1.1 Introduction
When considering prognoses given about the potential of certain technologies and technological trends, we most often remember the outrageous misjudgments rather than the precise assessments and are most likely witnessing similar false predictions these days—unbeknownst of their falsity. From IBM’s president Thomas Watson stating that the world will not need more than five computers in 1943 to promises about the various announcements of the immediate advent of self-driving cars on our streets within a few years, the list of wrong predictions is long.
Considering the complexity of reasons for certain technologies to become ubiquitous elements of many people’s everyday life, one would be inclined to refrain from those prognoses altogether. Yet, those prognoses themselves may function as self-fulfilling prophecies, by inspiring the public to think of a technology a certain way, thereby opening or closing minds and markets for certain devices.
Additionally to these self-fulfilling prophecies, unexpected breakthroughs in technological development, applicability, and compatibility, societal trends and ethical restrictions, economic (in-)stability and investments, politically forced acceleration or deceleration, successful marketing campaigns, or simply luck all take their fair share in the rise and fall of technological standards, applications, and devices.
For some individuals, new technologies represent hopeful progress toward a better future. For others, the very same technologies are viewed as threats to the way of life they are accustomed to and intend to keep unchanged. No matter whether one views technological changes as net positives or negatives, the common denominator seems to be that technology facilitates change.
However, not all types of change are facilitated by technology. Technological progress most often results in social change. It changes the way we relate to our environment, to each other, and often to ourselves. New communication devices allow for constant interactions with people thousands of miles away, while augmented reality will add another layer of interaction and information to our immediate surroundings. Technologically assisted medical progress allows for curing diseases that just decades ago were death sentences, while the latest autonomous battle drone can strike without being noticed by its target. Some social movements, like the Arab Spring, would not have been possible without ubiquitous access to social media. However, this access also allows oppressive regimes an even more oppressive grip on its population, as exemplified in the Chinese social scores. Technological progress may result in social change, but it does not guarantee social progress.
New technologies come with risks associated with their use, both individual and collective risks. In open societies, discourses about the acceptability of those risks ideally determine the overall acceptance of such technology. Artificial intelligence has so far been an elusive technology when it comes to its thorough risk-assessment and social response. Partly due to a certain AI illiteracy of the general public, leading to broken discourses about what AI can do, and partly due to the speed of its development, especially of the last decade, a coherent risk-assessment has been missing. This speed, often likened to a technological revolution, has also opened many philosophical questions that are just now slowly being asked and subsequently answered. Some of those questions will be asked here, and hopefully some answers will be provided.
1.1.1 Smart Fridges and Other Reifications
For a philosophical inquiry into issues of AI, one main obstacle appears right at the beginning: how can and should we understand the otherwise opaque concept of artificial intelligence? Due to the vast range of methods and applications that all are claimed to incorporate and exhibit some form of intelligent behavior, an all-encompassing definition will lose any practical purpose to limit any inquiry.
Take as an example: a “smart fridge.” Its intelligence-claim is based on the ability to scan items in one’s fridge and preorder those that, according to typical use based on someone’s consumption profile, will be used up soon, or warn about expired articles in the fridge. Calling the fridge smart, then, is a reification of AI, as it is not the fridge as a whole, but the added software and its connection to the cloud that is providing the smart function.
We are used to identifying intelligent beings as embodied entities occurring in nature, and this phenomenological basis is often the cause for misattributions and confusion about the source of (artificial) intelligence. This observation suggests that we require a fundamentally different approach to artificial intelligence than to natural intelligence: AI may come disembodied or might be re-embodied, duplicated, changed, and adjusted to the tasks at hand. It never is just one intelligent artifact, but an algorithm capable of operating on other hardware. To some degree, this argument also applies to approaches in philosophy of AI that concentrate on robots as embodied forms of artificially intelligent agents. Researchers have long argued that embodiment is a prerequisite for many cognitive capacities (Duffy and Joue 2000; Stoytchev 2009). However, this does not mean that robots are to be considered the intelligent entity, but that they operate with an intelligent algorithm. We should recommend, then, that philosophers carefully define the object and scope of their inquiries to avoid reification.
In this book, this object will be artificial speakers and their social impact. Artificial speakers are understood as computer programs capable of analyzing and reproducing natural language, that is the language human beings use to communicate with each other. Most of those speakers are not embodied, i.e., they do not appear with a physical presence, even though their application is certainly not limited to chatrooms or being personal assistants in mobile phones and at home.
The reason for seeking out artificial speakers from all current uses of certain types of AI is the capacity to speak with humans in their own language. This simple fact differentiates this technology from every other AI so far. No other technology produced by humankind so far has managed to enter stable, interactive communication based on people’s own language.
Engineers working in the area of natural-language processing (NLP), tasked with improving the skills of those artificial speakers, have no other way of proceeding than to imitate human language use, which ultimately results in deceptive copies of speaking robots that not only use human language but imitate human speakers. The better engineers follow this task, the more dubious their product becomes. With the incoming products of those engineering efforts, many of the topics discussed here are also being discussed under the term “human–machine communication” (HMC) or in media studies. In fact, many of the approaches of the social sciences take the development of humanoid robots as a starting point for their research (for example, Zhao 2006).
Take Google’s Duplex, advertised as a virtual assistant capable of seamlessly infiltrating human conversational practices by simulating human-specific features, like thinking noises and interjections (Leviathan and Matias 2018). Investigating the impact of such a humanoid robot on the social relationships between humans and other humans, but also between humans and those machines has become a central point for HMC (Guzman 2018, 16).
Yet, many of those analyses are approaching these artificial speakers and social robots from a media- or communication-science background. Reflecting upon those processes from a philosophical perspective, then, is needed to both provide tools to describe and to assess these social robots and their relationships with us. A philosophical approach to the way we interact and communicate with, rely on, and relate to these speaking machines allows for a normative approach not only to the way those machines are constructed, but also to our attitude toward the possibilities of building human–machine relationships.
Many children treat their plastic pets with the same care and empathy they would treat a living one. It seems that there are ways of relating to machines in ways unknown, unfamiliar, and possibly uncomfortable to us due to preconceived notions not only of what technology can do, but also of what relationships should entail. A bigger picture is needed to answer the questions of future human–machine relationships. It is important to keep in mind that artificial speakers are designed entities and thereby can take forms that we, as the designing community, ideally consent on democratically.
The core diagnosis of this book is that we do not have this bigger picture available yet, and constructing one is the task of philosophers of technology in the twenty-first century. This coming century will undoubtedly bring new ways of human beings relating to their technological surroundings, and one of these ways is to build social relationships with them. One element of this bigger picture, then, is the idea that our social categories with which we describe elements of the social fabric are woefully lacking differentiation. This lack of differentiation is the reason why some people rejected artificial speakers as any possibly relatable technology, similar to people who rejected the idea of relating to toy pets.
The limitations of social categories are driving the engineering goals to create more humanoid robots, fueling the fears of people resulting from this successful engineering, and limiting the imagination of human–machine relationships that are beneficial for everyone involved. This is accompanied by questionable presuppositions of what typical human features are, how they ought to be reproduced, and how those reproduced features form our image of exhibiting those features in real life.
The proposal for the bigger picture needed here, then, consists in offering alternatives to anthropomorphism and avoids several different problematic developments, justifying the program of this book.
1.1.2 What’s to Come?
For this program to work, some preliminary clarifications ought to be made. First, some methodological points are in order. These are presumably not necessary for the philosophically educated reader, but possibly quite useful for readers from other disciplines and backgrounds. The difference between an analysis and a reconstruction, for example, will carry some of the weight of this project, and it would be helpful for readers to follow this point. Second, we require a better idea of how artificial speakers work, what the main objective in creating them currently is, and why the assumption is justified that they will only increase in conversational sophistication. This chapter, in turn, may be of more interest to those of a less technical background. By pointing out that the current standard of programming AI is machine learning, the high hopes or concerns of soon arriving at a general artificial intelligence can be recalibrated. Most of artificial intelligence today consists in the convincing simulation of behavior and actions. However, as with any philosophy of technology, some speculation and extrapolation are required. Third, a relational approach is developed to lay the groundwork of understanding social human–human relationships. This relational approach includes human–pet relationships as a precedent for incorporating non-human agents into the social fabric by assigning them a unique social category. It also includes purely online-based human–human relationships as evidence that physical proximity is no longer a requirement for meaningful relationships. Fourth, the transfer of this relational approach to human–machine relationships is presented. This transfer should be considered the core chapter as it presents the arguments to incorporate artificial speakers into the social fabric by establishing a new social category, akin to a second domestication. Finally, the fifth chapter faces the consequences of such a move, by acknowledging that this position requires a stance on the debate on robot rights, human-based design, and the human–human consequences of emerging human–machine relationships.
References
- Duffy, Brian, and Gina Joue. 2000. Intelligent Robots: The Question of Embodiment. http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.59.6703. Accessed February 11, 2020.
- Guzman, Andrea L. (ed.). 2018. Human-Machine Communication. Rethinking Communication, Technology, and Ourselves. New York: Peter Lang.
- Leviathan, Yaniv, and Yossi Matias. 2018. Google Duplex: An AI System for Accomplishing Real-World Tasks Over the Phone. Google AI blog. https://ai.googleblog.com/2018/05/duplex-ai-system-for-natural-conversation.html. Accessed February 11, 2020.
- Stoytchev, Alexander. 2009. Some Basic Principles of Developmental Robotics. IEEE Transactions on Autonomous Mental Development 1 (2): 1–9.Crossref
- Zhao, Shanyang. 2006. Humanoid Social Robots as a Medium of Communication. New Media And Society 8 (3): 401–419. https://doi.org/10.1177/1461444806061951. Accessed June 6, 2020.