AI tools are fuelling a free-for-all
Anew phenomenon is catching on in the digital sphere. Users are flooding social media platforms like X, Instagram and Facebook with their portraits, family pictures and other images drawn in the style of legendary Japanese animator Hayao Miyazaki. He is the founder of the famed Ghibli Studio, which in the past four decades has created iconic animation characters. The typical style of drawing characters in light, pastel shades has come to be known as the Ghibli art. However, the art being posted online by users is not original but generated by artificial intelligence (AI) tools available for free on a trial basis. With just a few clicks, one can transform a picture into the trademark Ghibli art. Millions of users, including politicians, fashion and sports influencers, and celebrities, are using AI tools to create images in this style.
Though generative AI tools — which are used for content creation — have been around for some time, the current trend has been triggered by a new tool developed by OpenAI. Text-based or large language models (LLMs) models like ChatGPT take textual inputs and generate desired material like articles, book chapters, social media posts, etc. The latest multimodal generative AI tools can take text, images, videos, voice and music as inputs and generate desired output images and videos. Tools like Midjourney, Stable Diffusion and DALL-E can take in text as input and give output as images. For instance, one can ask DALL-E to generate a painting of a Delhi street in the style of MF Husain or Jamini Roy and it would do so within a few seconds. Then there are image-to-image tools like Lensa that can provide altered versions of the input images.
The images generated in the Ghibli style have caught the public imagination because they look cute and funny. The fun part, however, masks serious issues concerning text and image-generative AI systems. Any AI system is based on real data collected from various sources. AI tools should be ‘trained’ using massive amounts of data (text, images, music, etc.). If an LLM is asked to generate a small story in the style of Amitav Ghosh or Ruskin Bond, it would do so by digging into massive amounts of data that would have been based on the works of these authors. Much of this material is copyrighted, and AI companies have not bought the rights from authors to ‘train’ their systems. The same is happening with the works of Ghibli Studio or the paintings of masters like Husain.
AI companies say they are only using data available in the public domain, but the creative world is not buying this claim. The ‘fair use’ doctrine in copyright laws is being misused by technology companies. The existing copyright frameworks are inadequate to deal with this kind of infringement. Cases against technology companies are already in courts in North America and Europe.
The larger question that generative AI raises is about the assault on the very idea of creativity. A painting, book, cartoon character or a piece of music are all forms of creative expression by individuals. They are all cultural products, born out of individual experience and context. For example, the work of the Ghibli Studio is a window to Japanese society and culture, just as the productions of Walt Disney Studio reflect American culture.
Assigning such creative works as products of a machine or an AI system goes against the very idea of human creativity and passion. An artist may take months to finish a painting or a writer may take years to write a novel or a non-fiction book. Hundreds of hours of hard work go into the making of a cartoon or animation strip. Therefore, creators must get credit and money. AI companies are indulging in what critics say is data laundering and robbing creators of both credit and money. With every new generative AI model — built on the work of millions of creators — the valuation of AI firms rises by billions of dollars. They sell these models for thousands of dollars to users across the world, while creators get nothing. Image generators are not artists, but they are a challenge to artists.
Generative AI has already started disrupting the creative industry globally. The most obvious impact of these products is job loss for artists, illustrators and designers in the animation industry. While AI-generated images lack the depth and expression of human-created images, they are of generally acceptable quality.
By flooding the digital globe with images generated by using free versions of AI tools, like what has happened in the case of Ghibli art, technology companies want to normalise such machine-generated art. Commercial film studios and networks are accepting them purely for labour cost-saving. They are appointing a handful of AI technical directors instead of an army of animators. This trend should ring alarm bells in India as well since thousands of Indian techies are engaged in outsourced animation work for Hollywood studios. They use computer graphics and other tools, but generative AI tools can take automation to another level.
The problem has no easy solution. Fighting the legal battle over copyright is one of the options. Data protection laws and AI regulation are playing catchup with every new development. Some publishers and custodians of creative works like books and music are entering into formal agreements with AI companies to ensure revenue-sharing for allowing their books and other works to be used to ‘train’ AI systems. They feel that technology companies are already using their work for data training, so why not formalise it and take a share of the pie. Allowing the use of one’s work for data training with consent is another option, particularly when the work of individual artists is involved. Social media platforms, too, will have to be more sensitive towards allowing AI-generated images, videos and animation by users. The Ghibli trend may appear harmless and cute, but it is actually a wake-up call for us all.
Comments