Читать книгу Making Sense of AI - Anthony Elliott - Страница 8

Another way of reading AI against the grain – contesting the ‘official’ narrative of artificial intelligence – is to rethink its relation to economy, society and unequal relations of power. These are all key domains in which the discourse of AI can and must be situated. I have argued in the preceding section that what the idea of an intelligence rendered ‘artificial’ signifies is, among other things, the transformation and transcendence of human capabilities from natural, inborn and inherited determinations of the biological and biographical realms. AI consists in the project of transforming human knowledge into machine intelligence – and charging social actors with the task of integrating, incorporating and invoking such newly minted artificial automations into the living of everyday life. Such manufacturing of automated intelligent machines, however, works not only upon an internal register – the field of individual life, individualization and the development of human intelligence – but also outwards – across societies, economies and power politics. AI-powered software programs are today downloaded to multiple locations across the planet – at once stored, operationalized and modified. Contrasting the limitations of the human brain by cranial volume and metabolism with the extraterritorial reach of AI, Susan Schneider argues that automated machine intelligence ‘could extend its reach across the Internet and even set up a galaxy-wide “computronium” – a massive supercomputer that utilizes all the matter within a galaxy for its computations. In the long run, there is simply no contest. AI will be far more capable and durable than we are.’¹⁰

So, AI is also all about galaxy-wide movement and especially the automated global movement of software, symbols, simulations, ideas, information and intelligent agents. AI-powered information societies involve a relentless automation of economic, social and political life. This point is an important one to register, as many commentators invoke the spectre of globalization to capture the economic transformations of manufacturing, industry and enterprise as a consequence of AI technology and its deployment in offshore business models. Certainly, a great deal of academic and policy thinking has emphasized how the global digital economy has become ‘borderless’, with many frontiers now automated and regulated through the operations of intelligent machines. The rise of AI is intricately interwoven with globalization, it is often said. This is surely the case, though it is vital to see that globalization links together people, intelligent machines and automation in complex, contradictory and uneven ways. Understanding that AI is both condition and consequence of globalization has to be properly contextualized.

Many studies have cast globalization solely as an economic phenomenon. From this angle, globalization consists of the ever-increasing integration of economic activity and financial markets across borders. Some analyses have emphasized that globalization is the driver of economic neoliberalism, privatization, deregulation, speculative finance and the crystallization of multinational corporations operating across the borderless flows of the global economy.¹¹ It is obvious that such an image of globalization is well geared to rendering AI as simply an upshot of the corporate activities of IBM, Amazon, Google, Microsoft and Alibaba. Other writers have argued that globalization is synonymous with Americanization. AI here is viewed as a set of effects brought about by powerful actors, academic research institutes and industry labs, administrative entities and political forces promoting the Americanization of the world. Much AI research, as we will examine throughout this book, has indeed been funded by the American government, especially the US Department of Defense. Consider, for example, the extensive role of the Defense Advanced Research Projects Agency (DARPA), which during the 1960s poured millions of dollars into the establishment of AI labs at MIT, Carnegie Mellon University and Stanford University along with commercial AI laboratories including SRI International. As I discuss in some detail in chapter 3, the influence of the US Department of Defense upon the digital revolution was hugely consequential and brought in its train a global extension of emergent markets for artificial intelligence.

And so we come back to the big issue of who exactly commissioned the major AI projects that were launched in the 1950s and 1960s. Who was paying for the key AI research breakthroughs? What forms of power were these early commissions advancing and reinforcing? Obviously there were many divergent interests, although the history of the funding cycles around AI clearly suggests that nation-states (especially the United States and, to a much more limited extent, the United Kingdom) along with the biggest multinational companies were the principal actors. Beyond nation-states and corporations, however, another dimension of AI concerns the world military order. Understanding the connections between the techno-industrialization of war, automated techniques of military organization and the flow of AI technologies is very important to grasping the globalizing of AI. I seek to highlight these issues in terms of an institutional account of what I shall call algorithmic modernity, developed with reference to the operations of advanced capitalism, lifestyle change, social inequalities and surveillance, throughout the book as a whole. For the moment, however, it is notable that many of the early successes, as well as some fairly dramatic failures, in AI can be traced to overlaps between military power and the development of automated intelligent machines.

Some argue, rightly in my view, that the rise of AI sprang directly from challenges that the West faced in relation to Soviet communism and the outcomes of the Cold War. Certainly, the general imperative of establishing military dominance in world politics meant that, during the Cold War, the US military sought to automate the translation of documents from Russian and other languages into English. This situation led to considerable state investment in machine translation research. During this initial period of increased defence funding in AI research, a cluster of economic, political and military changes occurred around the late 1950s and early 1960s that were of essential significance to the building of better intelligent machines and advanced AI systems. First, Soviet communism delivered a major shock to the American psyche with the launch of Sputnik, the first artificial earth satellite, in 1957. Beyond this dramatic shock, further reverberations were felt throughout the West in the same year when Russia launched Sputnik 2, a spacecraft that put Laika the dog into orbit. The idea of a space future successfully colonized by Soviet-bloc countries spurred the USA into dramatically increasing spending – military and otherwise – on science, technology and research. Second, new research funding in AI – from machine translation to speech-recognition projects – was launched in America by agencies including the CIA, the National Science Foundation and the Department of Defense. This increasingly defence-driven system of research innovation resulted in a much greater speed-up of advances in automation as well as other breakthroughs in machine intelligence.

Third, during this period of state-led AI research investment in the 1960s, various socio-technical and cultural shifts took place as regards the promise, power and prestige of automated machine intelligence. The establishment of the Advanced Research Projects Agency (ARPA) in 1962 represented, for example, a gigantic effort to ensure that America was first to land on the moon. Beyond the space race, however, this entity ushered into existence other world-transforming contributions too, most notably breakthroughs in advanced computing and automated system architectures led by J. C. R Licklider. A psychologist with a passion for mathematics and mechanical engineering, Licklider served at the Pentagon and sought to expand ARPA (and subsequently DARPA, with the D added in 1972) beyond its narrow military confines by supporting multiple AI research projects and associated breakthroughs in advanced computing. As a chief networker among networked researchers and technologists, Licklider authorized support for many projects, including the work of John McCarthy, as well as projects at Carnegie Mellon University, SRI International and the RAND Corporation. His major legacy was to develop a computer network linking these colleagues and research projects together, initially pursued through Project MAC – the development of multi-access computing. This, in turn, culminated in the establishment of ARPANET – a computational network which was, in effect, the forerunner of the Internet and the World Wide Web. But it was ideas as well as inventions for which Licklider deserves a prominent place in the history of artificial intelligence. The digital transformation envisaged by Licklider was captured most vividly in his 1960 paper, ‘Man-Computer Symbiosis’. This was a dramatic advance beyond Turing’s notion that machines might one day think. Licklider’s vision, by contrast, was all about intuitive interactive computing, the interface of human and machine. In his compelling intellectual history The Dream Machine, M. Mitchell Waldrop argues that Licklider

was unique in bringing to the field a deep appreciation for human beings: our capacity to perceive, to adapt, to make choices, and to devise completely new ways of tackling apparently intractable problems. As an experimental psychologist, he found these abilities every bit as subtle and as worthy of respect as a computer’s ability to execute an algorithm. And that was why to him, the real challenge would always lie in adapting computers to the humans who used them, thereby exploiting the strengths of each.¹²

In this speaking up for interactivity, technological interfaces, decentralization and connectivity, Licklider can in many ways be said to have shaped AI as we know it today.