The Routledge Social Science Handbook of AI

Anthony Elliott, Anthony Elliott

• Publisher: Routledge
• Year: 2021
• ISBN:
  9780429582066
• Edition: 1
• Topic: Ciencias sociales
• Subtopic: Sociología

The Routledge Social Science Handbook of AI is a landmark volume providing students and teachers with a comprehensive and accessible guide to the major topics and trends of research in the social sciences of artificial intelligence (AI), as well as surveying how the digital revolution – from supercomputers and social media to advanced automation and robotics – is transforming society, culture, politics and economy.

The Handbook provides representative coverage of the full range of social science engagements with the AI revolution, from employment and jobs to education and new digital skills to automated technologies of military warfare and the future of ethics. The reference work is introduced by editor Anthony Elliott, who addresses the question of relationship of social sciences to artificial intelligence, and who surveys various convergences and divergences between contemporary social theory and the digital revolution.

The Handbook is exceptionally wide-ranging in span, covering topics all the way from AI technologies in everyday life to single-purpose robots throughout home and work life, and from the mainstreaming of human-machine interfaces to the latest advances in AI, such as the ability to mimic (and improve on) many aspects of human brain function.

A unique integration of social science on the one hand and new technologies of artificial intelligence on the other, this Handbook offers readers new ways of understanding the rise of AI and its associated global transformations. Written in a clear and direct style, the Handbook will appeal to a wide undergraduate audience.

Part I

Social science approaches to artificial intelligence

1
The complex systems of AI

Recent trajectories of social theory

Anthony Elliott

One of the principal aims of The Routledge Social Science Handbook of Artificial Intelligence is to provide students and teachers with a comprehensive and accessible guide to the major topics and trends of research in the social sciences of artificial intelligence (AI), as well as to survey how the digital revolution – from supercomputers and social media to advanced automation and robotics – is transforming society, culture, politics and economy. A unique integration of social science on the one hand and new technologies of artificial intelligence on the other, this handbook offers readers new ways of understanding the rise of AI and its associated global transformations. Another aim of the Handbook is to address the very wide array of phenomenon associated with the digital revolution, providing the most up-to-date coverage of developments in AI, machine learning (ML), robotics and supercomputing. Topics addressed where AI currently transforms or, in the future, promises to transform social, economic, cultural and political processes include:

• AI within discrete apps, embedded within operating systems, and operating systems based on AI.
• Single-purpose robots throughout home and work life.
• Bigger, faster, superdata analytics: ‘Colossal data’ (bigger big data) which will necessarily involve new data curation and analysis approaches that enable more patterns from an ever diversifying range of data.
• Low-power computational hardware, including neuromorphic computers that are more suited to some applications of AI than traditional computers.
• Miniaturized quantum encryption devices, which will underpin the security and trust that will be required before new technologies are widely applied. This particularly applies to applications with high-consequence failure modes (such as implants with direct access to the brain).
• Advances in ML.
• Advances in battery technology: enables stand-alone and mobile intelligence in a wide range of applications.
• Advances in machine cognition systems
• Mainstreaming of human–machine interfaces. This would enable a host of new applications, the easiest to imagine being those using new brain-machine interfaces.
• Massively parallel computational architectures and quantum computing.
• Advances in generalized robotics, such as multipurpose labourer robots.
• Advances in AI, such as the ability to mimic (and improve on) many aspects of human brain function.

The Handbook provides representative coverage of the full range of social science engagements within the AI revolution, from employment and jobs to education and new digital skills to automated technologies of military warfare and the future of ethics. A principal aim of the work is to help cross C.P. Snow’s ‘great divide’ – in this instance, that between technical specialists and social scientists on the topic of AI.

A globalizing world of AI

As the great wave of digital technology breaks across the world, artificial intelligence creeps increasingly into the very fabric of our lives. From personal virtual assistants and chatbots to self-driving vehicles and telerobotics, AI is now threaded into large tracts of everyday life. It is reshaping society and the economy. Klaus Schwab, founder of the World Economic Forum, has said that today’s AI revolution is ‘unlike anything humankind has experienced before’. AI is not so much an advancement of technology but rather the metamorphosis of all technology. This is what makes it so revolutionary. Politics change dramatically as a consequence of AI. Not only must governments confront head-on the fallout from mass replacement of traditional jobs with AI, algorithms and automation, they must ensure that all citizens are adaptable and digitally literate. It will be fundamental to almost all areas of policy development.

Recent technological breakthroughs have resulted in advanced AI transforming manufacturing, the service industry and business platforms, impacting significantly on most jobs including many professions seemingly immune from digital disruption. Research in the US, UK, Japan and Australia, including both academic reviews and government inquiries, estimates that approximately 40% to 50% of existing jobs are at risk from AI technology and automation in the next 15 to 20 years. Other researchers point to a trend of increasing job polarization accompanying automation. At the same time, it has been estimated that AI could contribute approximately $16 trillion to the global economy by 2030.

Given the intricate interconnections between employment and self-identity, it is easy enough to see why more and more people are troubled by AI. Artificial intelligence is, in short, quickly changing the global economy and, fundamentally, everyday life and the self. Smart algorithms run large tracts of enterprise, executing trades, controlling new additive manufacturing, billing clients, automating customer services, navigating aviation flight paths and guiding surgical care. While there is a public fascination with chatbots and self-driving cars, however, very few people understand how AI actually functions and is changing the world in front of their very eyes. Or maybe this is the issue: AI, like electricity, is invisible. It is a general-purpose technology that works its magic behind the scenes. The contours and consequences of AI remain elusive to us – we can’t see them in action, but we still somehow experience the impact. Like other general-purpose technologies, such as the internal combustion engine, telephony and the silicon chip, AI is becoming ubiquitous. It is everywhere and nowhere at once, both omnipresent and unnoticed.

Whilst there is a lack of agreement among researchers about how to characterize the main defining elements of AI and its related technologies,¹ there is some measure of agreement in the area of public policy and governance. The UK government’s 2017 ‘Industrial Strategy White Paper’, for example, defines AI as ‘technologies with the ability to perform tasks that would otherwise require human intelligence, such as visual perception, speech recognition, and language translation’.² It is perhaps useful to begin with such a definition, one geared to state-promoted AI, if only because such an account is clearly quite narrow, and leaves unaddressed some of the most important deep drivers of AI. It is crucially important, for instance, to underscore the intricate interconnections between AI and ML. A key condition of AI, one not captured by the UK government’s white paper, is the capacity to learn from, and adapt to, new information or stimuli. Among the deep drivers of AI are technological advances in the networked communications of self-learning and relative autonomy of intelligent machines. These new systems of self-learning, adaptation and self-governance have helped to reconstitute not only the debate over what AI actually is, but have also impacted the relationship between artificial and organic intelligence.

While AI generates increasing systems of interconnected self-learning, it does not automatically spawn a common set of human reactions or values in terms of those engaging with such technologies. The relation between AI and its technologies, including particularly people’s experiences of or views on AI, is a complicated one. As a first approximation we can define AI, and its related offshoot ML, as encompassing any computational system that can sense its relevant context and react intelligently to data. Machines might be said to become ‘intelligent’, thus warranting the badge ‘AI’, when certain degrees of self-learning, self-awareness and sentience are realized. Intelligent machines act not only with expertise but also with ongoing degrees of reflexivity. The relation between AI and self-learning is considered to operate at a high level when intelligent machines can cope with the element of surprise. After all, many ML algorithms can easily be duped. Broadly speaking, AI can be said to refer to any computational system which can sense its environment, think, learn and react in response (and cope with surprises) to such data-sensing.³ AI-related technologies may include both robots and purely digital systems that employ learning methods such as Deep learning, neural networks, pattern recognition (including machine vision and cognition), reinforcement learning, and machine decision-making. Let us take a closer look at some of these approaches and technologies.

Machine learning

Machine learning is one of the most important advancements of contemporary AI technologies, where computers execute tasks through processes of ‘learning’ or ‘information gathering’ that draw from (but are not reducible to) human intelligence and human decision making. ‘Machine learning’, writes Toby Walsh,

is an important part of computers that think. It tackles the bottleneck problem, the problem of pouring into a machine all the knowledge we have developed over thousands of years. Programming all that knowledge ourselves, fact by fact, would be slow and painful. But we don’t need to do this, as computers can simply learn it for themselves.⁴

Through analysis of massive volumes of data, ML algorithms can autonomously improve their learning over time. ML relies on algorithms ranging from basic decision trees through to artificial neural networks that classify information by mimicking the structure of the human brain. The rise of neural networks, a kind of ML loosely modelled on the structure of the human brain, consisting of deeply layered processing nodes, has been especially significant in the spread and efficacy of AI. So too, deep leaning – a more recent spin-off of neural networks – which deploys multiple layers of AI to solve complex problems has underpinned much of the explosion of interest from businesses, media, the finance sector and large-scale corporations. The essential scientific aspiration here has focused on replicating general intelligence, which for the most part has been understood largely in terms of reason, cognition and perception, as well as planning, learning and natural language processing.

Natural language processing

Natural language processing (NLP) is a fundamental aspect of AI and encompasses all AI technologies related to the analysis, interpretation and generation (of text- and speech-based) natural language. NLP has prominent applications including machine translation (such as Google Translate), dialogue systems (including Google’s Assistant, Apple’s Siri and Amazon’s Alexa) and automatic question answering (for example, IBM’s Project Debater). NLP has matured rapidly over the past 10 to 15 years as a result of the unprecedented amount of language being produced, shared and recorded in electronic and spoken forms.

The social impacts of NLP as conjoined to AI technologies have been massive, and the likely trajectory of development is set to skyrocket. From Amazon’s Alexa to Google’s Home, people are busy talking to intelligent machines as never before. It is estimated that more than 60% of Internet traffic is now generated by machine-to-machine, and person-to-machine, communication. IT advisory firm Gartner has predicted that by the mid-2020s the average person will be having more conversations with chatbots and robots powered by NLP than with their partner. These claims may seem the stuff of science fiction, but they spell significant change as regards society, culture and politics. I have previously looked at these developments in some detail, focusing on the likely impacts to social interaction and transformations in communication and talk. My argument was that digital devices deploying NLP programs and AI technology are plainly quite divergent from the ordinary conversations of people. Machine talk occurs as part of pre-programmed sequences built up through machine learning. As a result, machine talk – to date at any rate – can usually only respond to conversational contingencies in quite minor ways. Digital devices might be programmed to convey an impression of ‘immediate talk’ geared to the needs of the user, but the production of machine talk is, in fact, drawn from an enormous database of code, scripted utterances and network conversation. For example, most chatbots and virtual personal assistants consist of programmed ‘appropriate replies’ to even the most obscure conversations. This is underscored by Brian Christian’s argument that machine language is a kind of conversational puree, a recorded echo of billions of human conversations. But even this is now under challenge as a result of technological breakthroughs in AI and NLP: for example, Google’s Duplex. Chatbots, softbots, and virtual personal assistants have become increasingly integral to our daily lives and our identities, even if we are not always aware of their role. If talking to chatbots and virtual personal assistants becomes the new normal, we should be aware of the ways they could change how we talk to each other and how we relate to ourselves. One thing is certain. AI is having a profound impact on experiences of the self, what identity means, and of how selfhood intersects with others (both human and non-human) in the wider world.

NLP advances and breakthroughs over the past decade have been achieved with specific tasks and datasets, which are driven largely by big data. However, NLP is only ever as good as the dataset underpinning it. If not appropriately trained and ethically assessed, NLP models can accentuate bias in underlying datasets, resulting in systems that work to the advantage of some users over others. Significantly, NLP is currently unable to distinguish between data or language that is irrelevant and socially or culturally damaging. These are matters of significant social and political importance.

Robotics

Robotics has been characterized as the intelligent connection of perception to action in engineered systems. Robotics include not only human-like robots but any kind of technological system that uses sensors such as cameras, thermal imagers or tactile and sound sensors to collect data about the operational environment and construct an automated response-world of actions.

The scaling up of robotics today is hugely significant throughout much of the world. Industrial robots transforming manufacturing – from packaging and testing to assembling minute electronics – is the fastest growing source of robotic technologies. From the early 1960s when one of the first industrial robots was operationalized in a candy factory in Ontario through to the 2010s where new technologies facilitated robots working hand-in-hand with workers, there has been a growing expansion in robotics and...

Table of contents

• Cover
• Half Title
• Title
• Copyright
• Contents
• List of figures
• List of tables
• List of contributors
• Acknowledgements
• Foreword: The World in 2062
• Part I Social science approaches to artificial intelligence
• Part II Fields of artificial intelligence in social science research
• Index