Language Revolutions: Unraveling the Evolution of Linguistics and AI
As The Language Matrix blog series attempts to explore how human languages and AI language models influence each other, let us first understand the linguistic histories of both these fields.
Linguistic evolution refers to the application of the theory of evolution to study the ways in which language has evolved through history, from mere primeval utterances to a sophisticated medium of articulation of thought. Evolution then is not merely a process that affects biological aspects but also refers to cultural evolution and the ways in which different aspects of culture change over time. Linguistics is a specially interesting area to explore here because it exists at the bridge between biology and culture, as biological evolution is not enough to explain the development of the thousands of complex languages we have today. As linguistic evolution continues to occur even today, it becomes important for us to unravel its history.
Language is an ever-changing construct: pronunciations evolve, new words are borrowed, while some go extinct, syntax structures change, or the morphology may develop amidst other things. Even as the evolution of language is hardly noticeable on a daily basis, the changes become much more drastic and noticeable over generations. Over a century, the changes are tremendous, and over a couple of centuries, even the same language is hard to read. Therefore over a period of time, the same language splits into many diverse branches. Thus, it is hard to believe that Sanskrit and Latin share the same language root in Proto-Indo-European. Language then evolves because of a variety of processes as it is transmitted from one generation to the next. Each generation reconstructs grammar based on their immediate cultural influences. Processes like migration and trade bring languages into contact with one another, thereby influencing and merging them through borrowing sounds and words. Further different social groups introduce distinctive variances in languages to mark their group as different from others.
While this forms the broad history of linguistic evolution, the history of the evolution of AI is also quite fascinating. Around the 1950s, scientists in Europe and America were already immersed in the scientific possibility of artificial intelligence that would allow machines to use available information for problem-solving purposes. The first person to credit here would be the British mathematician Alan Turing, for forming the basis for artificial intelligence through the Turing test. However, as computers remained extremely expensive, artificial intelligence remained a pipe cream. In 1956, at the Dartmouth Summer Research Project on Artificial Intelligence, John McCarthy and Marvin Minsky fostered a collaborative platform for open-ended dialogue on artificial intelligence, a term that McCarthy coined at the same conference. In the following decades as computers were able to store more information and became cheaper, machine learning algorithms were also improved as computer experts were able to write better algorithms for problem-solving. Here, Joseph Weizenbaum’s ELIZA is worth noting as a remarkable demonstration of the ultimate goal of interpreting spoken language by computers. Such attempts bolstered government funding towards the creation of Machines that could transcribe and translate spoken languages. In 1980, the development of deep learning techniques by John Hopfield and David Rumelhart allowed computers to use past information and experience for future decision-making. This radically improved the prospect of the creation of AI as computers could mimic human decision-making to some extent. However, despite some minor victories such as the defeat of chess champion Gary Casper in 1997 by IBM’s deep blue computer program and massive public investment in the creation of AI could not faster a major breakthrough. However, the recent advent of big data in the 21st century which has allowed the collection and processing of huge sums of data has ushered in the new age of AI owing to these tremendous amounts of data available which can be used to train computer algorithms through machine learning and predictive analysis to finally be able
Given that computers now have the ability to mimic human language and voices, we now enter the age of AI. As vast amounts of research are still underway, AI language technologies can fundamentally reshape the way we engage with language. This blog series is an attempt to understand precisely this influence.
Sources: