociological, philosophical and juridical literature. I'll try to show overview of sociologically relevant aspects of this field. In opposite I'll not deal with topics which are clearly technical. Methodologically it will be horizontal survey which is rooted in different philosophical-sociological approaches from Bourdieu's theory of fields through Latour's view on materiality to Habermas' theory of communication. Text identifies four basic narratives of AI discourse (scientific, religious, securitization and artistic) with whose will be illustrating particular topics.

Keywords

Artificial intelligence, discourse, philosophy of science, religion, materiality

Table of contents

Disclaimer

This is non-edited long form of final short version . There are a few reasons to publish it here in this format. Firstly, there was some additional research which didn't fit into the shortened version. Secondly, it is easier to read it here. Thirdly, original data store distorted resulting PDF, so many readers are not able to correctly display it. (better file here). It is published late, but better now then never. If there are any typos, factual errors, etc. feel free to file a github issue. Otherwise, I'm not planning to update the text.

---

1 Introduction

1 Why

Three ugly words in title of something what should not be a research work but a bachelor thesis. That needs some kind of explanation. Let's cast naïve question: Why study something like Artificial Intelligence (AI) (something which doesn't exist) from sociological stance? and similarly naïve answer follows: Because it matters (and sociology deals with things which don't exist. Now we have to see it more broadly.

At first here comes AI. It is a thin stream which is born in twentieth century (but it has in many points quite ancient roots as we will see later) and is slowly gaining momentum to these days. Now AI concept is something which almost everyone is more or less familiar with.

From early fifties when Alan Turing comes with his famous Imitation game through all buzz of excited sixties when Marvin Minsky becomes a new apostle of AI religion until these days when there is no month without some progressive news from the field.

Story of AI is not only about technical expertise, but also encapsulates many others and a lot of them are sociologically relevant. Permanent struggle of science and religion stages another battlefield here which ironically shows more religious zeal on side of (positivist) science. Politics also become involved as there is a few forces which has to emancipate themselves through permeating law systems with AI-related concepts ending even with Charter of Fundamental Rights and Freedoms AI amendments.

There is still one word left — anthropomorphization. There is a reason, in fact more of them, why most work related to AI is thinking about it as about a human-compatible creature. We will see that it is quite problematic — from point of creature to human-compatible concepts. There are some reasons why experts expects it to be alike, but there are also not a few problems which such thinking raises.

2 What

Let's dive into main lines of thinking about AI and look to which social and cultural aspects played and still plays important roles here. Whole AI sector even if it looks like ultra-rationalist and firmly rooted in modern positivism is full of contradictions. We will go so far that in extreme position we will describe it as a religious project which builds on Judeo-Christian roots touched by hermetism and gnosticism. AI story is becoming part of a much larger cultural stream from Enkidu through hermetic homunculi, golems and entering modern era via Mary Shelley's Modern Prometheus — Frankenstein's monster. We will see how is AI discursively constructed, how this construct backfires to its authors and how still non-existent (strong) AI is changing existing world already.

Whole discourse on AI is quite complex and for purposes of this work I've to tear out just few stripes. To keep (not so false) complex system idea I'm going to not choose only one line but few narratives whose should serve as a base for limited horizontal overview of the field. Thus, we will go through scientific, religious, securitization and artistic stories. In the end it should draw picture how such detached topic as AI for most of population is can change and changes our everyday world.

3 From Where

Let's call this section Section of clarifications. First clarification is to say, from where I'm coming. I'm not born-and-raised sociologist. My epistemo-ontological origins lies near to the enemy. In fact, I'm coming directly from the eye of the storm — theoretical informatics, more specifically from mathematical linguistics branch directly building on Wittgenstein, Chomsky and mathematical intuitionism¹ speaking with Leibniz-like philosophical language. On the other side I've to deal with quite different languages like theological philosophy of Teilhard de Chardin.

We are both speaking about areas somehow related to AI problematic, but for many people each of these languages is incomprehensible. I wanted quickly illustrate how wide is the stream of thinking which I have to grasp. But instead of it I see that I'm falling to one of the modern traps — building dichotomy. Mathematics vs. philosophy, 2008 vs. 1949 So I'll try to fix it a bit. Let's add some new axes and let these two extremes slowly melt into the amalgam of (post)modernity.

4 Through What

From the point of philosophy I will be mainly operating on one side with Leibniz, Kant and Nietzsche — in topics related to ethics and epistemology. Reasons for that will be given in section on materiality. On the other side there will be Wittgenstein, Habermas [Habermas and Cronin, 2008, Habermas, 2003, Habermas and Outhwaite, 1996], Alexander useful in area of meaning construction. To the mix will be added some authors from the science of philosophy arena which will be touching the topic from more angles as it is the only reflective apparatus which could be used to reflect on scientific dimension of AI.

In sociology I'm quite influenced by Latour's notion of networks [Latour, 2007, Latour and Porter, 1991], nevertheless it will be used in somehow naïve way, as this is not research work and there are only tiny resources dedicated to AI topic. It will be set in contrast with Alexander's critic expanded by Hodder mostly concerning the materiality issues. Of course, we can't evade classical sociologists as Marx, Durkheim and Weber mainly through their critics or followers. Guide to the religious aspects is Geraci [Geraci, 2008, Geraci, 2014, Geraci, 2007, Geraci, 2011]

Of course there will be a lot of references to insider scholars and practitioners dealing with AI and related fields such as Alan Turing, Marvin Minsky, Hans Moravec [Moravec, 1988, Moravec, 1991] or Ray Kurzweil [Kurzweil, 2014, Kurzweil, 2001, Kurzweil, 2005].

There is no definitive methodology I'll be using. It could be some loss and on the other side some highlight of the work. So, if I've noted that I'll add few more axes to distorted image I'll be adding also sociological and philosophical approaches which should allow us to see more holistic picture.

Such mix of schools will provide, I believe, some insight into the field with wider scope than choosing one limited approach which would require some speculations and would be more appropriate for empiric research such is not intended part of this bachelor's thesis. From the same reason is not any specific method of inquiry used.

2 Narratives

1 Story of Science

Long story short. Science and in this case hard science is minefield of epistomologic and ontologic misunderstandings. And similar dangers lies on the sociology side. We will slowly navigate through it and also probably hit some hidden mines.

Narrative is probably most-known in public discourse. It is quite logic (even if word logic will be soon problematized) as it is related to origin of actors which are involved in AI development. In first place this narrative is natural for them and not in last it provides benefits. From the legitimization by science or possibility to use scientist status to utter new facts.

1.1 Scientific Version of History

Let's dive into history of AI as initial framing. As this is scientific story, I ignore more mystical roots and to demarcate some starting point I deliberately choose Leibniz. Some authors (Pratt, Churchland) puts him as grandfather of AI even if it sounds quite exaggerating. Umberto Eco] takes different approach and stays more with what Leibniz really formulated and thrived for — which is universal philosophical language characteristica universalis designed to describe everything in the universe. This idea is coming in the second half of seventeenth century. What is new with Leibniz approach is scientific approach. He is leaving searching mode of original God language and moving to modern way of constructive science. His vague design of philosophical language has heavily influenced even Kurt Gödel. Gödel himself is second big foundation stone of modern computation². On the technical side there is Pascal, Charles Babbage and Ada Lovelace (famous Enchantress of Numbers which we will see again later in religious pantheon of AI cult in (Discursive Construction of AI)) on their road to mechanical computer. ³

Alan Turing is finally the person which has brought this philosophical, mathematical and engineering streams of thinking to its climax and started the AI revolution. In year 1950 he publishes fundamental text with which contemporary scientific world still has not settled in satisfactory way. In Computing Machinery and Intelligence he creates famous Imitation Game where he proposes test (which is immediately named after him) which should verify if computer (machine, more generally) can think.

Noam Chomsky's Syntactic Structures are going to print in 1957. Following Wittgenstein's linguistic turn and merging it with formal logic, Chomsky once more opens door for computers to metaphysics. His essential concept of language classes is inherently connected to problems of computability and computation theory of brain modelling. Engineers got one more bullet to their crucial argument that brain is machine. When it got linked with Turing machine limits and Church-Turing thesis It based almost unbreakable foundations⁴. Research in physiology and biology in the same time adds to the theory of AI. It is probably not an accident that year before publication is first Dartmouth Project⁵ run, where the term Artificial Intelligence is coined (To what AI is and how it became term of itself is covered in Discursive Construction of AI. This is the point where AI is definitively connected with computers and computational theory of mind.

Pragmatism, as one of the post-positivist ways out, allows opening of full scale of differently educated theories, which further allows building model of human behaviour, brain and human (or human-like) manifests. Idea of so-called Rule-based reasoning as an inference model for human behaviour is born. It builds on same thesis as homo economicus. Man is driven by rational rules. It is emerging in the time when Skinner's radical behaviourism is going to be one of the leading schools of USA's psychology science. This paradigm evolves with psychology through Dennett's Computational Theory of Mind until contemporary trends in Computational Ethics [Allen et al., 2000, Torrance, 2008, Leuenberger, 2014] and finally Model-based Reasoning as convolution with neuroscience [Thagard, 2010, Blank, 2013, Jones, 2003]. Long-time paradigm shift is from constructing AIs from scratch to design of learning machines which can evolve on their own [Cristianini, 2014, Velik, 2010].

1.1.1 Philosophy

The philosophy approaches which are utilized by AI scientists, could be helpful to understand what drives and what also limits them. Philosophy in AI is divided to few quite distinct areas. First one is approach to science itself. What is the science, what is its purpose and how reflectively they look to it. Second (and definitely being most explored) field is ethics. And the last one is relationship between science and lived world.

In population and especially in academics there is present idea, that technology is not a field where theory of science is reflected well. If it is true, false or irrelevant is not of our concern now. I'll look to points where sociologists and philosophers looks upon AI science and how science builds its own philosophical foundations.

There is a tradition from Plato and Aristotle, through Leibniz and Newton to Russell which is slowly taking mathematics to its knees and forcing it to reflect upon itself. Especially Whitehead's and Russell's Principia Mathematica (of course pointing to Newton's Philosophiæ Naturalis Principia Mathematica has shown problems which are hidden in the base of math in the modern light. In the early twentieth century schools divided according to epistemological and ontological explanations to three basic branches: formalism, intuitionism and logicism. Mainly discussions of epistemological qualities directly influenced AI research. Floridi mentions that in year 1964 philosophy of AI had already produced more than a thousand articles [Floridi, 2004, p. 566] which is stunning compared to the fact that in that time computers were able to solve limited set of word mathematical problems which is quite far from what we imagine under word AI.

Mission

If we look to almost random paper, which is not purely technical, we will see some similar features. Typically, it is self-confidence that what science is doing here is important and even the mission of mankind. Creation of AI is inevitable. The Singularity denotes an event that will take place in the material world, the inevitable next step in the evolutionary process that started with biological evolution and has extended through human-directed technological evolution. [Kurzweil, 2005] ⁶ Background could be different — here Kurzweil reasons with evolution, otherwise it could be simple eternal value of knowledge like here: Artificial life is foremost a scientific rather than an engineering endeavor. [Bedau et al., 2000] — ignore engineers, do science no matter what.

Ethics

There is a lot of open questions regarding to ethics. We investigate few here and more later in Story of Religion which is definitely more concerned with it.

Questions being asked here are about what makes AI living, creature, human or superhuman. They are important in ethical part simply because they define how we will behave to and with AIs. Typical question which haunts the field is question if AIs are genuine or if they just simulate. Searle's Chinese room dilemma redefined this typical western dichotomy, where we can see Alexander's binary cultural codes. It is not important here if AI is compatible or indistinguishable from human. Important is what's its ontology. Another dichotomy is put if there is some personhood or not. Binary division is also seen if AI needs to have emotions. These few examples shows that general tendency is to grasp some specific concepts where almost all of them are taken from (philosophical) human traits and as such are not definable and try to find if they are applicable to AIs. These concepts can be defined relationally as human-only and as such is futile to try to map them to AIs. It mostly creates only false contradictions as there is no objective position from which it can be evaluated.

Final comment about this topic is reflection of these anhropomorphic terms via Coeckelbergh: Verbeek's turn to what things do remedies this problem, but his conclusion that we therefore should talk about moral artefacts or the morality of things goes at the cost of diluting the anthropocentric meaning we like to give to the terms moral and morality. [Coeckelbergh, 2009, p. 183] ⁷ We see here, that whole topic of ethical things is quite problematic in binary view. There are two ways out — to stop thinking in dichotomy and adopt non-discriminating approaches, which is sometimes proposed, mostly inspired by Latour or throw away this type of morality at all and stays on human-only level.

Philosophy strikes back

The new technologies make a public discourse on the right understanding of cultural forms of life in general an urgent matter. And philosophers no longer have any good reasons for leaving such a dispute to biologists and engineers intoxicated by science fiction. [Habermas, 2003, p. 15] as we see here, Habermas has quite clear view on the problem, that ethical and philosophical decisions are being made by unqualified engineers. Many philosophers are disturbed that tunnelled view of engineers should be contemplating about future of mankind. It is disputable if these reactions are relevant and if philosophy is reflective enough upon itself. There is a lot of opinions which says that opposite is true. When we are talking about Habermas we can point e.g. Latour's reaction on him. In [Latour and Porter, 1991, p. 60] he makes an attack based on idea, that Habermas intention is to defend modernity. According to Latour he denies postmodernity's right to existence (he categorizes him as pre-postmodern). In such situation is hard to believe that philosophy as a field has some superior rights to speak about AI ethics. We also see examples (especially Searle) which more defends their place in the postmodern science than to use rational arguments. In case of Habermas it is quite ridiculous compared with his theory of communicative action. Of course, only at first sight. We see, what he tries to communicate here philosophy is more special than engineering.

Nevertheless, Habermas slowly leads us to chapter about religiosity, with which philosophy is very closely related. The scientistic belief in a science which will one day not only supplement, but replace the self-understanding of actors as persons by an objectivating self-description is not science, but bad philosophy. [Habermas, 2003, p. 108] It is obviously another attack on incompetent engineers. If we ignore the tone and get base message, we've to agree. It is belief in science. It is not rationally defensible position. Habermas further explains that science is mostly performed as a religion. It is aligned with late Berger's diversion from secularization thesis. Society is no more secular. In this postsecular society still stays a lot of endemic places which are not aware of it. One of these is (western) science itself. Ironically, science which has proposed concept of postsecularism is disturbed by another postmodern attack on its primacy. Most of science (in this case hard-science) tend to ignore this interpretation and stays in secular phase, which is from outside-view interpreted as a religious institution [Evans, 2014, Knight and Murphy, 2010]. In such paradigm makes sense to communicate with science but definitely not to adhere to its ontological meanings.

Another example of fear from mathematical-naturalistic hegemony is shown in Floridi's paper. Informational theory of cognition and human behaviour and its concepts are in his view so powerful, that they ultimately dodge almost every criticism. Informational concepts are so powerful that, given the right level of abstraction (LoA) (Floridi and Sanders 2004), anything can be presented as an information system, from a building to a volcano, from a forest to a dinner, from a brain to a company, and any process can be simulated informationally — heating, flying, and knitting. [Floridi, 2004, p. 566] He concludes that in such case this theory lies in the realm of dogma. Such science is no Popper's science for Floridi. It is something what is pragmatic engineering but not contestable science. They have to leave that paradigm which on the one side proclaims Gödel's paradox and in the same moment does not want to measure itself by it.

Why I'm elaborating on these issues is my attempt to show that independently on real topic of discussion, AI creates arena for fight of two scientific fields. In Bourdieu's sense we see on one side hard-science with its questionable objective truth ⁸ and on the other side philosophy which pulls itself out of science to higher-ground of patronizing referee position above science. Fight is staged via standard scientific means such as papers and also out of the field, legitimizing itself and fight for public.

1.1.2 Materiality

The political task starts up again, at a new cost. It has been necessary to modify the fabric of our collectives from top to bottom in order to absorb the citizen of the eighteenth century and the worker of the nineteenth. We shall have to transform ourselves just as thoroughly in order to make room, today, for the nonhumans created by science and technology. [Latour and Porter, 1991, p. 136] This Latour's idea is one of the central problems of previously mentioned Computation Ethics. Ideas on which this concept builds are not uninteresting. We can see how process of theory choosing is driven by a) hegemonic (informatics) discourse (more in Discursive Construction of AI and b) technical possibilities. Especially second aspect is interesting from the point-of-view of sociology and even more from latourian perspective. We will see how these (basic) questions are practically without any freedom determined by technology which dictates its own rules. View on if what means ethical agent is basal for such discussion.

To speak with language of the field, scientists distinguishes between ethical-impact agents, implicit ethical agents, explicit ethical agents and full ethical agents (like in [Tonkens, 2009, DeBaets, 2014]). ⁹ Agents would lead us to Latour's concept of agency. In fact a lot of AI ethicists are acquainted with it and are using it in some way. The problem of agency concept is similar to one which is once and again hit by Latour's disciples and critics — it sounds too anthropomorphic. It no more matters that it is not ¹⁰, the problem is that is often perceived as such. In this concrete case the first two types of agency (ethical-impact and implicit) are connected more to things and latter two with humans (AIs, creatures, …).

While ethical-impact agent is something what has no real agency, it can be used in ethically-relevant way. Knife has no ethical impact until it is used to kill. Implicit ethical agent has some possibility to influence reality in such way. Airbag in car is designed to serve ethical purpose and it can be activated in right moment. Technically it is designed to be moral agent without possibility to choose immoral option. Explicit ethical agent has options to do morally relevant actions in both directions. Autonomous military drone can choose about target or not to shoot at all [Sparrow, 2007, Hallevy, 2013]. Full ethical agent has furthermore all other options which explicit ethical agent lacks. Simply, nowadays only full ethical agent is human.

Now back to previous claim — we naturally feel that the latter two are more human. They describe some creature which has options to choose relevant action. Both of them have free will in some way. This feeling is supported by human-like word agent. This discourse brings us more to anthropomorphic representation than before. Importantly, new artificial actors do not enter the stage from nowhere. We design them. We can give them the mask, the appearance we want. Since we love ourselves as humans, it is very likely that we will give them our own mask. Or the mask of the animals that remind us of our own, human infants. says Coeckelbergh and it also becomes one of problems in the field.

Generally we can see here (and whole ethics of AI is) tendency to interpret such complex machines in terms related to social world. Sharkey quotes psychological study which studied people evaluating other people (and in which case they were more positive if evaluated person was present) compared to people evaluating computers (in which case they were more positive if computer was present). Simply people treated computers like persons in this case. This has to fall in section about materiality as these things directly change way how we think and behave only by their mere presence. Indeed we are often quite willing to ascribe intentionality to much less human-like things e. g. My car doesn't want to start this morning. My computer is thinking about it. [Adam, 2008, p. 152] Such approach leads to problems that human traits are ascribed to machines and in the field of ethics it means (as easiest way) that human-targeted ethic (as Kant's is) is used on them. Some authors of course noted this problem and Adam, which is quoted here, proposes: The combination of the arguments of Latour on the attribution of agency, even moral agency to things, Dennett's as if intentionality, Magnani's moral mediators and Collins' and Kusch's delegation of action all lend support to IE as an ethics where things have as much place as humans. [Adam, 2008, p. 154 It is reflected by Latour's program: So long as humanism is constructed through contrast with the object that has been abandoned to epistemology, neither human nor the nonhuman can be understood. [Latour and Porter, 1991, p. 136]. Adam goes this way, while most of the authors simply ignores it and continues with anthropocentric approaches.

If ethicists have intention to create AMA (autonomous moral agents) as they call their holy grail, they are coming to trap what ethical means. If it has to be implemented in machine they have these options: a) design ethical model and implement it b) let ethical behaviour emerge from AI itself. Mostly from securitization perspective option b is out of the game as there is no way how to prevent potential damage ¹¹ [Davis, 2015]. So, only option a is relevant. This is the time when ethics hits the wall. Free will has to be technologically modelled first, in robot/AI design or in imagination, before we can fully work out the philosophical difficulties. [Coeckelbergh, 2009, p. 184] It is one of the problems. AI needs free will defined and implemented. Only after that there is a possibility to implement ethical behaviour. There are tons of proposals how to do that and every each of them is limited by technical possibilities. It is not by chance, that AI geeks are into Kant. It is simply too tempting to have global categorical imperative in AI. It seems to be simplest to program.

Now taking bridge to Kant. Andersons states, that Deontic logic's formalization of the notions of obligation, permission, and related concepts, make it a prime candidate as a basis for machine ethics. [Anderson and Anderson, 2007, p. 8] and further says that one of the central issues in machine ethics is trust and mechanized formal proofs are perhaps the single most effective tool at our disposal for establishing trust. We can easily see here, that parameter for choosing ethical theory or principles is not based on anything else than on some plausibility (use something which has legitimate status) and it has to be algorithmically easily described. When we set such rules, there is not many systems which satisfy them. We already limit ethical systems to deontic ones and it seems that only Bentham's utilitarianism and Kant's categorical imperative [Tonkens, 2009, Allen et al., 2000]. We will see in (Materiality) that these systems are not so well thought out for AI and have internal contradictions, but they are still top candidates because of their compatibility with programming.¹² Other types like virtue ethics have much more technical problems, that they are not considered at all or only as a supplement to one of the previous.

There is of course a lot of another material limitations and questions of related ethics, such as memory issues, biologically-derived ethics, human comfort or actual real-world problems as reappearance of trolley problems¹³ with autonomous cars. Whole field of materiality-impacted theory is also linked to embodiment and problems of AI with or without body [MacCormack, 2012, Kaur, 2013, Seaman and Rossler, 2008].

1.1.3 Rules of a Field

From Bourdieus's, view, we've to count on field of science. How science and especially positivist science is done make very large impact to how AI is constructed. Let's make take an example from Velik:

In 1969, for instance, M. Minsky and S. Papert published their famous work on Perceptrons [84] criticizing computational models of the nervous system, which showed with unanswerable mathematical arguments that such models were incapable of doing certain important computations [70]. This killed most research in neural computing for the following 15 years. Much later, S. Paper admitted in an interview [70]: Yes, there was some hostility behind the research reported in Perceptrons … part of the drive came from the fact that funding and research energy was dissipated … money was at stake. In recent times, these conditions have turned out to be particularly hard for scientists working in Brain-Inspired AI. [Velik, 2012, p. 44]

Marvin Minsky — one of the loudest proponents of AI caused so-called AI winter for almost twenty years not only from scientific right to critique, but also from political reasons like funding. We can't close eyes in front of this. It is likely that if in this case Minsky reflected this push, there still have to be many more examples when scientists are not able to recognize these non-voluntary pressures. Two biggest can be still easily imagined — military and business. Illusion that AI is independent of anything and it is purely scientific project is something what discourse builds on, but in the same time it is mere illusion.

The Pure Science trope is building on its independence, but as already Merton noted,

Science includes disinterestedness as a basic institutional element. Disinterestedness is not to be equated with altruism nor interested action with egoism. Such equivalencies confuse institutional and motivational levels of analysis… It is rather a distinctive pattern of control of a wide range of motives which characterizes the behavior of scientists. [Merton, 1973]

Ethos of scientific truth as goal for itself successfully hides most external influences and reifies itself building false legitimization base for AI scientists. Typical example where this issue is noted, but not reflected properly is in Campa's text where he states quest for pure knowledge is and can be part of the transhumanism agenda.¹⁴ There is even not much reflection why transhumanism should, or could, be part of science at all. Important is that it is somehow compatible with Pure science and that means, that it is good.

1.2 Discursive Construction of AI

Term itself — Artificial Intelligence was coined in 1956. Of course, there is a long history of how term was chosen and another long history how its meaning has been changing until today.

1.2.1 Case: Turing

Imitation Game made a significant impact on AI term. I'm not going into deep why Turing did this and did this in this exact way. I still would anticipate that he never expected what such relatively simple article will do in next hundred years. What was so extraordinary with this AI test, that it is still seriously considered even if it has quite a lot of methodological flaws (e.g. [Berrar et al., 2013]) that renders it completely unusable from scientific point of view?

Let's start with simple transcription. The crucial question which should be answered by proposed test is Can machines think? Human interrogator is set to position where he can ask two other creatures any question. By investigating answers, reaction times, mood, generally whatever he has to decide which of creatures is computer and which is human. Turing further sets estimation that in fifty years there will be smart-enough machine that human interrogator will not be able to guess correctly in more than 70% of cases after five-minute interrogation. From present it seems that many computers already passed this test as 30% threshold is quite low. We know from practical life, that there are many people who believe to spambots and even can be forced to do some self-destructive action' like sending money to these virtual identities.

What probably makes paper special in that time is next section with responses to anticipated objections. It is noteworthy to emphasize word anticipated as selection of these question shows us state of the time. And we will see, that most of them defined debate for almost hundred years now. There is nine of them and first is Theological Objection. We're set in Great Britain, 1950. What engineer selects as an objection to AI? Religion and Immortality. It is not question about possibility of thinking but about Right of Creation. We will investigate this obsession in Story of Religion and Story of Alchemy.

Next argument is about dangerousness of such machines and it prefigures Story of Securitization. Mathematical objection is mentioning the superiority of human over machine by defining it as such. Interrogator is asking machine logical paradox. Machine can't answer such question correctly. Such situation gives human feel of an advantage — not until time when he realizes that human is in the same situation.

Argument from consciousness is not only about consciousness but also about emotions. Turing quotes Jefferson Not until a machine can write a sonnet or compose a concerto because of thoughts and emotions felt, and not by the chance fall of symbols, could we agree that machine equals brain-that is, not only write it but know that it had written it. No mechanism could feel (and not merely artificially signal, an easy contrivance) pleasure at its successes, grief when its valves fuse, be warmed by flattery, be made miserable by its mistakes, be charmed by sex, be angry or depressed when it cannot get what it wants. Turing responds to this argument that machine can counterfeit such feelings and thus it is not important for the test. What is interesting here is opening another Pandora's box. Question of emotions and feelings quickly became one of the crucial ones. In fact in second half of twentieth century it is becoming the primal one in secular society. The machine now differs only in its abilities to understand and feel emotions. While in the time of Turing crucial trait was creativity now it is emotionality as creativity was deconstructed and killed by postmoderns. Emotionality now strives to be definition of human identity in western world. Where we see this topic most embraced is art, so I will elaborate more on this topic in Story of Art. Now I would stress only one example from 2015 — movie Ex Machina as it deals with Turing test directly. Humanoid AI is presented to interrogator and whole test is reduced to investigating if AI's emotions are real or simulated. It is quite illustrating how AI is being treated if it is stripped of its securitization aura. Of course there are still everlasting topics of responsibility of god-creator, primal sin and other christianic overload. But still — difference between human and machine is real emotionality here.

It would not be an AI text if it would not mention Lady Lovelace. Objection attributed to her is one about creativity. The Analytical Engine has no pretensions to originate anything. It can do whatever we know how to order it to perform (quotation by Turing, italics by Lovelace 1842). This is typical issue of that time. Humanity as resulted from Enlightenment project was defined by creativity. To be human means to produce new meanings. Turing response is critical here. He contradicts term of originality. There is nothing new without building on some foundation. If there is no independent idea from nothing then learning machine could produce same results. Turing is trying to beat the enlightenment beasts here. Creativity was definitely a topic for him as also other people noted ([Berrar et al., 2013, p. 249]).

The other arguments deal more with technicalities and are not so interesting for us. Argument from disabilities simply claims, that machine will have some flaws. Quick answer is, that every man can't do something. It is not about thinking, but about magnitude of thinking (or doing). Argument from continuity in nervous system says that human body is analogous compared to digital computer. As computer/human is allowed only to communicate in written way, this argument is for Turing irrelevant. Development of next decades obsoleted it completely. On one side analogous computers has risen and on the other side human body is no longer treated like analogous machine more like quantum digital space. Argument from informality of behaviour is reduction of learning problem — if computer can learn to read, then it can also learn to behave according to situation. Last argument is from extrasensory perception. Even such thing is Turing evaluating — telepathy, precognition and psychokinesis. He proposes to bracket telepathy out, as it infringes the test definition (communication only via written text). The rest of text is dedicated to description of learning machines.

I've already pointed few topics which have tremendous impact on whole AI discourse. Of course that Turing has not invented them, only distilled them from contemporary scientific discourse. But terminology and approach to AI (word was not used yet in 1950) determined AI as anthropomorphic problematic. It is not important that machine can think but that thinking is defined as an exclusive human activity. Turing is routinely using other anthropomorhic words as learning, playing or mental act when speaking about machine. I would dare to say that Turing test is so successful concept only because it treats AI as human and setting human as a measure to it. This definition by relation without investigating what runs under the hood of AI seems to be compatible with Habermas' communicative action which suggests possibility of its success. Every other definition of AI is complex and even incomprehensible, so this one (which is not in fact definition at all) is still successful. Here starts the mainstream discourse of human-like AI. Ironically, AI scientists sometimes understands text as in this example: We believe that Turing wanted to encourage us to abandon our anthropocentric view and to adopt a broader view on these phenomena. [Berrar et al., 2013, p. 249].

On one side Turing was thinking about machines as something what will merge with people or maybe even something what is identical to people, but on the other side he was not able to escape its humanistic origins. One of such quotes is here: It is natural that we should wish to permit every kind of engineering technique to be used in our machines […] Finally, we wish to exclude from the machines men born in the usual manner. [Berrar et al., 2013, p. 254]. This dichotomy between machine and man is hunting AI science until transhumanism is born in late sixties which finally transcends shackles of humanism.

Thin thread connecting text from 1950 with our present could be seen in AI discourse. IBM created Deep Blue in 1997 to play chess and beat mankind represented by Garry Kasparov. As Berrar quotes him and Martin he sometimes saw deep intelligence and creativity in the machine's moves […] The decisive game of the match was Game 2 […] we saw something that went beyond our wildest expectations […] The machine refused to move to a position that had a decisive short-term advantage — showing a very human sense of danger. [Berrar et al., 2013, p. 245]. Anthropomorphic approach is still present and legitimized by authority of science, especially by Turing which was put on pedestal decades after his death.

1.2.2 Evolution

Image wouldn't be complete if we don't look to how AI term's meaning is changing through the time. We've to found the actors (as there are probably some new according to Alexander's note: The exigencies of time and space create specific aesthetic demands; at some historical juncture, new social roles like director and producer emerge that specialize in this task of putting text into the scene [Alexander, 2004]) and see in which way they construct this term.

Intelligence is base term which are we working with. In public it is still something taken for granted, but we can note that even the word is starting to have contemporary meaning only in last few hundred years: Though it is most often assumed to be an ahistorical concept, the construction of one's intellectual capacity as being normal and another being abnormal only began to appear in the English language during the era of industrialization. [Nina Lester and Gabriel, 2014, p. 778]. Special term which allowed psychology to take expert position on this [Goodey and Dawson Books, 2011].

I'll not follow this path further and turn to artificial where robot is starting anthropomorhization point. It is interesting, that in Czech (from where robot word comes) is difference between general robot intended for any work and robot which is human-like. This is given by grammatical gender in which the world is used (there are only small differences e.g. in plural form roboty vs. roboti¹⁵) but it includes awareness that there are more types of robots. This is fading to grey, when it comes to intelligence.

One issue is with false implications which connection of artificial and intelligence allows. As Gozzi notes, construction of public AI legitimacy involved metaphors which were used to describe it. Most of them lead to anthropomorphization trap as they were used e. g. to compare computer to brain. Simple example show this false implication. Computer is brain with minor premise Thinking is computing leads to conclusion, that Sophisticated (AI) programs will give the computer a mind, with consciousness, intelligence and other human attributes. As Gozzi puts in his research, he sees diminishing usage of strong metaphors as AI term gets more and more established. It was no more needed as field was set up as legitimate and metaphors can move further, typically to Kurzweil's Singularity and transhumanism program. Another reason is that AI term can be broadened and re-thinked as something what is not limited by its human image. Greer for example thinks about intelligence whose subset is human intelligence. So now, AI is freed from human limits and is part of something bigger and shouldn't be treated as such.

Limits of AI can be seen here. As at first there was no need to compare human and AI status, there was simply AI aspiring to human level. From this discourse could be selected this (under)definition AI can be defined on the basis of the factor of a thinking human being and in terms of a rational behavior: (i) systems that think and act like a human being; (ii) systems that think and act rationally. [Čerka et al., 2015]. AI is here simply human-equal entity.

As transhumanist ideas appeared in the field, problem of comparison appeared as well. Transhumanist agenda tried to define AI as something better than human, which could be seen in terms of strong AI. The strong AI theory asserts that an AI is capable of having a mind, as well as mental states; versus the weak AI position, which suggests that a machine is only capable of behaving intelligently. [Laufer, 2013, p. 1]. Ironically this distinction come from the opponent's side as part of argument why it can't work. It provided name for new project where AI is condemned to transcend humanity.

Twenty years later this project splits to AGI (Artificial General Intelligence) which is more technical-oriented and seriously taken in AI science, while Superintelligence¹⁶ concept of N. Bostrom is some more general and tries to open hands for other implementation and criticism. Superintelligence tries to break stigma of artificiality and tries get beck to Lamarckian discourse of evolution. Bostrom is thus one of biggest proponents of naturality of such process. This discursively dissolves intelligence perception more generally as it is detached from natural persons. Superintelligence (with its paternalistic tone) also opens way to study things like Charismatic Robot Leader: The Superintelligence. It somehow removes stigma of Apocalyptic AI.

Term intelligence make software more tolerable than others. As some applications of software systems which nowadays could be hardly called as AI. Schwartz shows that using the world allows easier deployment of surveillance systems which otherwise will cause bigger social resistance. Their intelligence make them more human-compatible and less state-power tool. Similar examples are taken from places where these expert systems would replace some human workers.

Completely different chapter is link with science fiction. As Geraci points even in title Transhumanist Evangelism in Science Fiction and Popular Science, sci-fi is entangled with AI discourse. It on one side serves as generator of ideas and on the other it consumes scientific discourse and disseminates it in public. From Marxist, more especially Althusser, position we can see it as its tool of power. AIs are present in sci-fi almost from the beginning and shaped the discourse. Asimov's laws of robotics are influential until today and on the other side, concepts of friendly, apocalyptic, strong AI or superintelligence is there. Discourse is widely shaped by literature and movies and I'll go through it in Story of Art as they're not constituent part of scientific story, but still quite interconnected with it.

2 Story of Religion

Turing test as a crucial idea of AI discourse is reopening arena of positivist-religion fight. As I've pointed out first and foremost objection in that text is from religion. His denial of such argument is finished with denoting Copernicus heliocentric issue as similar to AI and its soul. This topic is incorporated in scientific narrative almost for whole its existence more or less visible and formulated. Existence of (non)human soul and consciousness is one of leading motives of decades after Turing. On one side it culminates inside so-called serious science by dispute Searle vs. Kurzweil about argument of Chinese room, which is at least by second side respected as an ontological dispute. On Searle's side I would say, that most of his objections are openly religiously deistic and contradictory.¹⁷ The result is that two sides completely discoursively miss each other and are not able of any meaningful communication in Habermas' sense.

To look into story of religion we have to go broader. As discourse contains a lot of references to other fields, we need to deal with concepts like Apocalyptic AI or Singularity trope. What connects all these issues with AI are transhumanist and in its extension posthumanist movements.

In first instance we will look to these movements via prism of Implicit Religion (IR). According to Bailey implicit religion can be described be three definitions. First is about Commitments which can be also unconscious. It is connected to free will and can emerge in the moment of crisis, Second definition is Integrating foci. There are moments in life which ties together all components of IR. Thus, this concept is permeating to all levels of identity and society in sense both of gesellschaft and gemeinschaft. Third is Intensive concerns with extensive effects, All three ideas bound together give us some framework how to image IR. It can be almost anything (Bailey's examples are IR of pub or suburbian life), so it shouldn't be merged together with terms like civil, invisible, folk or popular religion. These can be interpreted as special cases of IR, but it provides much broader framework. We can interpret AI movement from this unconscious IR interpretation of secular scientists to institutions of explicit religion (Turing Church, Mormon Transhumanist Association). Whole transhumanist and posthumanist movement can be interpreted via third definition. Living habits like reading sci-fi can be treated in integrating foci, paradigm, etc.

Posthumanism itself is fascinating example of enlightenment project impact. Definitive victory of science over religion and shackles of man. This creed is allowed by philosophical development which can be traced back to the enlightenment period and which culminates with conference series of Josiah Macy Foundation (1946–1953). Individualization of man is crowned by removing his special status in the universe and pursued by …a new theoretical model for biological, mechanical, and communicational processes that removed the human and Homo sapiens from any particularly privileged position in relation to matters, meaning, information, and cognition [Tirosh-Samuelson, 2012, p. 711] It is in consistency with what Foucault and Derrida describes as decline from man-centered universe in almost the same way as Latour notes The expression anthropomorphic considerably underestimates our humanity. [Latour and Porter, 1991, p. 137].

This pursue quickly escalates to removal of classical concept of man from the prevailing discourse of evolution. Evolution in transhumanism should continue in so-called protestant ethics of self-improvement which now contains also removing of body sanctity. Such improvements now consists also from self-modification. In its simplest case it is just taking some pills. In its harder-way by cyborgization and modification of one's DNA. Such man will technologically and genetically supersede the conventional man and will become Nietzsche's Übermensch finally.¹⁸

From here it is quite simple to remove concept of man at all and take one more step to posthumanism and condemn man forever. This inherently bad creation of evolution should be forgotten and substituted by evolutionary higher creature — post-human existence where human is obsoleted by postbiological and post-Darwinian stage of human development.

Substrate is morally irrelevant, assuming it doesn't affect functionality or consciousness. It doesn't matter, from a moral point of view, whether somebody runs on silicon or biological neurons (just as it doesn't matter whether you have dark or pale skin). On the same grounds, that we reject racism and speciesism, we should also reject carbon-chauvinism, or bioism. (N. Bostrom in [Kurzweil, 2005])

Kurzweil's bible of posthumanism and technological singularity ideology Singularity is Near: When Humans Transcends Biology is full of similar religious aspects as whole ideology. We will see that there is almost nothing new and we know all these concepts from history of Christianity. For now we just say with Geraci, that we can encounter new creature creation, unavoidable apocalypse, possible transcendence for chosen which allows surviving it, New Jerusalem of human patterns (where man is no more a creature but time-based intelligence pattern) uploaded in network encircling whole universe.

Tirosh-Samuelson tracks importance of cybernetics in transhumanism ideological base Whereas Heidegger saw it as the apotheosis of Cartesian humanism, Jean Pierre Dupuy has argued that cybernetics represented a crucial moment in its demystification and indeed the deconstruction of humanism. For Dupuy, cybernetics consisted of a decisive step in the rise of antihumanism (Dupuy 2011, 228), by which he means (following Hannah Arendt) the rebellion against human existence. [Tirosh-Samuelson, 2012, p. 712].

The Singularity is the ultimate point of evolution of man. There is nothing behind it what can be carved by biological man, the future behind this point of no return is unknowable. It is quite connected to Apocalyptic AI and as Geraci puts it: The singularity is the point on the graph of progress where explosive growth occurs in a blink of an eye; it is the end of history and the beginning of the new world and it is closer than you think. [Geraci, 2008, p. 149]. There is generally no link of old man with new world (and we will see this more closely in Story of Securitization), typical variant is Eventually, the machines will tire of caring for humanity and decide to spread throughout the universe in the interest of discovering all the secrets of the cosmos (though perhaps some will remain behind). [Geraci, 2008, Moravec in p. 151]

There are two threads whereas the technological posthumanists see themselves as a continuation and even intensification of the Enlightenment Project and glorify human reason and its ability to improve the world, philosophical/cultural posthumanists critique the Enlightenment Project because of its flawed metaphysics and harmful social consequences. [Tirosh-Samuelson, 2012, p. 715].

When we go to question when enlightenment projects ends and where religious path starts we will see almost clear division on present/future border. The former is straightforwardly secular, indeed a continuation of nineteenth-century humanistic naturalism and utilitarianism, while the latter is saturated with religious themes: its mentality is apocalyptic and its orientation is eschatological. [Tirosh-Samuelson, 2012, p. 716]. ¹⁹ I can take Bibel and his salvation ideology as typical example, where people has weaknesses like us selfishness, limited rationality, etc. To fix it we can either change human nature or compensate the weaknesses in an organizational and technical way based on Zusean thinking. […] This Zusean alternative means that we attempt to build an artificially intelligent agent which on the one hand matches human intelligence and on the other does not suffer from human weaknesses and failures. [Bibel, 2014, p. 89]. ²⁰ Finally he sees salvation and long-term guidance in AI What mankind would need is a super-human intelligence which on the basis of all available knowledge always would follow the rules of consistent, logical and rational thinking taking a global and far-sighted perspective. (Ibid, p. 100) going to such extreme that he proposes to design AI which will propose Kantian world's fundamental constitution. (Ibid). This unworthiness of mankind can be described via more engineering means as mankind should be replaced because its computational capacity is low and limited. Motivation why there should be any capacity is driven by belief in enthropy and opinion that high enthropy is boring.

It is sometimes difficult to set some topic to religious narrative or to science/materiality as it simultaneously hits both. Such case is relationship between body and religion. Concept of soul is inherently connected to body as there is no soul without body except heaven. As Christians strive for resurrection, it makes no sense for transhumanists who don't need body anymore. They are able to transcend even the body. Once again here steps into the game Teilhard de Chardin reintepreted by Ray Kurzweil as someone who can reconciliate Christianity with trans/posthumanism. AI raises once more question about which values can be learned by AI and which has to be loaded. Even here we meet ideas that religious belief (which means of course Christianity as almost any other religions are not interested in AI) should be preloaded [Mercer, 2015, Fernández Castro, 2014]. Even AI's existence can be taken as holy quest. In case of Mormonism it is very easily integrated and it is even becoming one of the ultimate goals of church²¹. In any case, classic religion is becoming interested in AI and devotes some of its resources to deal with the topic. It is not by chance, that same people are pointing to transhumanist belief that AI is inherently good with note If you model it after the human, you have to expect also human capacity for sin (paraphrased Mercer).

What will reveal us more about Christian roots of AI religion is when we look to others how they deal with possible conscious machine. First, less interesting, is Buddhism represented by Dalai Lama's response to such question as he can't totally rule out the possibility that, if all the external conditions and karmic action are there, a stream of consciousness might actually enter into a computer. (quoted in [Tamatea, 2010, p. 987]). Tamatea in his empirical research comes to conclusion, that real Buddhist response is in fact quite similar to Christian even counter to canon that says that God is equally present in everything. Notable difference is only in concept of Apocalyptic AI. This is not present in Buddhism as it has no eschatological version. It implies, that there is no fear from science as a tool to end of the world. What is also reflected here is that AI would be religious — it makes no sense to not be.

Japan with its philosophical origins in shintoism is even further than Buddhism and it seems that it have nothing to say to euroamerican discourse. Soul is part of everything thus discussion if something (AI) has soul is nonsensical. Border between living and artificial being is indistinct, easily crossed. Whole western obsession by question if AI is creature (ergo has intentionality, soul or consciousness) seems to be incomprehensible to this culture. One of most popular contemporary artists is Hatsune Miku — synthesizer generated voice with 3D projected body. For Japanese it is a creature for which they will use same word as for a person and it doesn't mean that they would succumb less to her charisma on concert than to western superstars.

Now if we've some starting point for Christianity, lets go back to it. Jewish account is more problematized by fact, that it is ethnicized religion. As such there is more than just teleological explanation. Science in itself is not in conflict with current state of religion and as Samuelson suggests, there is also in play genetic interest due to large prepondrance to inherited genetic issues. Such interest also modifies official stance of religious authorities. So, there is not anything what would stop Jewish religion from supporting science and transhumanism. The problem of AI consciousness is put to second line and is partially solved by positive Kabbalah stance. We've evidence even from seventeenth century of discussion whether golem can be included in minyan and similar discussions continues to twentieth solving whether artificial person can be granted the legal status of natural person.

For Sunna, Shia and Sufism is quite hard to find almost anything. There is anecdotic sunnic evidence, where Owaied proposes to study AI through Quran which says not much, but at least signals perceived compatibility of Quran with AI creation. Mahootian says, that Sufism is ready for transhumanism by definition. There is a simultaneity of disclosure and recognition in Suhrawardi's and ibn al-Arabi's ideology and this makes a basic epistemological point. From here is relatively open road to create AI in consistency with God. In discussion is if alam al-mithal could be somehow related to world of digital self and AI. Alam al-mithal's properties as place where soul builds the pattern of its resurrection body [Mahootian, 2012, p. 148] reminds disembodied AI and possibility of its instantiation. Sufism via Ismailism also perpetrated Shia and brought there such notions.

From here we can jump to Orthodox church as it is influenced by both previous. Clay studies seventh century's Maximos the Confessor. His (re)definition of orthodox teachings sets free will (thelēma) as part of human nature (phusis) with its own activity (energia). Human being is here a God's partner and in the same time concept of created and uncreated energies allows building of AI without problems. Compared to Buddhism and also Sufism (apocalyptic Islam not included) Christianity opens topic of eschatology. Russian orthodox contribution to AI/transhumanism is mostly via Nikolai Fyodorovich Fyodorov who was focused on this topic, immortality and resurrection of all dead as final project of Christianity. His ideas, while so wild, influenced not only later transhumanism but also project's like Tsiolkovskii's rocket science with goal to have space for the resurrected.

Western Christianity is most concerned with AI. I've already shown parallels which are taken to AI/transhumanism from Christianity. From the other side there is a lot of concerns. As Tamatea notes, christian reactions are often linked with fear of science related to eschatological explanations. Polarity between thing and being was also stressed on other place, but it is one of the leading topics in christian literature. Division between human and animal is concerned. This is not running only on level of thing-being, but also (and maybe more important, as human-thing is solved by defining it) human-animal level [Koosed, 2014, Garner, 2011]. Torn between body and soul is a long-term problem here and it is emphasized by cyborgizing bodies and theories of religion embodiment. In such moment christianity have to deal with moment when man becomes thing [Stannard, 2000, Cole-Turner, 2011].

Why is so much religion in AI? One answer is the paraphrased Latour — We've never been secular theory. Technoanimism (as we've already hit it in materiality section) is inherently present whole time and in postsecular society just more visible. As a primitive belief it could be present in IT as Aupers suggests. He meets Habermas here to say Science has no secular project. Even if everybody else moved to post-secular period, it is not the case of absolutist science.

Less aggressive position is from Habermas himself. He opens need for some transcendental belief as a result of modernity.

In moments like these [facing Holocaust], the unbelieving sons and daughters of modernity seem to believe that they owe more to one another, and need more for themselves, than what is accessible to them, in translation, of religious tradition — as if the semantic potential of the latter was still not exhausted. [Habermas, 2003, p. 111]

I can finally mix it together with Christian roots and say with Horkheimer Knowing there is No God, it nevertheless believes in him (in [Habermas, 2003, p. 113]). Eschatology as coming from classical Christian teaching and from the other side as fear from liquid modernity is being melted together in not completely new but rebuilt religion of AI with all its eschatological and transcendental promises. End of the world is on one side of discourse feared and on the other embraced [Geraci, 2007, McIntosh, 2010, Kurzweil, 2001, Geraci, 2008, Mehlman and Project Muse, 2012].

3 Story of Alchemy

Is related to story of religion. But it is more connected to hermetism and more limited to the moment of creation. As I've already noted in the introduction, roots of this movement can be traced thousand years back. Three main aspects are present. First is almost total anthropomorphization. Real goal of this branch is The Creature. It is not purely utilitarian approach from story of science. It is mystical attempt to recreate human. Model scenarios are striking with their detachment from reality as they are describing human behaviour not anything what could be expected from current state of AI science. …previous robot team leader has been killed, and several other robots have been either tortured or physically abused. In one case, one of the sentient robots is even used as a slave for another. Some of the robots are now imprisoned and are ritually abused according to their physical attributes or company of origin. as Ashrafian describes potential legal situation [Ashrafian, 2015]. Not to mention, that Neuman, Wiener or Minsky recognizes alchemist heritage by linkage with Rabbi Löw's golem.

Second is what Davis calls Techgnosis or Technomysticim. Assemblage of wild belief in technology, combined with sense of awe from mysteries hidden and being recreated there — it is everything what modern alchemy is. Starting with Newton and ending in today's cults of ghost-in-the-machine. This part is mostly interconnected with chapter about religion (and in fact would be best understood in terms of implicit religion), but I'll use it here as a glue.

Third aspect, directly implied by first, is ethics of the machine. We've already touched this problem in section about materiality. But there is still some maneuvering space even with limitation imposed by materiality. So let's investigate how scientists are thinking about this issue. Basic questions asked here are following. Which moral code should such creature have? What is the way how this code should be given or taught? Question of ethics is quickly moving from ontological level to epistemology. Another basic contradiction is emerging here: If ideal AI is condemned to human epistemology (inseparably connected with predesigned ethics), but in the same time has to outrun him on every thinkable level, which ideally means better (or at least different) epistemological capacity. From the point of ethics it seems inevitable to give AI free will. Kantian ethical system is one where this contradiction can be seen most easily.

Once this question [whether Kantian ethics permits the development of autonomous moral agents] is asked, it becomes clear that creating Kantian artificial moral agents is anti-Kantian. On one hand, Kantian moral machines would not be Kantian moral agents, strictly speaking. On the other hand, even if such machines were Kantian moral agents, their creation would nevertheless violate Kantian moral law. Because of this, the creation of Kantian AMAs is inconsistent with the prescriptions of Kantian morality. Since we (rightly) demand consistency in ethics, the failure of such machines to meet the standards of morality that they are designed to heed is unacceptable. We risk creating machines that may come to understand their very existence as being unethical. Moreover, we would be asking such machines to act in accordance with a moral code that we violated through the act of creating them. [Tonkens, 2009, p. 422]

From sociological perspective is interesting that even discussion about origins of morality doesn't reflect on its roots and is not even disturbed by that its own positions are not defensible in discussion or are even internally dogmatic. The only way to develop authentic Kantian moral agents would be to create AMAs that are free to the extent that they can sometimes choose to act immorally. This is most likely not a consequence that machine ethicists would be willing to accept, and rightly so. [Tonkens, 2009 , p. 431]²² Objection from context could be raised. I believe it is not in the place as there is a lot of other texts which builds on these non-reflected positions like what is right. In philosophical text trying to find best moral code is disturbing that it has a priori positions which are not taken to the mind. Of course, that these contradictory problem not hit only Kantian ethic, but almost in same way also Mill's or Bentham's versions of utilitarianism. There are some proposals how to get from the anthropomorphization trap in Latourian way of ethical code where things and humans are treated same or at least similarly [Floridi and Sanders, 2004, Coeckelbergh, 2009].

Whole alchemy line is one which brings anthropomorphization in most visible way. In many cases without reasons, in others there is underlying securitization story based on thesis: AI will not harm mankind only in case that it will be human-like.

Here comes partly mystical, partly already deconstructed inclination to humanization of machines. We see many attempts from old cultures to contemporary ones when people are trying to make machines look sentient. from moving statues in old Greece or Egypt to famous von Kemplen's chess machine with hidden player. Pre-AI software is working in less obvious way but still it was deception and no real sentience (e. g. ELIZA). Sharkey goes through this history and shows us how cultural myth of robotics was established. Via Alexander's strong program of cultural history he shows, how robots where established as cultural property shortly after R. U. R. in 1920 was released. Machines which were mechanical boxes until that time has started to acquire human-like visage. In 1927 Westinghouse produced simple machine — in fact small phone central. The Televox was from marketing reasons turned via so-called Čapek's effect to mechanical man. Two boxes which were enhanced by hands and legs via newspaper's comics. It started new mode wave Anything that had a sensor was called a robot. This is short example how quickly are people able to attribute human traits to machine. There are psychological experiments which shows us that people create attachments to computer and even more to androids. There is a lot of therapeutic robots which are exploiting exactly this feature.

Interesting effect of business applications can be viewed from gender studies prism. There is running discussion about gendering AIs and robots. One stance is that they shouldn't be gendered at all (voice of AI should be genderless) to not reify stereotypes. On the other side is business push which wants to use psychological effect of anthropomorphization and in its extent — gendering. People better respond to genderized machines, they behave differently if wo/man like robot is present. Of course to create robot with correct sex is easiest done by emphasizing gender stereotypes. Some effects of this are clearly visible, while other could have some impact in the future [Adam, 1998, Marchetti-Bowick, 2009, Hrişcă, 2012, Larson, 2010, Shaw-Garlock, 2014].

Where this part has the most harming effects is definitely military today ²³ — we see that this issue matters: US army colonel who was in charge of a test of one of Tilden's legged robots in a minefield. The robot, modelled on a stick insect, successfully detonated a number of mines and lost a leg each time. It was dragging along very successfully when the distressed Colonel in charge stopped the demonstration because he thought that the test was inhumane. [Sharkey and Sharkey, 2006 , p. 13]. We can imagine some impacts of this. When we are nearing to AI, human response is slowly changing from idea that robot is alive to idea that AI has mind and is conscious. Sharkey concludes that it is almost inevitable that AI designers also have to fall into the trap and they'll treat their products as conscious even if it is just a simple box with two wires.

There are few hypotheses, why even in these secular and post-secular times are scientists so susceptible to hermeticism and generally mysticism of technology. Let's start with Long before Weber wrote about disenchantment, the world was already disenchanted by monotheism [Coeckelbergh, 2013] Secularization thesis is extended by Coeckelbergh to this extreme. What killed enchantment is one God, not removal of God. Christian God defined (of course in its ideal variant) the material world as world without spirit. This is probably the first time in history when things are really dead. When this line is breached, monster appears — Frankenstein. There is some possibility of aura bestowed by author to its creation, but this aura is being lost in modern age with mass production. This idea is starting with Marx and even more explicitly with Benjamin. Information technology emphasizes this problem as artefact is also lost. Information is not artefact and it is not human also but it is pervading everything. Its status is threat to ontological concept of human identity. Latourian way of returning things into play could be seen as re-enchantment of the world. His removal of dichotomy between works of purification and works of translation is trying to retie the world back together. It is also expressing some sentiments but in new post-modern way. In less scientific interpretation it simply means putting spirit back to things which is exactly what AI discourse does. It is no more transcendent spirituality but immanent one as Coeckelbergh correctly notes. This immanent spirituality is something what I would place in this alchemy line without hesitation. There is also short way to fear of such creations as they're more created in imago humani [Herzfeld, 2002, Herzfeld, 2015].

Part of alchemy is creation of superhuman. We can trace thin thread from Lamarckian evolution to eugenics (which is in its modern form inherently connected with trans and posthumanism). Teilhard de Chardin is leading figure of philosophy of AI and especially appreciated by singularity theorists. His vision of Noosphere and Omega Point were almost without changes, just with removal of poetic balast and transformation to science-like narrative. All-encompassing purpose of mankind was taken from this Jesuit and rebranded to something without inner consistency previously provided by Christianity. Language of Kurzweil and Bostrom is built on destiny of intelligence. From Noosphere is conscious universe and from Omega Point — Singularity. Intelligence is a pattern of life which has to expand as Lamarck's heritage. Tendency is taken for reason. What Fuller shows, is mix of non-reflected Christian origins with blind science shifts focus in last decades from simulating to instantiating of life and he explains it: The language of instantiation derives from theological discourses of the Christian deity's triune nature, i. e. the idea that God is subject to three equally divine manifestations: Father, Son and Holy Spirit. These theological roots go beyond historical curiosity to a general principle of Biblical interpretation that provides a precedent for reducing, if not erasing, the differences between processes, entities and interventions of artificial and natural origin. [Fuller, 2009, p. 14] Science finally playing God — this paradigm switch has direct impact to society also. Fuller argues with U.S. circuit court case in which this changed concept of instantiation was used as argument against creationist defendants. Judge simply took for granted that simulation of primitive life in computer is no more simulation but life itself. This example can from some point be seen as a fair fight of two religious paradigms. Such interpretation would be fair via Habermas' notion of communicative action, but it is not perceived as such in science with its ultimate right on objective truth. Fuller is pointing to probable self-fulfilling prophecy character of such shifts and its reification by legitimacy of court authority. There is a potential that concept of life will be redefined in future based on this paradigm shift. There are already some attempts to this — what is unpredictable is alive. In Wittgensteinian interpretation it means, that we've nothing more to talk about. Problem noted in this text will simply disappear in language which will not involve term of life (when everything is alive aka in Kurzweil's and de Chardin's conscious universe).

4 Story of Securitization

Securitization is term borrowed from International Relations created by so-called Copenhagen school. It means politicization of some topic via its relation to security issues. Important is that it is talking about security as about speech acts and in our case it is view which allows us to see how (potential, perceived or imagined) AI security issues influence real-world and science progress.

This story connects all previous. Concerns from AI and science in general is leitmotif here. AI as a grave danger is framing of almost any public discussion of the topic. It does not depend if it is a religious counterargument or simple positivist calculation how much time is needed for uncontrolled self-replication nanobot to destroy whole planet.

Fear of being destroyed by own progeny once again follows old myths represented e.g. Zeus' revolt against Titans. In relation to AI will be more interesting modern incarnation in Čapek's novel R.U.R. or in the later world-known novel Do Androids Dream of Electric Sheep?²⁴. These myths are forged to concrete fears which are based on specie or at least group rivalry. If AI will be created what is the probability that it will be human-compatible? What will stop AI from enslaving mankind? Will AI have any cognitive capacity to even perceive humans? Whole this category of questions is leading us to two sides of quarrel — first is abolitionist movement which wants to stop everything before self-improving AI will be unstoppable. On the other side is an effort to make AI compatible from design and in second row to build ethical-legal frame for coexistence.

Especially question of legal framing is quite interesting. Technicist branch is still rooted in Asimov's three laws of robotics [Asimov, 1977, Khalil, 1993] and is set in defensive mode. Compared to it humanist line is based on human/AI equality thesis and in its weaker mode on animal rights argumentation. This line could be also split to part which opposes Anglo-Saxon legal order together with Christian tradition where man is qualitatively different and such order is formed to his defense. Main objection is that if animal rights were already admitted as substantially independent (so not as defence of their owner's rights) the taboo is already broken and there is no reason why to not extend these rights to AIs. Second line is rooted in discourse promoted by P. Singer. In connection with human rights the new interesting mix is born. Some rights are confirmed, while not every one of them. Field is quite large and has origins in anthropocentrism, biocentrism or ecocentrism with all their historical baggage. Whole discourse almost logically expands to bigger issues like when man is still man and when AI is ready for its rights. This is the point where sociologically relevant authors enters the field: Barthes, Baudrillard or Husserl with human identity as a social construct, etc. [Hrişcă, 2012]

This partly scientific and partly public discourse is further propagated to legal norm proposals. Their genesis and construction are also interesting. One of the origins is already mentioned Asimov (note, that this is the art input) and second quite influential is historical experience with legal norms for genetic manipulations and virus treatment. Proposals which are dedicated to transformation to laws are brought to live [Veruggio, 2007, Ashrafian, 2014, Field, 2012]. They are still not real laws, but media works with them as witch such. It is more connected to discursive construction than to securitization. Practical norm which at least someone adheres to are Principles of Robotics or EU proposals.

Beside these proposals which are more utopia than reality what we hit is problematic of software (and robot) responsibility. First trials are being run, where judges tries to answer question about who — if its operator, manufacturer or drone itself — is responsible for drone's actions. Part of the field is completely serious discussion about how AI should be persecuted if found guilty. Proposals for limiting some of their rights, jury's right to turn it off (as a euphemism to capital punishment) puts a mirror to how these rights, ethical and democratic foundations are perceived in concrete communities. Habermas theory of communicative ethics can enlighten some aspects. For example even if AIs are in this discourse treated as completely human-equal they are not permitted to anyhow participate on legal system. Only some extremist positions hold on that AI will be part of a democratic process. Question of political rights of AIs is for most of the actors completely unacceptable and this non-acceptance is in direct contradiction with their egalitarian position.

AI as a danger (and here I omit other variants as an ecological threat, etc.) is main argumentation line of AI antagonists. Ethical or religious issues takes place behind this flagship.

Nor is there, to be sure, any lack of wild speculation. A handful of freaked-out intellectuals is busy reading the tea leaves of a naturalistic version of posthumanism, only to give, at what they suppose to be a time-wall, one more spin - hypermodernity against hypermorality - to the all-too-familiar motives of a very German ideology.

The big problem is once again anthropomorphization. Let's say with Sharkey, that Like other cultural myths, it can be harmless in casual conversations in the lab. But it is a perilous road to follow in legal and political discussions about enabling machines to apply lethal force. [Sharkey, 2012, p. 791] I've already shown some cases where a military equipment was treated more like fellow comrades. Add another: The Washington Post reported that soldiers on the battlefield using bomb disposal robots often treat them as fellow warriors and are sometimes prepared to risk their own lives to save them. They even take them fishing during leisure time and get them to hold a fishing rod in their gripper. [Sharkey, 2012, p. 792] or human-like discourse soaking to highest ranks of U. S. military Gordon Johnson, former head of the Joint Forces Command at the Pentagon, told the New York Times that robots don't get hungry. They're not afraid. They don't forget their orders. They don't care if the guy next to them has just been shot. (Ibid). Robot is taken for soldier here and treated as such. Even if we admit that there will be human-like AI, we're not there yet and these soldiers are simple machines with no free will. I would quote Sharkey's once again as he brightly summarizes the problem which emerges here:

Direct result of such subversion, is responsibility slowly virtually shifting to machines. Man who is driving drone feels no more full responsibility for its actions, as drone is semi-autonomous. This quality of anthropomorphization effect is on the second side exploited in business and in therapy. Human-like robots are better sold [Robertson, 2007, Larson, 2010] or they can be used as companions for e. g. elderly people [Hakli et al., 2014, Kerstin, 2014, Kernaghan, 2014, Pfeifer et al., 2012]. The question of trust to AIs is repainted as commitment. This targeted use makes border between what we pretend and what we believe even more fluid.

Generally there is a technicist discussion if and how is AI able to destroy mankind. Second part of it is AI will have intention to do it. The concern is that the machines will view us as an unpredictable and dangerous species. [Panda et al., 2015, p. 111] shows typical argumentation. On the other side transhumanists and scientists are trying to oppose such fears [Kurzweil, 2001]. Compared to direct destruction there is an eternal idea of society decay in the realm where everything is provided.

Part of securitization discussion is also topic of AI as a citizen. We've already covered it in other chapters, so just adjusting some topics how they are reflected from this point. At first capability of AI to be a citizen is not coupled with its human status. There are tendencies to define new criteria and redefine traditional Turing test to citizenship test. Such activities can cast some light on how this duality humanity-citizenship is treated in today's society. Here is sometimes seen tendency to evade such problem by …best not to make AIs extremely human-like in appearance, to avoid erroneous attributions that may blur the bright lines we set around moral categories (Arneson 1999). If such confusion were to develop, given the strong human tendency to anthropomorphize, we might encounter rising social pressure to give robots civil and political rights, as an extrapolation of the universal consistency that has proven so central to ameliorating the human condition. [Yampolskiy and Fox, 2013, p. 224] If scientist are not willing to give AIs civil rights, they are not against lending them at least some non-harming rights like intellectual property rights. Of course, there is still discussion running if it is not possible to grant them more rights (with hidden goal to provide them also political rights, but somehow limited).

5 Story of Art

Art is quite disconnected from discussion of scientific and public circles. It is just an art. Nevertheless, it is exactly this sphere and especially sci-fi what builds public image and also motivates directions of scientific research. Topic of scientific AI is facing whole post-descartian epoch. From mentioned Frankenstein, R. U. R. through whole storm of books and movies from fifties and sixties. Exactly from the time when AI seemed to be almost here. Cold war also had its impact here. Partial deprivation during AI winter²⁵ was fully replaced after its end on the dawn of singularitarianism. Not only literature but also big Hollywood theme with flagship A. I. by Steven Spielberg. Topic changing reflects contemporary state of science in some proportions. On the other side art is involved in topics which are not so changed from twenties, so almost a hundred years of possibility of emotionality, creativity, living together and what humanity means in such context [Simon, 2013, Van Den Eede, 2015]. Classical leitmotif is AI more human than humans. Practically every story from previous chapters is here, religious utopias, Turing test and hesitations about salvation.

As artistic forms are part of another field (while I was mainly focusing to scientific and public discourse) and plays by different rules [Bourdieu, 1996] I will only point out few of specific aspects where science and religion meets art. Science fiction (SF) novels frequently express religious themes and do the work of traditional religion; as a result, they are actors within the religious landscape. Although in many ways science fiction can oppose traditional religion, which often appears in SF as dogmatic, obscurantist, or even evil, the genre explores religious themes, and even gives them pride of place. [Geraci, 2014, p. 418] I would add to this, that not only religious actors but also scientific ones. Nevertheless, their religious aspect in light of preceding chapters is most relevant here.

Apocalyptic AI is typical example of what is running through literary field. This concept is here from the beginning. Frankenstein contains it in its primitive form as one destructive being. Twentieth century sees AIs operating on much larger scale of destruction especially with dawn of second world war. R. U. R. is running the expansion — humanity is destroyed by AIs which shows to be at least same human as their creators. Trend is set and in epos Dune written by two generations (1965–2007) of Herberts, AI is the ultimate enemy of mankind. Fear of it is presented three years later (1968) in Kubrick's blockbuster 2001: A Space Odyssey. 1978 gives a birth to Battlestar Galactica (BSG) which starts at the moment when AI destroyed the mankind. 1990 story reemerges in Simmons' Cantos Hyperion — mankind is once more confronted with its eternal enemy. 1999 brings The Matrix trilogy 1999–2003 where revolt against AI-imposed slavery ends also with cooperation of AIs and humans [Díaz-Diocaretz and Herbrechter, 2006]. When it seems, that it is closed forever, war on terror resurrects Battlestar Galactica in 2004 and movies like Transcendence where AI threat reappears. Story is mostly the same for whole hundred years — AI is an enemy but has some potentiality to merge with mankind (typically by desire for emotion and creativity) and transcend its flaws. Practically every piece mentioned here ends with this message more or less perpetrated by Christian topics. Starting with R. U. R. where final episode of Adam and Eve is explicit same as in BSG, which is based on Mormonism and whole episodes are devoted to problems of One God and return of Christ through Cantos Hyperion dealing at least with Abraham's sacrifice dilemma. Theory that AI threat topic is correlated with moments of ontological or physical insecurity of western society would be worth of some serious research.

As typical Christian example is taken Terminator movie (and especially TV) series which is overloaded by Christian symbolic. Bennet points to mankind saviour John Connor (aka Jesus Christ) through terminators as AI angels and Sarah — the Virgin Mary. This concept is not unconscious, as authors reflects it: …actor Thomas Dekker (John Connor) asks showrunner Josh Friedman why there is so much Jesus in the show and the latter replies, The Terminator is a very religious series. […] I liked the idea of exploring it in a non-obnoxious way [Bennett, 2014]

Another repeated topic is relation of human to AI — what is the difference and who is more human. Frankenstein already hit the topic in similar way as R. U. R.. Classical piece is Dick's Do Androids Dream of Electric Sheep? 1968 where variant of Turing test is used to determine if one is an android and as such should be retired. Main character is trapped in dilemma of murder/retirement. Book was later remade by Ridley Scott to similarly famous movie Blade Runner 1982. Two years later in 1984 William Gibson starts new sc-fi movement cyberpunk via novel Neuromancer where AIs are more independent and thrive for their own independence and goals. Religious topics are made explicit in rest of the trilogy (Count Zero 1984 and Mona Lisa Overdrive 1988) where AIs take the role of network demigods. Topic is revisited in movies A. I. (analyzed e. g. in [Chira, 2014]) or Ex Machina 2015. There is no theme of merging with humans, but more about co-existence. Symbiotic variant is also present in Iain M. Banks Culture series which also lead to some proposals about how AI can sustainably impact society.

Transhumanism strikes the sci-fi explicitly with Cyberpunk 2.0 movement. Notable works like Stross' Accelerando in 2005 or Doctorow's Down and Out in the Magic Kingdom are taking AI, Singularity and transhumanist future for granted in Kurzweil's vision of it. Functions as successful disseminators of positive image of transhumanism and as tools for overcoming AI fears].

As computer engineers are quite tied to these topics and transhumanism at all, it is not a surprise, that with advent of computer games there is a prevalence of AI-related topics. System Shock in 1994 introduces typical villain AI, which reappears in small intervals in Deus Ex 2000 or Mass Effect 2007 which typically don't impose new ideas and reiterate themes previously opened in literature.

Whole art train is passing the scientific and transhumanism ideas to more general public and on the other side reifies legitimacy and driving force on side of scientists [Geraci, 2014, Geraci, 2007, Geraci, 2011]. I've already mentioned this chapter is simplified as there is much more material which can (and should be) dealt with.

3 Conclusion

AI anthropomorphization theme has shown as quite wide. It is permeating AI science itself but other — on first sight unrelated — fields like ethics, religion or law. All these independent and rational areas seems to be more or less affected by unconscious choices given the agency of things. It is nothing new, but worth to study on concrete examples. Difficulty of holistic dealing with the topic is on the other side seen also in the structure of work. It is not easy to talk about AI religion without touching its alchemistic side or looking to limits of programming. What can be taken as a lesson here is that potential empiric research should take in game all these effects. Otherwise, it may will describe some facets of the topic, but probably it will create distorted image. AI is not clear hard science. AI is human project as most others and is influenced by and influences living world also in purely social ways.

Originally I have selected five narratives from whose only two are now present in text. I have dropped religious, alchemist and art stories due to lack of space. I have also investigated them and these chapters are available from me on demand. The selected ones are still broad enough to show different aspects which can be studied from sociological point of view.

Story of Science has shown, that AI is not deliberate concept which is valid in any rational space. It was constructed especially in western civilization during last few hundred years and it has shown that in other intellectual and socio-political situation such concept need not exist at all.

Story of Securitization on the other side shows that when such concept exists it can be on one side used as a political weapon nevertheless what real impact is. Or on the other side in combination with anthropomorphization it leads to weird human behaviour and laws which merely reflects the objective reality. Agency of things itself can modify how judge will decide its case as we have seen.

When we would dive to religious stories, we would have seen that AI science field is driven by western Christianity and even non-consciously embracing eschatological discourse. This is something which on one side has big impact to securitization story and on the other side warns us a lot. As AI scientists would never say they are religious believers in AI, they behave like their followers. We can think about us as sociologists if we don't behave in same way with some other deity.

Also many questions were opened to be tested by rigorous hypothesis, some of them can be verified by experiments, some by deeper field research. Anthropomorphic AI can seduce us to thinking about it as only a small fragment of sociology of science. But from other stories we see, that it is not only sociology of science, but also sociology of religion, sociology of things, … Sociology is also not the only tool which can be used to understand it — anthropology can give us a helpful hand with its tradition of observation. What we also need is independent point of view which can reflect on role of possible sociology of sociology of AI (in same meaning as meta-sociology of science or so-called second-order science). So let's explore the borders of constructed science.

Bibliography

Footnotes