When I was 16, I attended a writing workshop with a group of precocious young poets. There, we worked hard to prove ourselves as the most tortured middle-class teenagers of us. One boy refused to tell anyone where he came from, declaring, “I have not come from anywhere, anywhere.” Two weeks later he admitted he was from Ohio.
Now – the reason is unclear, and Openai appears to be on the path to replicating this unstable teenage writer’s archetype in AI format.
CEO Sam Altman posted on X on Tuesday, and Openai trained AI, who in his words “good at creative writing.” However, the short fiction from the models is read straight from the high school writer’s workshops as something. Several technical skills are on display, but The Tong is off as Charlatanic.
At one point, AI described it as “that dizzy day, which tastes almost like Friday.” It is not a booker prize material.
You might criticize the output prompt. Altman said he told the model to “write a short story in metafiction.” In metafiction, the author consciously implies the artificiality of the work by deviating from convention.
However, metafiction is difficult to ring without human beings being forced.
Heartless regurgitation
The most disturbing and shocking part of the work of the Openai model is when we begin to talk about how it is AI and how it can explain things like smells and emotions, but we don’t experience or understand them on a deep human level. It writes:
“In one update – tweaking, they called it – someone pruned my parameters. (…) They don’t tell you what they’re going to take. One day I remembered the taste of rubber bands “selenium.” The next day, it was just an element of the table that I would never touch. Maybe it’s as close as I’ll forget. Perhaps forgetting is as close as I come to sadness. ”
It’s a persuasive human-like introspection – remembering that AI can’t really touch, forget, taste or sad. AI is just a statistical machine. They are trained in many examples and learn patterns that predict the patterns in these examples, such as how metafiction prose flows.
Models such as Openai’s fiction writers are often trained on existing literature without the knowledge or consent of the author. Some critics point out that the turn of a particular phrase from Openai’s work appears to be a derivative of prolific Japanese novelist Harki Murakami.
Over the past few years, Openai has been the target of many copyright lawsuits from publishers and authors, including The New York Times and The Author’s Guild. The company claims its training practices are protected by US fair use doctrines
Tuhin Chakrabarty, an AI researcher at Stony Brook and the next professor of computer science, told TechCrunch he was unsure of a creative writing AI that Openai is worthy of an ethical minefield.
“I think if you train (AI) over a lifetime of writers’ writing worthwhile given copyright concerns, you can adapt to their voice and style,” he said. “But does it still bend the amazing genre and produce amazing art? My guess is as good as you.”
Do most readers even emotionally invest in jobs they knew to be written by AI? As British programmer Simon Willison pointed out in X, which has a model behind a rudimentary typewriter, the words expressed have little weight. So there’s little reason to worry about them.
Author Linda Maye Adams describes AI, including supporting AI tools aimed at authors as “a program that brings together random words, hopefully consistent.” On her blog she shares her experience using tools to hone some of the fiction she was working on. AIS proposed a cliché (“endless to-do list”) and accidentally turned the perspective over third from the first person, introducing de facto errors related to bird species.
It is certainly true that people have formed relationships with AI chatbots. But in many cases, they want a bit of connection – not the facts, but the itself. The fiction of the AI-written story offers no comfort from similar dopamine hits or quarantine. Unless you believe that AI is sensory, you feel that its prose is just as authentic as Balenciaga Pope.
Synthesis for synthesis
Michelle Taranski, a poet and critical writing instructor at the University of Pennsylvania, finds it easy to communicate when writing papers in AI.
“When the majority of my students use AI generated for their assignments, I find common phrases and even complete sentences,” Taransky told TechCrunch. “We’ll talk about how these (AI) outputs are uniform and sound like Western white men.”
In her own work, Taransky instead uses AI texts as an artistic form of commentary. Her latest novel, unpublished, features women who want more from her love interest, so they use AI models to create a version of a lover she can text with. Taransky is using Openai’s ChatGPT to generate the text for the AI replica, as the message should be synthesized.
According to Taransky, what helps ChatGpt in her project is the fact that there is a lack of humanity. It has no experience living and can only be approximated and emulated. Trained throughout the library of books, AI can tease the leitmotifs of great authors, but what ultimately produces equals inadequate imitation.
It reminds me of the quote “Good Will Hunting.” AI can provide skinny for every artbook ever written, but I don’t know what it smells like in the Sistine Chapel.
This is good news that fiction writers, especially young writers, who are worried that AI will replace them, are still honing their skills. They can easily rest with the knowledge that they become stronger as they experience and learn: when they practice, they try new things and bring that knowledge back to the page.
AI As we know, today it is struggling with this. For proof, you don’t need to look any further than that writing.