Rinat M. Nugayev

The success of deep learning in natural language processing raises intriguing questions about the nature of linguistic meaning and ways in which it can be processed by natural and artificial systems. One such question has to do with subword segmentation algorithms widely employed in language modeling, machine translation, and other tasks since 2016. These algorithms often cut words into semantically opaque pieces, such as ‘period’, ‘on’, ‘t’, and ‘ist’ in ‘period|on|t|ist’. The system then represents the resulting segments in a dense vector space, which is expected to model grammatical relations among them. This representation may in turn be used to map ‘period|on|t|ist’ (English) to ‘par|od|ont|iste’ (French). Thus, instead of being modeled at the lexical level, translation is reformulated more generally as the task of learning the best bilingual mapping between the sequences of subword segments of two languages; and sometimes even between pure character sequences: ‘p|e|r|i|o|d|o|n|t|i|s|t’ → ‘p|a|r|o|d|o|n|t|i|s|t|e’. Such segmentations and alignments are at work in highly efficient end-to-end machine translation systems, despite their allegedly opaque nature. But do they have linguistic or philosophical plausibility? I attempt to cast light on this question, in the spirit of making artificial intelligence more transparent and explainable.

Natural science, properly understood, provides us with the methodological key to the salvation of humanity. First, we need to acknowledge that the actual aims of science are profoundly problematic, in that they make problematic assumptions about metaphysics, values and the social use of science. Then we need to represent these aims in the form of a hierarchy of aims, which become increasingly unproblematic as one goes up the hierarchy; as result we create a framework of relatively unproblematic aims and methods, high up in the hierarchy, within which much more problematic aims and methods, low down in the hierarchy, may be improved as scientific knowledge improves. Then, we need to generalize this hierarchical, aims-and-methods-improving methodology so that it becomes fruitfully applicable to any worthwhile endeavour with problematic aims. Finally, we need to apply this methodology to the immensely problematic task of making progress towards as good a world as feasible.

The crisis of our times is science in a world without wisdom. The immense intellectual success of modern science and technology have given some of us unprecedented powers to act, which has led to all the great benefits of the modern world, and to the grave global crises we now face. Before modern science, we lacked the power to do too much damage to ourselves or the planet; now we have science, wisdom has become, not a private luxury but a public necessity. Unless we learn to become a bit wiser, sooner or later we will destroy ourselves. But how can we learn to become wiser? We can learn from the last place many would look - namely science! We can learn from scientific progress how to achieve social progress towards a wiser world. This is in fact an old idea. It goes back to the 18th century French Enlightenment. But, in developing the idea, the philosophes of the Enlightenment made three gross blunders. They failed to capture correctly the nature of the progress-achieving methods of science; they failed to generalize these methods properly; and, most disastrously of all, they applied their misconstrued scientific method, not to the social world directly, not to the task of making social progress towards an enlightened world, but instead to the task of improving knowledge of social phenomena, to the task of developing social science. This botched version of the profound Enlightenment idea was developed throughout the 19th century, and built into universities in the early 20th century with the creation of disciplines of social science. As a result, academic inquiry today is a botched version of what we need. Instead of giving priority to problems of living, and devoting itself to helping humanity acquire a bit more wisdom, academia today devotes itself to the pursuit of knowledge. That is the key disaster of our times. Our institutions of learning seek knowledge, but fail disastrously to devote themselves to helping humanity acquire a bit more wisdom. If the world is to acquire global wisdom, it needs institutions of learning rationally devoted to the task. As a matter of supreme urgency, we need to bring about a revolution in academia so that the basic task becomes, not just to acquire knowledge, but to seek and promote social wisdom.

In my book Understanding Scientific Progress, I argue that fundamental philosophical problems about scientific progress, above all the problem of induction, cannot be solved granted standard empiricism (SE), a doctrine which most scientists and philosophers of science take for granted. A key tenet of SE is that no permanent thesis about the world can be accepted as a part of scientific knowledge independent of evidence. For a number of reasons, we need to adopt a rather different conception of science which I call aim-oriented empiricism (AOE). This holds that we need to construe physics as accepting, as a part of theoretical scientific knowledge, a hierarchy of metaphysical theses about the comprehensibility and knowability of the universe, these theses becoming increasingly insubstantial as we go up the hierarchy. Fundamental philosophical problems about scientific progress, including the problems of induction, theory unity, verisimilitude and scientific discovery, which cannot be solved granted SE, can be solved granted AOE. In his review of Understanding Scientific Progress, Moti Mizrahi makes a number of criticisms, almost all of which are invalid in quite elementary ways.