Good content makes good AI

The phrase “Garbage In, Garbage Out,” frequently quoted in the IT industry, describes how poor input data simply results in bad output data as well. It does not matter how advanced the AI or semiconductor technologies are: if you input garbage, garbage will inevitably come out. Conversely, AI with good content becomes more powerful.

Leading global AIs, such as Google’s Gemini and OpenAI’s ChatGPT, appear to have already mastered all the information on the planet. They are absorbing all materials that can be digitized, including books, encyclopedias, news articles, and conversations and posts on social networks. There are even predictions that AI learning materials will run out by 2032.

The issue is that AI indiscriminately absorbs quality content without paying. Methods such as ‘crawling’ or ‘web scraping,’ which randomly search for information on the web, were also used in this process. These behaviors undermine the motivation of creators, even leading to giving up on creating. Some creators set up poison (toxic pixel) in their digital creations, traps that will render AI learning useless when it learns from it. This is another digital struggle in the 21st century.

The collapse of creators will be followed by the demise of AI. The void by creators will be replaced with garbage, which will be all that AI has to learn from.

The Korean government initiated research on enhancing AI-related copyright systems a year ago and released a ‘Generative AI Copyright Guide’ at the end of last year. However, these actions may be perceived as belated and insufficient. Protecting creators is not about regulating AI, but about preserving its potential. Establishing a robust ecosystem where created content and AI can complement each other is a prerequisite for becoming an AI powerhouse. Advancing this mission is crucial to prevent Korean AI from learning subpar content.