Impact of AI to literature | Galina Yuzefovich | TEDxEleftheriaSquare

The Poem "Fields of Gold" and AI Authorship

• A poem titled "Fields of Gold" is presented as an example, describing a scene with flowers, a bumblebee, and a gentle dance of life, which is considered nice but not a masterpiece (16s).
• The poem's style and language suggest that the author is likely a woman, not too young, and possibly old-fashioned, with similarities to Victorian European poets like William Wordsworth or Walter Landor (51s).
• The poem's tone and imagery evoke a sense of warmth and a kind-hearted amateur author, possibly a cat owner, which could influence the reader's perception of the poem (1m38s).
• However, it is revealed that the poem was actually created by a deep learning neural network, which raises questions about the reader's emotional response and trust in the author's identity (2m7s).

Impact of AI on Literary Criticism and Writing

• The speaker shares a personal experience as a literary critic and academic writing teacher, where students began using AI for their essays, leading to skepticism and difficulty in trusting written pieces (3m36s).
• The speaker emphasizes the importance of understanding why humans feel confused or deceived when learning that a piece of literature was created by AI, as it affects trust in human creativity and literature as a whole (2m58s).
• The potential consequences of not addressing this issue include losing touch with literature, connection with human creativity, and ultimately, a part of human nature, as literature is inherent to human culture (4m56s).

The Case of RI Kudan and the Akutagawa Prize

• A Japanese writer, RI Kudan, was awarded the prestigious Akutagawa Prize, but confessed that 11% of her award-winning novel, "Tokyo Tower of Compassion," was written by a neural network, sparking a huge scandal (5m48s).
• Many people who bought her book wanted their money back, and critics insisted that the prize should be revoked, but the judges ultimately decided not to change their decision (6m13s).
• The judges' decision was likely due to their lack of experience and strong opinion on the matter, as well as the absence of a legal procedure for dealing with AI-generated content (6m53s).
• The incident highlights the difficulty in shaping an attitude towards AI-generated literature, as people feel that something is wrong, but can't quite put their finger on what it is (7m26s).

Indie Writers and AI Assistance

• Indie writing has become a popular option for writers who can't find publishers or literary agents, with many using platforms like Amazon Kindle Direct to self-publish their work (7m46s).
• Indie writers often need to select a specific niche, such as young adult mermaid dramatic fantasy, and produce content quickly to keep their audience engaged (8m32s).
• The pressure to produce content quickly has made Indie writers perfect clients for AI services, which can assist with tasks such as adding descriptions, expanding or shortening text, and suggesting plot twists (9m24s).
• About 52% of Indie writers currently admit to using AI in their work, and prefer to call themselves "assisted writers" (10m9s).
• AI is currently unable to create a book by itself, but can provide valuable assistance to writers, including recreating an author's style (9m39s).
• A neural network called Sudowrite was given a text lacking individuality and asked to recreate a passage in the style of Ernest Hemingway, resulting in a passage that mimicked Hemingway's style, including the use of typical words and literary devices such as inversion (11m11s).
• The recreated passage demonstrates the potential of AI to assist writers in producing content that resembles their own style, which can be valuable for writers who need to produce content quickly and consistently (11m54s).
• However, many readers are hesitant to support writers who use AI assistance, with 48% of indie writers not admitting to using AI assistance, and many readers believing that they can create similar content themselves (12m21s).
• The most common criticism of AI-assisted writing is that readers don't want to pay for content that was created with the help of a machine, rather than a human (12m35s).

Readers' Perceptions and the Author's Identity

• When asked what makes a perfect book, readers typically respond that it should provide a fresh experience, provoke thought, and exceed expectations, but few mention the importance of the author's personality or character (13m25s).
• The obsession with an author's figure is puzzling, given that many famous writers are not well-known as individuals, and some, such as Shakespeare, may not have even existed as we think they did (14m21s).
• In fact, many classic works of literature, such as the Iliad and Odyssey, were likely written by multiple authors over time, and the concept of a single author is not always clear-cut (15m4s).
• A survey found that 32% of readers don't even remember the names of the authors of books they liked, highlighting the disconnect between the importance of the author's identity and the reader's experience of the book (15m40s).
• The question of why we are reluctant to accept AI as an author or co-author is raised, highlighting the need to find the border of our tolerance for AI participation in the writing process (15m57s).

Book Recommendations and AI's Role

• Literature is not just about writers, but also about readers, and readers need to find out about new books from somewhere, with 61% of Goodreads visitors saying they don't trust recommendations generated by AI (17m24s).
• However, 20-45% of reviews on Amazon.com are AI-generated, and many people find them valuable, relevant, and helpful, indicating that we do rely on AI in supporting our reading patterns (17m42s).
• People find out about new books mainly through recommendations from people they know personally (64%), followed by recommendations on platforms like Goodreads (14%), and advice from bloggers and opinion leaders (21%) (18m37s).
• The influence of opinion leaders is illustrated by the example of Victoria Beckham, who named her daughter Harper after Harper Lee, causing the book "To Kill a Mockingbird" to become a bestseller, even though Beckham later admitted she hadn't read the book (19m12s).
• The difference between trusting a person who hasn't read a book and a well-trained neural network is questioned, highlighting the need to reflect on this issue to avoid a future where artificial writers produce artificial literature that is reviewed, recommended, and consumed by artificial readers (20m2s).

The Future of AI in Literature

• The need to answer questions about the role of AI in literature is emphasized, with predictions ranging from 15 years to a more optimistic timeline, before AI-generated novels become a reality (21m1s).
• In the future, it is predicted that an AI will be able to write a novel on its own that is not only indistinguishable from a human-generated novel but also impressive and possibly great (21m12s).
• Before this point is reached, it is necessary to clarify the relationship with literature in order to keep it as a significant part of culture (21m33s).
• Reflecting on and giving thought to this question is crucial at the present time (21m49s).
• If managed correctly, there will be a chance to answer the most important question concerning literature, which is whether literature is inherent to any type of thinking entity (22m8s).
• This will also provide insight into whether a new type of intelligence, specifically artificial intelligence, would want to write novels without being asked (22m21s).