I’ve written a fair bit in the last month on the development of AI art tools, but what about language? AI, as you’re probably aware, is not only able to mimic artistic styles. Its developers also want it to generate words, and to all appearances, they are succeeding. If visual artists are in trouble, how are journalists, novelists, and academics implicated in the AI revolution?
I have a background in English, literature, and creative writing, so naturally, this AI issue hits a bit closer to home. Suppose an AI program could compose a short story with the prose quality and cohesive style of Ernest Hemingway. Could AI eventually produce news content, thus substituting the human reporter or journalist? As it turns out, writers’ jobs are pretty safe, despite the sophistication of AI language generators. Why? Linguistic anthropologist Joseph Wilson explains, writing,
The displays of AI-generated language are impressive, but they rely on a very narrow definition of what language is. First of all, for a computer to recognize something as language, it needs to be written down. Computers capable of chatting with a human being, or writing what could be deemed Beat poetry, are programmed with software applications called neural networks that are designed to find patterns in large sets of data…But this excludes all unwritten forms of communication: sign language, oral histories, body language, tone of voice, and the broader cultural context in which people find themselves speaking. In other words, it leaves out much of the interesting stuff that makes nuanced communication between people possible.”
Like AI art tool Stable Diffusion, language generators depend on a human programmer to work. Wilson also helpfully points out that not all languages are written down. Some cultures are still primarily “oral.” Until the Gutenberg press, language was usually communicated via the ear, not the eye. Even then, physical books were often a luxury. AI language generators can’t replicate body language, oral intonation, and facial expression, and so limit language’s full scope.
AI also suffers from a lack of context, according to Wilson. All language is context-dependent, and this is nowhere more obvious than in everyday conversation. Wilson writes,
The importance of context in understanding language is obvious to anyone who has tried to convey sarcasm or irony through email. The way someone says the words, ‘I love broccoli,’ for example, determines its meaning more than do the words alone. Nonverbal cues such as tone of voice, rolling of the eyes, or an exaggerated facial expression can nudge listeners toward interpretations that are sometimes the exact opposite of the words’ literal meaning.”
Language, then, is infused with uniquely human traits—tone, expression, body language, etc. People are always more well-rounded and fleshed out than computer code.
The question of creativity also haunts the AI project to replicate human language. As noted above, all AI depends on the natural intelligence of human beings to function. In that sense, machines will never be “linguistic.” They are always derivative of human innovation. Furthermore, if language is a fundamental trademark of human life, then it seems that language and culture are inextricably related. Language enables us to share ideas, values, and perspectives. Words communicate the relatedness of things, and so indicate that our world can be, at least in some ways, deciphered, uncovered, and made inhabitable. Especially in the literary arts, language tries to find ways to explore the meaning of being human. Fiction delves into the inner thoughts and feelings of characters. Poetry seeks to bear witness to the beauty and suffering of daily existence. Can AI generate the same kind of depth and meaning? Can AI short stories move the heart, or AI “poetry” aptly do justice to the range of human emotion?
While the benefits of AI are ample, the arts, which at their best speak to truths no algorithm can express, should be seen as a uniquely human endeavor. At the end of the day, language is about one person speaking to another, with emphasis on the person.
Mind Matters features original news and analysis at the intersection of artificial and natural intelligence. Through articles and podcasts, it explores issues, challenges, and controversies relating to human and artificial intelligence from a perspective that values the unique capabilities of human beings. Mind Matters is published by the Walter Bradley Center for Natural and Artificial Intelligence.

source