Artificial intelligence: looking though the Pygmalion Lens

摘要

The AI debate brings to our notice, on the one hand, the danger of singularity, and on the other, the enchantment of AI Futures of a data-driven world. Singularity is seen in terms of human beings 'enslaved by enormously intelligent computers', supported by the claim that 'humans are no more than biological machines'. The enchantment of AI Futures is stimulated by entrepreneurial opportunities offered by deep learning and machine learning tools in domains ranging from health and medicine to Industry 4.0 projects. Whilst these narratives continue to evolve, we feel the wrath of the 'god of algorithms' in feeling helpless when confronted by the customer relations sermon, "the Computer Says NO", bereft of any common sense decisions. As tempting though the digital sermon of the intelligent machine may be to the tech prophets, the concern here is with how would we cope with gaps between the complexity and ambiguity of our living world and the unpredictable algorithmic miscalculation. In exploring this concern, our attention is drawn to a new universal narrative of "Dataism" (Harari 2015), propagated by the new high-tech 'Platonians of the Silicon Valley'. This narrative legitimizes the authority of a giant data flow system, defined by algorithms and inhabited by emails, blogs, Apps, Facebook, Twitter, Amazon and Google. It is as if the Pygmalion AI philosophers of today are enthralled by the universality of the Turing Machine, and are engaged in anthropomorphizing the robot 'Eliza' into a 'robotic duchess' of human society. This algorithmic manipulation is not only continuing the historical disconnect of language from its cultural bearings, it is leading us to subconsciously accept the imitation machine as an 'unpalatable truth', or leading to our 'willful' blindness' limiting our ability to imagine the 'unthinkable'. Phil Rosenzweig (https://www.mckin sey.com/business-functions/strategy-and-corporate-finan ce/our-insights/the-benefits-and-limits-of-decision-model s) asks us to understand the limits of the predictability of data-driven decision models, technically dazzling as they are, for example in detecting fraudulent credit-card use and predicting rainfall. But these predictions can neither change the behaviour of card users nor of the farmers to benefit from weather predictions without wise counselling of card users and without the wisdom of experiential knowledge of farmers to manage and improve crop yields. Data-driven decision models, in computing predictions of complex and large databases 'may relieve the decision makers of some of the burden; but the danger is that these decision models are often so impressive that it's easy to be seduced by them', and to overlook the need to use them wisely. As Rosenzweig says, 'the challenge thus isn't to predict what will happen but to make it happen, and how to control and avoid the adverse happenings'. Whist social media in the form of Facebook, twitter and Google, powered by the intelligent machine, draws and captures our attention, we are in danger of becoming mere passive observers and losing sight of the new social, cultural, ethical, and political tensions created by the intelligent machine. These new tensions exacerbate the already current conditions of conflict, vulnerability, and instability arising from globalization. In the pursuit of a new paradigm of artificial intelligence for common good, we need to reflect on the potential and limits of the dream of the exact language and the limit of digital discourse promoted by the proponents of the intelligent machine.