Sora’s Imagination: Bridging Text and Video with AI
Video taken from openai.com, Open AI’s Sora model’s landing page
Open AI, the maker of Chat GPT, has launched a new Generative AI model called Sora which can instantly generate videos from text. The video that I have shared here is one such video generated by Sora. The quality of the videos is far superior to the ones generated by the previous models launched by the other tech giants. It is capable of generating videos in different resolutions and aspect ratios. It is not publicly available yet.
While these videos appear more polished than those generated by earlier models, closer examination reveals occasional errors, such as discontinuity between frames and surreal scenes in realistic videos. You might see strange things like cats with three legs or a person’s arm disappearing into nowhere.
You might notice a certain oddness about them.These videos don’t exhibit the careful attention to detail and artistic flair that comes with human expertise in scripting, cinematography, and film production. Rather, they’re generated by a machine piecing together knowledge from a vast pool of data. As someone immersed in the world of technology, I’m always eager to see advancements and encourage collaboration between humans and machines. However, the belief that AI could undermine creativity seems unfounded.
I keep thinking about the definition and essence of the word creativity. Isn’t creativity about original ideas? These models draw their creative inspiration from actual past work. They learn from what’s been fed to them, which helps them come up with creative stuff. But what if humans stop making new creative things? Then the models would start learning from the things they make themselves, which could lead to a problem called “conceptual collapse”. If AI systems primarily rely on existing data and patterns to generate new content, there might be a lack of truly original or groundbreaking ideas. Instead, the output could become repetitive or constrained by the limitations of the training data. This could potentially limit innovation and creativity, leading to a cycle of recycled ideas rather than genuine novelty.
It can be argued that humans too draw inspiration from the work of others, nevertheless there is a seed of originality from which great work blossoms. I’m a neuroscience enthusiast and have read some interesting articles about the neuroscience behind creativity. According to some neuroscience studies, creativity springs from three main brain networks: the Executive Attention Network for focus, the Imagination Network for imaginative thinking, and the Salience Network for managing transitions between them. First, the Imagination Network generates fresh ideas by processing emotions and senses. Then, the Salience Network shifts focus to the Executive Attention Network for polishing. How well these networks connect and work together predicts creative ability. Creativity is highly subjective and different for every individual and is shaped by the person’s genetic makeup, her environment and experiences.
While machines can process immense amounts of data and generate permutations and combinations, they struggle to grasp the depth and nuance of subjective human experiences. They often lack the human touch of creativity. Human ingenuity is enriched by emotion, intuition, and personal experiences, which are challenging for machines to replicate.
So, I feel that while it’s possible to admire the output of machines and recognise their usefulness in time-sensitive situations with limited creative scope, describing their output as “creative” might be stretching the definition. Here’s where I think that creative people don’t need to worry about AI taking over their jobs. AI will definitely help and work alongside humans, but I don’t think it’ll replace the special touch that comes from creative minds.
By Rajashree Rajadhyax
Co-Founder, Cere Labs
Comments
Post a Comment