Language and technology
The development of writing enabled knowledge to be communicated across distances of time and space and the information sharing mechanism thus emerged advanced our civilization further. The ability to use language for rational discussion among humans and the distribution of information, lead to the cultural evolution of complex societies and the emergence of knowledge underlying modern technologies. Therefore, language is intricately linked with technology and these two disciplines intersect on multiple levels. In this lecture, we will be looking at some of those aspects closely.
Language is often literally connected to technology or is even part of what the technology is. Consider some of the technologies we use daily, such as computers, mobile phones, and even robots. While these devices are tangible physical objects, they depend on software, which is essentially a form of language or code. Both the physical components and the linguistic aspects are integral to the technology's operation. Code alone cannot create any real-world impact; it needs hardware, among other software components, to function. In the same way, hardware relies on the language of software to execute tasks.
In contemporary information and communication technologies (ICTs), the interplay between language and technology is undeniable. To understand the agency and essence of these technologies we need to include the discussion of language in the equation. Even at a technical level, these technological tools exhibit a dual nature, comprising both material and linguistic elements simultaneously.
When it comes to communication technologies, language takes up the crucial role of the interface which enables interactions between the human user and the technology. It is through language that we interact with technology. Take the case of personal computers, the internet or social media, and social devices like home assistants and social robots. These technologies use text-based or voice-based interfaces that rely on language. All of this points to how language is both a part of technology and a technology in itself. It functions as an interface technology, a communication technology, and an information technology.
One approach to understanding the impact of language on technology is to suggest that language mediates human–technology relations. It functions as a non-neutral medium which is capable of actively influencing and moulding technology itself. It also influences our perception of the world, and our interactions with that technology. Our relationship to technology is always mediated, and part of that mediation is accomplished by language.
Post phenomenology is a philosophy of technology that understands technologies in light of how they mediate human-world relations by co-constituting the subjectivity and objectivity of experience. That is how technology influences our perception of the world. It posits the role of technology as not a mere instrument but capable of shaping our experiences and our actions. While the pioneers of post phenomenology have rightly pointed out the role of material technologies as mediators, the mediating role of language and its varied and sometimes complex relations to technology, is often neglected. Not only technology but language also shape our experience and our actions. I hope you recall the discussion we had on the same, in one of our previous lectures.
An important thing we have to emphasise is that Language encompasses more than just individual words and sentences; it extends to the creation of broader structures like discourses and narratives. Similarly, technology is not just about the physical devices but also the discussions surrounding them. Specific technologies, such as robots and computers, are closely tied to the narratives and discussions we construct about them. Both as individuals and as societies, our reactions to technologies, or even the conceptualization of these technologies, are influenced by the meanings we assign to them. These assigned meanings shape the essence of the technology.
Within these discourses, fiction also plays a significant role. For instance, in the discourse about robots, stories like Frankenstein and movies like Terminator have contributed to various fears surrounding robotics and artificial intelligence, impacting the use and development of these technologies. Engineers and computer scientists are aware of this influence. If there is excessive fear within the general public, often driven by narratives like Terminator, they may face resistance to their technology. Consequently, they make efforts to reshape the discourse in a direction they deem more favourable. They might emphasize the materiality of technology, such as emphasizing that robots are mere tools or machines.
As observed by twentieth-century theorists like Foucault, discourse is intricately linked to interests, knowledge, and power dynamics. Various stakeholders with diverse backgrounds and positions are involved in shaping these discourses, as highlighted in fields such as science and technology studies. In doing so, they implicitly acknowledge that discourse isn't just mere language or text. Much like technology, language significantly impacts how we perceive the world and our actions. Interventions from scientists and technology developers, which encourage a focus on the objective facts and material reality of technology (for example, asserting that a robot is a machine, not a human being), rely on the belief that words and discourse are indeed influential. When they intervene to guide the public on how to use language, it means that they take language very seriously.
Many individuals who engage in critical discussions about technology in today's world often do so by drawing a contrast between technology and human values, principles, and the human way of life. Technology is frequently perceived as existing within a distinct, non-human realm, one that differs from or is even antagonistic to the human sphere. For instance, in the context of artificial intelligence, there's an emphasis on the need to ensure that human values are upheld as if AI was entirely disconnected from human values to begin with. These critiques often assume that contemporary technology is problematic, and they sometimes idealize a past state, a sort of Eden before technology, which was considered good and harmonious until technology caused a "Fall."
For instance, people might argue that the Internet and mobile devices have negative impacts to society compared to television, which used to bring families together, whereas now, each individual has their own screen. The aim here is not to assert that these criticisms are entirely false (there might indeed be some truth to them), but rather to reveal the recurring patterns of discourse and narratives that influence our present perspectives on technology. Language, in the form of culturally and historically developed discourse and narrative, significantly shapes the way we perceive technology, and in turn, it shapes our conceptions of what technology is and what it should be.
On the other hand, building machines that can understand language has been a central goal of the field of artificial intelligence dating back to its earliest days, even though the task has proved rather difficult. As our civilization rapidly went through technological advancements, the need to create computational devices that could interact with humans through natural language gradually emerged. This form of communication would facilitate human-machine interaction and allow for the development of useful systems. There was also, a motivation for the development of intelligent systems and, as the language used by humans is distinct from the forms of communication used by other animals, it helps to support the belief that intelligence is closely linked to human language.
Since language is widely regarded as the quintessence of human social intelligence, an AI that can truly understand language the way a human would, by implication would be capable of any other human-level intellectual activity. Put simply, to solve language is to solve AI. The association of intelligence with the ability to understand natural language is generally attributed to Safir-Whorf's hypotheses (Whorf et al., 1956), which state that the structure of a language affects the worldview or cognition of its speakers and, therefore, people's perceptions are related to their spoken language.
This profound insight is at the heart of the “Turing test,” introduced by AI pioneer Alan Turing in a groundbreaking 1950 paper. Though often critiqued or misunderstood, the Turing test captures a fundamental reality about language and intelligence.
Human-level comprehension and production of language still remain an AI-hard problem. Humanity has yet to build a machine intelligence with human-level mastery of language. But over the past few years, researchers have achieved startling, game-changing breakthroughs in language AI.
AI’s intersection with human language manifests in the field of Natural Language Processing (NLP). NLP has evolved alongside the earliest advancements in AI, focusing on teaching machines to understand and generate human language. Since language is integral to human social intelligence, in a future where human capabilities are enhanced by AI tools, language will continue to play a vital role. Recent developments in NLP have resulted in highly adaptable pre-trained language models capable of handling various comprehension, text generation and reasoning tasks.
As the field of NLP and other AI technologies increasingly get better at understanding and generating human language, there is a growing concern about the negative impacts of technology on language.
A recurrent concern which arises is, Is technology capable of corrupting language?
The anxiety about technology changing language is nothing new, for example, Greek philosopher Plato was against the act of writing things down, as he believed it would have a detrimental impact on our memory. Throughout history, nearly every subsequent innovation, including the printing press, radio, telegraph, television, movies, and the internet, has faced accusations of corrupting language.
Even though the displays of AI-generated language are remarkable, they are rooted in a very narrow definition and understanding of what language is. For a computer to analyse, process and classify something as language, it must initially be in written form.
Computers, which can engage in conversations with humans or even produce what might be considered Beat poetry, are equipped with software applications known as neural networks. These neural networks are specifically designed to identify patterns within extensive datasets. Over time, these networks acquire the ability to replicate the identified patterns. Similarly, other language models developed by companies like Meta (Facebook) or OpenAI, are trained using data extracted from public platforms such as Reddit, Twitter, and Wikipedia.
This approach neglects all the non-written means of communication, such as sign language, oral traditions, body language, intonation, and the broader cultural context that enriches nuanced human interaction.
Writing, a relatively recent human technology emerging just 5,400 years ago, stands in stark contrast to spoken language, which has roots dating back at least 50,000 years. Unlike spoken language, writing is a more recent innovation that most humans find less instinctual. Children naturally acquire the ability to speak within a few years, whereas mastering the abstract conventions of spelling and syntax often takes many years of formal education.
Moreover, writing is not a universal practice across the world's approximately 7,100 "natural languages." Only about half of these languages have a written form. To bridge this gap, audio recordings and voice recognition tools can be employed, but these solutions require algorithms to be trained on extensive datasets, ideally encompassing millions of diverse speakers.
Many oral languages are spoken by small, historically isolated populations, both socially and geographically. When these oral languages are incorporated into databases, they must undergo transcription and encoding into written formats. However, a significant challenge arises from the fact that written words are never a perfect representation of spoken language. Initial transcriptions necessitate determining what constitutes the "standard" dialect and encoding the numerous non-linguistic cues that accompany spoken language. These decisions involve uncomfortable value judgments, especially when made by linguists or anthropologists from outside the community. Often, these choices reflect more about the power dynamics within the community of speakers than they do about how the language is actually used by the majority.
A more fundamental issue lies in the fact that the spoken nature of many languages is what provides them with their practicality and the ability to infuse culture with vitality. Take, for example, many Indigenous languages in contemporary North America, where the act of storytelling is inseparable from its contextual surroundings. Documenting these stories in writing and freezing their meaning may strip them of their capacity to remain as living, dynamic cultural elements.
While transcribing marginalized oral languages can be a means of preserving them, this process is full of complex ethical considerations. For some Indigenous communities, who have endured trauma due to decades of forced assimilation through residential schools, the written script itself can be perceived as an instrument of colonization and exploitation. Anthropologists share part of the responsibility, as certain scholars have left a detrimental legacy by transcribing and publishing sacred stories, often never intended for wide public consumption, without seeking permission from community knowledge keepers.
When viewed from the perspective of linguistic anthropology, it becomes evident that cars and chatbots developed for "natural language processing" do not truly possess command over language. Instead, they execute a limited subset of language competence, a fact often overlooked when the technology industry emphasizes sensational claims of AI having sentience. Language, in its dynamic existence, is significantly more complex.
The fundamental question of Can machines ever reach a point where they comprehend language as humans do? is humorously depicted in the following XKCD comic strip, highlighting that language is not a rigid system but rather a complex and unpredictable phenomenon.
The comic underscores how the same words can carry vastly different meanings depending on subtle cues like gestures, intonation, pauses, or cultural nuances, even within the same region, not to mention across diverse cultures and languages.
In everyday life, conversations are a complex interplay of various forms of communicative cues. Real conversations are disorderly, with participants occasionally talking over each other, negotiating for speaking rights, and pausing to find the precise words.
The importance of context in grasping language is apparent to anyone who has attempted to convey sarcasm or irony through written communication, such as email. The manner in which someone says something has a more substantial impact on its meaning than the words themselves. Nonverbal signals like tone of voice, eye movements, or exaggerated facial expressions can lead listeners towards interpretations that may, at times, contrast with the literal meaning of the words.
Moreover, speakers frequently employ subtle signals in their discourse that are only comprehensible to individuals familiar with the same cultural customs. For instance, in North America and various parts of Europe, people often employ conventions like air quotes or introduce a statement with phrases like "She was like..." to quote someone else's speech. At times, a speaker's voice modulation indicates when they are quoting someone else’s words. Additionally, nonverbal behaviours like nodding and common interjections such as "uh-huh" serve as culturally specific forms of "back-channelling" that encourage the speaker to continue their train of thought. Unfortunately, these nuances are lost when language is transcribed in written text.
Nonetheless, computer scientists and computational linguists have achieved remarkable progress in the capabilities of large language models, such as OpenAI’s large language models. In specific contexts, such as text-based conversations, machine-generated text can be nearly indistinguishable from that produced by a human. Nevertheless, whether examining purely oral languages or considering the non-written cues present in everyday dialogues, spoken language proves to be far more complex and captivating than what can be conveyed on a page or screen.
This distinction underscores the truly unique and inherently human nature of the world of language.
One potential scenario that holds significant promise is the pivotal role of technology in shaping the future of language. As artificial intelligence continues to advance, the landscape of language could undergo profound transformations. We can anticipate the emergence of increasingly sophisticated machines that possess the ability to comprehend and generate human language, marking a noteworthy turning point in the evolution of communication. This intriguing prospect raises questions about the potential development of entirely new languages or specialized dialects specifically tailored for machine use.
Conclusion
The development of writing facilitated the transmission of knowledge, contributing to the cultural evolution of complex societies and the emergence of modern technologies.
Language and technology are intricately linked, as evident in the technological devices we engage with in our everyday lives. And both physical components and linguistic aspects are integral to the operation of these technologies.
In communication technologies, Language functions as an interface for human interactions with technology.
The Discourses and narratives surrounding technology, significantly impact how individuals and societies perceive and react to technological advancements.
In AI, Natural Language Processing (NLP) plays a key role with recent breakthroughs in language models raising concerns about the potential negative impacts of technology on language.
The evolution of writing, which is a relatively recent human technology, stands in contrast to spoken language, and the transcription of oral languages poses complex ethical considerations.
Spoken Language is inherently complex, with non-verbal cues and cultural nuances as functions as important contributors of meaning in communication.
The future of language may be shaped by advancements in AI technologies, potentially leading to the development of new languages or specialized dialects tailored for machine use.
