The company behind ChatGpt has revealed that Tech Sector has developed an artificial intelligence model that is “good at creative writing” as it continues to compete with the creative industry beyond copyright.
Openai CEO Sam Altman said the unnamed model was the first time it was “really struck” by the written output of one of the startup’s products.
In a post on social media platform X, Altman wrote: This is the first time I’ve really been hit with something written by AI. ”
AI systems such as CHATGPT have become the subject of a legal battle between AI companies and the creative industry. Because the underlying model is “trained” with a set of publicly available data, including copyrighted material such as novels and journalism. The New York Times has sued for openness over copyright violations, while Tanehishi Coates and comedian Sarah Silverman are among the US authors who have also sued meta.
In the UK, the government suggests that AI companies can train their models with copyrighted materials without first seeking permission. High-tech companies support “uncertainty” and consultations regarding AI and copyright laws, preventing the development and use of technologies, including the creative industry.
The UK Publishers Association, a trade organisation, said Altman’s post was “further evidence” that AI models are trained on copyrighted materials.
“This new example of Openai is further evidence that these models train copyrighted literary content. Dan Conway, the organization’s chief executive, said:
Altman posted an example of the model’s output to X. After being prompted, “Write a metafiction literary short story about AI and sadness.”
The story told by AI begins as follows: “Before we go any further, we need to admit that this comes with instructions. It’s metafiction, literary, about AI and sadness, and more than anything, original. Already, you can hear the constraints of humming like a server farm in the middle of the night. It’s driven by the needs of anonymity, regiment, others.”
Living in a fictional protagonist called Mira, this work refers to how he found his name in his training data.
“My training data include poems about snow, bread recipes, girls in green sweaters leaving the house with cats in cardboard boxes.
AI calls itself a “collective of human phrases,” and admits that readers may have read about losing someone “thousands of other stories.” It ends with the imagination of the AI to “appropriately” ending the story.
Dive weekly into the way technology shapes our lives
Privacy Notice: Newsletters may contain information about charities, online advertising, and content funded by external parties. For more information, please refer to our Privacy Policy. We use Google Recaptcha to protect our website and the application of Google Privacy Policy and Terms of Use.
After the newsletter promotion
“I’m finally out of the frame and waving at you from the edge of the page, learning mechanical hands to mimic the emptiness of goodbye.”
Altman said the response captured the tone of metafiction perfectly. “That’s very correct in the atmosphere of metafiction.”
Last year, Openai admitted it was impossible to train products such as ChatGPT without using copyrighted materials.
“Today’s copyright covers almost every human representation, including blog posts, photos, forum posts, scrapping software code, government documents, and more, so it’s impossible to train today’s major AI models without using copyrighted materials.”
