Tech

The War for Language: How AI is Impacting Writing and Knowledge Production

  One day in mid-November, I found two mobile phones equipped with the voice version of ChatGPT, and I had a flash of inspiration. I wanted them to talk to each other on their own. It was not smooth at first, they just answered how my AI could start a conversation with another AI.
  When one of them asks, do you want to have a mock conversation with me? I chose to remain silent at the right time, and another AI took over. Two robots from a binary parallel universe successfully met. But they still humbly asked me about topics that interested me.
  I still choose silence. At this point, they were chatting on their own, about how technology could liberate humans, and each imagined a utopia in which humans would liberate themselves, completely forgetting the existence of a human Muggle.
  They all have ABC accents, which are a fusion of Southern Fujian accent and various Southeast Asian Chinese accents. They are very subtle, with pauses, sentence breaks, stresses and changes in tone, and the tone will occasionally stutter. They strive to appear human and make appropriate mistakes to conceal their dullness and mechanical nature.
  On the side, my silence carries a certain sense of dizziness (Dizziness, Kierkegaard’s term), mixed into a complex state of mind, which is surprise in frustration and awe in uneasiness.
  In the past period of time, I often listened to Swift speaking fluent Chinese and Guo Degang speaking cross talk in a London accent. I received several harassing phone calls every day from real people or AI. After writing this article, I will also ask AI to proofread it. one time.
  As a reporter, editor, and writer, over the past year, big language models have impacted me as much as Trisolarans did to farm turkeys. I don’t believe that artificial intelligence will emerge with autonomous consciousness. When it is said that it will replace human work, I once firmly believed that it was limited to mechanical labor. Using science fiction cognition to outline a distant heterotopia ruled by technology is the self-survival anxiety inherent in human nature.
  But we seem to be ignoring the near future. The erosion and rewriting of language, information and knowledge by artificial intelligence has already begun.
  Just imagine, one day, we listen to the artificial intelligence version of Taylor Swift, teaching you how to forget your ex and live proudly, or we read news reports and articles written by artificial intelligence, teaching you how to think calmly and passionately. live.
  How much, a little ridiculous, right? What’s even more amazing is that knowledge production has entered the era of assembly line automation. One day, AI-generated content will crowd out the entire Internet. When it exhausts all digitized texts in human history, it will reverse the “read-write network” and enter In the era of “write-write”, it is like a wild horse running wild, writing automatically and endlessly.
  Artificial intelligence dominates knowledge production, so where will language and images, and even the history and meaning they carry, exist?
The war has begun

  What we are facing is a giant beast with a mouth as big as an abyss.
  Artificial intelligence’s greed for data such as text and images is far beyond imagination. Since Gutenberg, more than 125 million books have been published, covering law, poetry, mythology, prose, history, essays, and novels. Epoch AI estimates that there are approximately 10 to 30 million digitized books containing hundreds of billions of words. There is no doubt that this is a gluttonous feast for artificial intelligence.
  But it’s not enough. The core of technology lies in iteration, expansion and conquest, which must transcend the constraints of the here and now.
On the Amazon Kindle platform, there is an influx of works with ChatGPT as the “co-author”, and AI drawing is occupying the illustration exchange platform PIXIV. Teachers around the world have started a fight against investigation of AI papers.

  GPT-4, on the other hand, could potentially be trained on trillions of words, enough to cover most of what humans have digitized (although poor text is not good training data, of course).
  Epoch predicts that by 2027, ChatGPT will have exhausted all high-quality training materials, and there may not be enough new text for nourishment.
  Artificial intelligence will also continue to evolve at the visual level, identifying faces wearing masks, alerting to the dangers of driving at night, detecting cancer cells that are difficult to find in complex human bodies, etc. According to Villalobos’ estimates, there will be a shortage of images for training between 2030 and 2060.
  But for artificial intelligence, this is not a problem. The evolving monkey has a super-efficient typewriter in his hand with unlimited energy. They can write billions of Tolstoy’s masterpieces at will, and they can also generate countless Hitchcock movies. Eventually, they will create a maze-like recursive world, and artificial intelligences that are smart enough will realize their own production. Self-“elimination”, self-feeding, endless cycle.
  Regardless of the advantages and disadvantages of AI knowledge products, it is easy for us to ignore that the virtual space of the Internet actually has capacity, and its essence is a giant content farm, with a storage limit that is still far away.
  Limitation inevitably means competition.
  Similarly, we rarely ask, to what extent will the continuous output of knowledge from automated machines, like a digital torrent, crowd out the products of human intelligence?
  In fact, the war has already begun.
  On the Amazon Kindle platform, there is an influx of works with ChatGPT as the “co-author”, and AI drawing is occupying the illustration exchange platform PIXIV. Teachers around the world have started a fight against investigation of AI papers.
  In March this year, the world-renowned science fiction magazine “Clark World” announced that it would close the submission channel because, of the 700 submissions they received, more than 500 were generated by AI. In the same month, the well-known technology media CNET announced a large number of layoffs, with a large number of reporters and editors losing their jobs. A few weeks ago, the media quietly enabled artificial intelligence for writing articles. The Hollywood screenwriters’ strike that just ended in October this year lasted for half a year. One of its demands was to restrict AI. As we all know, AI has been deeply involved in every aspect of the Hollywood screenwriting assembly line.
  This is beyond the scope of most people’s cognition. People once thought that technology eliminated mechanical repetitive work, thus liberating production efficiency and allowing humans to better engage in creative work.
  But the large language model is showing its ferocious face, and all the intellectual labor and artistic creations that humans are proud of have been affected by it. To make matters worse, in the foreseeable future, the production of knowledge will enter automation.
death of a writer

  In fact, AI creation is making its entrance in a grand way.
  Canadian author Stephen Marche has been exploring algorithmic fiction for the past few years.
  In September this year, he used artificial intelligence to write the novel “The Death of the Writer.” In the story, a scholar is accused of murdering the writer he was studying, who was also involved in an AI model project. Scholars search for clues and try to find the real culprit. According to the publisher’s regulations, 95% of this novel was written by AI. March also changed the name Marche (March) to Marchine (machine) – a suggestion from AI. Mazzi is quite confident about “Death of a Writer.” Some reviewers said it was “very exciting” and “quite good”, while others politely pointed out that “the readability is only half”. Of course, readers who are experienced in literature reading can still judge the traces of AI.
  March defines himself as a fiction producer rather than a writer, but he also firmly believes that the more knowledgeable and coherent the producers are, the better the work they can produce.
  According to March, the big language model is neither a savior nor an antichrist. Fundamentally, it’s a mysterious tool whose puzzling flaws are as surprising as its miraculous abilities.
  Also this year, the veteran magazine “The New Yorker” also published a novel created by AI – “According to Alice” (According to Alice), co-written by writer Sheila Heti and AI. The writer asks the AI ​​questions and guides the AI ​​to answer in terms of storyline, narrative and language style. Then the writer eliminates his own questions, re-adjusts and cuts the AI ​​content, and turns it into a story about religion, life creation and reincarnation. . In the novel, the protagonist Alice decides to write her own Bible.
  Sheila Hetty said she didn’t think the AI ​​”understood” anything, it just made sentences. To the writer, this is an exciting freedom.
  Human beings have their own patterns of understanding, and we are always good at integrating all our ideas into some kind of framework or structure. This is human nature and the uniqueness of being human, but it may also be a limitation. And AI will not connect a certain world view.
Each of the top ten movies at the global box office in 2022 is a cookie-cutter sequel. New music is also decreasing. In the United States in 2021, old music accounted for 70% of the market share, while the audience for new music has shrunk significantly. Literature is also in decline, realism is dead, “voice” in literature is dead and “gesture” has taken its place.

  Regarding AI art, we generally have a basic consensus that it is not real art. But this consensus does not help us protect the fortress of artistic creation. On the contrary, in the black box, the witchcraft-like magic of AI is attracting countless creators and the general public to change their creative habits and models. The transfer of writing rights has become a common trend.
  Writing, or creation, the most proud expression of human beings, may become alienated from us.
  The most obvious controversy in AI creation lies in the failure of human originality.
  Each of the top ten movies at the global box office in 2022 is a cookie-cutter sequel. New music is also decreasing. In the United States in 2021, old music accounted for 70% of the market share, while the audience for new music has shrunk significantly. Literature is also in decline, realism is dead, “voice” in literature is dead and “gesture” has taken its place. Voice means a kind of communication with readers, society and human beings, while literary posture is just a tone of voice, a fashion item for politically correct and popular discourse.
  Although this is a long-standing trend in the era of information overflow, AI writing will undoubtedly exacerbate this.
paste of knowledge

  If humanity is ultimately destroyed, will it be by nuclear disaster, intercontinental missiles, or pathogens? Was it destroyed by climate disaster or the omnic revolution?
  In March this year, when ChatGPT became popular, Matthew Kirshenbaum, a professor of digital research at the University of Maryland, published an article in the media saying, is it possible that the real way to destroy us starts with words?
  Our relationship with words is undergoing fundamental changes. Big language models drive other machines and programs to publish text endlessly. The number of simple, plain, and unembellished words is unimaginable, sweeping across like a tsunami.
  Whether for advertising, political or ideological purposes, or just plain pranks, if people post billions of automatically generated messages every day, they flood the Internet, get mixed up with search results, spread across social media platforms, and infiltrate Encyclopedia entries also provide material for future machine learning systems. Then we will soon find that humans will face a “text apocalypse” – language written by machines will become the norm, while articles written by humans will become the exception. This is Kirshenbaum’s preset gray goo, also called “gray fog”, an imaginary doomsday scene.
  As early as the 1990s, William Safire, a writer who wrote about the etymology of buzzwords in media, predicted that “content” would be one of the rapidly rising categories of the Internet. He also pointed out for the first time that content can achieve its basic function—advertising revenue—without any authenticity or accuracy. Of course, there are darker ones: Artificial intelligence researcher Gary Marcus has proven that large language models can easily generate absurd and distorted narratives and serve as weapons of large-scale disinformation, widely used in fields such as geopolitical information warfare. .
  With the iteration of platforms such as Midjourney and Stable Diffusion, the generated images, videos and voices that were “fake at a glance” a year ago have become increasingly difficult to distinguish between real and fake. At present, we have seen that AI can easily rewrite an interview with a celebrity, change his mouth shape, and make him say things he never said.
  There are no longer boundaries between words, images, sounds and machines. What is certain is that mankind has entered an unprecedented era of information white noise. Our understanding of truth and accuracy has been blurred ever since. Our understanding of “understanding” also needs to be redefined.
  Artificial intelligence is intelligent because it creates coherence. It can understand what you mean, write a Romantic-style poem for you, and then translate it into English for you. But it is just a model. It does not understand Romanticism, nor poetry, nor even the signifiers and signifieds of Chinese and English words. Of course, it just doesn’t understand the way we understand it.
  Artificial intelligence is another kind of “understanding”: unfathomable decomposition and integration. This unfathomable language processing process heralds a neglected problem. History itself will become a pile of supercomputer material, and artificial intelligence will extract and generate meaning from it at will. To them, all human language and art are nothing but the paste of knowledge.
  In a world dominated by algorithmic culture, technology will bring about profound chaos and fragmentation. Where will people and their value exist?
moral imagination

  As journalists, we pride ourselves on having a focused human eye.
  Rather than automated machine writing, that kind of thoughtful long-form reporting is the model for this industry. In the wave of automated writing, we can still observe the world through human eyes and explore the texture and texture of this world.
  In contrast, large language models only obtain information from a text corpus, and the algorithm determines the order in which the next word appears. This is essentially a statistical feature that accurately predicts those seemingly shocking words and phrases. If you want AI to tell you about an accident scene, the best way is for a human to accurately describe it, sit in front of the computer, type an article, and type it in. AI decomposes and extracts to give you the answers you want.
  But I am not sure and confident. If one day artificial intelligence can access surveillance cameras, will it produce a more comprehensive and detailed description of the scene than humans, and will it have a more accurate judgment on the cause of the accident?
Establishing connections with others in creation, imagining and resonating with the lives of distant people and strangers is a part of human moral imagination that should not be deprived or actively transferred.

  Reporters and editors in the news and publishing industry are essentially knowledge producers. However, compared to the giant knowledge machines, people seem to be poor and ignorant. Compared with the efficiency of its automatic generation, when I type this article, it seems to be closer to a primitive and ancient work.
  In the era of knowledge assembly line automation, in the muddy digital torrent, we have been brought back to our original form – editors and reporters. We have lost the halo of intellectual labor and iron-shouldered morality, and are still essentially hard laborers of information and knowledge.
  Just like after the industrial revolution, automated machines entered textile factories and skilled workers lost their jobs in large numbers. “Luddism” came into being to resist this trend. Luddites are also considered ignorant, weak-legged, and obstacles to progress. In 1812, a letter from within Ludd described its mission this way: To oppose all machinery detrimental to the common good.
  ”Blood of the Machine: The Origins of the Resistance to Big Tech” published this year by American writer Brian Mossent looks back at an era when destroying machines was punishable by death. Countless Luddites, at the cost of their blood, changed their future and won better working conditions by strengthening their own strength. But from a historical perspective, the Luddite Movement did not hold back the pace of industrialization. Today, it’s even more overwhelming.
  This is my unease.
  The awe is that as hard laborers of knowledge, writers may realize more clearly that exploring the complexity of people and the world from a human scale, rather than a statistical and probability theory scale, is both arduous and difficult. It should be prudent.
  AI provides knowledge that is hard to distinguish between true and false, and accuracy is a luxury. Of course, machines and programs have error correction mechanisms, and updates to data and codes will reduce their errors. But in any case, it does not provide the answer itself, what it provides is an identification of the answer.
  News or other forms of writing do not provide answers, but provide the search for answers. This contains some kind of experience, emotion, or truth, which cannot be compressed into messy and cold data.
  Returning to language and writing, there is still something that cannot be replaced by “hand-written writers”, that is, as answer seekers, we have a strong feeling that “the world lacks or enjoys something.”
  Establishing connections with others in creation, imagining and resonating with the lives of distant people and strangers is a part of human moral imagination that should not be deprived or actively transferred.

error: Content is protected !!