# Stunning! AI “Creativity” Is Highly Predictable, Researchers Find Source: https://www.youtube.com/watch?v=Pyg-fsaia3A Recap page: https://rapidrecap.app/video/Pyg-fsaia3A Generated: 2025-08-11T15:32:33.646+00:00 --- ## Quick Overview AI models are highly predictable and lack true originality, according to research that suggests their creativity is largely based on modifying existing data rather than generating novel ideas. **Key Points:** - AI models demonstrate high quantity and fluency in idea generation but often lack originality, primarily modifying existing data. - Research suggests AI creativity is highly predictable, with some studies showing LLMs outperform humans in certain creative tasks like elaboration. - Noam Chomsky critiques AI output as "high-tech plagiarism," while others view AI's rapid idea modification as a form of creativity. - LLMs can generate many ideas quickly but tend to recycle or slightly alter concepts, unlike humans who may produce more novel ideas at a slower pace. - A study found that LLMs' collective creativity is equivalent to 8-10 humans when questioned multiple times. - The video contrasts these findings with the idea that AI creativity is ultimately just modification, lacking genuine novelty. ![Screenshot at 00:02: The video's title card poses the question "Is AI Creative?", featuring the OpenAI logo and setting the central theme of the discussion.](https://ss.rapidrecap.app/screens/Pyg-fsaia3A/00-00-02.png) **Context:** The video discusses the ongoing debate about whether Artificial Intelligence (AI), particularly Large Language Models (LLMs), can be considered truly creative. It references research and expert opinions to explore the nature of AI-generated content and its comparison to human creativity, touching upon concepts like divergent thinking, originality, and predictability. ## Detailed Analysis The video explores the concept of creativity in AI, referencing research that suggests current AI models, particularly large language models (LLMs), excel at generating a high quantity of ideas but often lack true originality. It highlights that AI creativity can be highly predictable and relies on modifying or recombining existing data. Sabine Hossenfelder discusses two studies: one suggesting AI is 90% predictable and another indicating LLMs are more creative than humans on divergent thinking tasks, particularly in elaboration and fluency, but fall short in originality. The video contrasts the views of Noam Chomsky, who calls AI "high-tech plagiarism," with those who see AI's ability to modify and generate ideas quickly as a form of creativity. The presenter also showcases a "Hoverpen" product as an example of good design, contrasting it with the limitations of AI creativity. ### AI Creativity vs. Human Creativity - AI models excel in generating a high quantity of ideas and are proficient in elaboration and fluency, but often lack originality and novelty, relying on modification of existing data. ### Research Findings - Studies indicate AI creativity is highly predictable, with some research suggesting LLMs are more creative than humans on divergent thinking tasks in certain aspects, but they fall short in originality. ### Expert Opinions - Noam Chomsky views AI output as "high-tech plagiarism," while others, like Sam Altman, see AI's ability to modify and generate ideas rapidly as a form of creativity. ### AI's Strengths and Weaknesses - AI can generate many ideas quickly, but often recycles or slightly modifies existing concepts, whereas humans, though slower, may produce more genuinely novel ideas. ### Product Showcase - The Hoverpen is presented as an example of good design, highlighting its magnetic base and refillable nature, contrasting with the limitations of AI creativity. ![Screenshot at 00:02: Title card asking "Is AI Creative?" with the OpenAI logo.](https://ss.rapidrecap.app/screens/Pyg-fsaia3A/00-00-02.png) ![Screenshot at 00:11: Noam Chomsky is shown with a speech bubble stating "High-tech plagiarism".](https://ss.rapidrecap.app/screens/Pyg-fsaia3A/00-00-11.png) ![Screenshot at 00:36: A research paper abstract is shown, detailing a theory of creativity in convolutional diffusion models.](https://ss.rapidrecap.app/screens/Pyg-fsaia3A/00-00-36.png) ![Screenshot at 01:03: An illustration of a woman holding a knife with a red circle around it, questioning AI creativity.](https://ss.rapidrecap.app/screens/Pyg-fsaia3A/00-01-03.png) ![Screenshot at 01:23: A chart displays "Creativity score" with bars for Fluency, Flexibility, Originality, and Elaboration.](https://ss.rapidrecap.app/screens/Pyg-fsaia3A/00-01-23.png) ![Screenshot at 02:11: Screenshot of a "Scientific Reports" paper titled "The current state of artificial intelligence generative language models is more creative than humans on divergent thinking tasks".](https://ss.rapidrecap.app/screens/Pyg-fsaia3A/00-02-11.png) ![Screenshot at 03:51: Highlighting text from a study stating that when questioned 10 times, an LLM's collective creativity is equivalent to 8-10 humans.](https://ss.rapidrecap.app/screens/Pyg-fsaia3A/00-03-51.png) ![Screenshot at 04:41: An AI-generated image of a woman in a witch hat with a unicorn, with the word "creative?" superimposed.](https://ss.rapidrecap.app/screens/Pyg-fsaia3A/00-04-41.png) ![Screenshot at 05:09: Images of Noam Chomsky and a man \(likely referring to a TechCrunch author\) are displayed, representing different viewpoints on AI creativity.](https://ss.rapidrecap.app/screens/Pyg-fsaia3A/00-05-09.png) ![Screenshot at 06:06: Close-up shots of the "Hoverpen" in silver and black, showcasing its design and functionality.](https://ss.rapidrecap.app/screens/Pyg-fsaia3A/00-06-06.png)