Recent events confirm his words. For example, the AI development company Anthropic agreed to pay $1.5 billion to authors whose works were used without permission. A lawsuit was filed against Anthropic for using a shadow library of 500,000 books that was collected illegally. This library was intended to create accessibility to literature; however, Anthropic used it to train its models. Court documents indicate that despite the downloads, the books were not used in their training.
The judge ruled that the company was liable for creating unauthorized copies, regardless of actual use. Each author whose books were affected was awarded $3,000 for each work.
The court administrators in the case of "Bartz et al. v. Anthropic" provided a list of affected works on a special website where anyone can check if their book is included in the library. However, as is often the case with class action lawsuits, authors rarely receive significant amounts, as most of the money goes to legal fees. In this case, only half of the amount goes to the authors, with the rest going to the publishers, and the chances of receiving compensation for textbooks are even lower.
As an author, I checked the list myself and found that four out of five of my books published in New York are in this library. I am proud that my ideas can be accessible to millions of people thanks to AI. At the same time, I feel sorry for the authors whose works did not make it into this database.
Anthropic claims that it did not use this library to train AI; however, the question of the legitimacy of using digitized books remains open. For example, Google digitizes books for search, showing only small fragments. Can the same copy be used for training AI? The court ruling suggests that this is possible if the books were obtained legally. Anthropic was fined not for training AI, but for having copies for which no payment was made.
This is just the beginning of many questions regarding copyright and its application to new technologies. Rights to text copies do not work in the context of AI training, which does not require storing information, only reading it. Perhaps we need new licenses and intellectual property rights to regulate this process. But the question of rights is just part of the problem. A new audience of readers has emerged—artificial intelligence.
Over time, we will accumulate experience in AI training. The selection of training materials will become a key aspect in making decisions about how we will use AI. Some users will want AI to be trained on materials that align with their views. For example, conservatives may prefer AI to provide answers consistent with their beliefs, while liberals will insist on training on liberal data. However, for most, it is important to get the best answer, regardless of political bias.
500,000 books is an impressive number for training, but there are already millions of works in the world that AI cannot read due to unclear copyright or language. AI training is a process that is still ongoing. Creating a database for inspiration will become an art. Ultimately, AI will be able to read all of humanity's works, and having only 500,000 books in its arsenal will soon seem insufficient. That is why authors should strive to have their works included in AI training.
Young people, quickly mastering AI capabilities, are increasingly trusting its assessments. I myself have noticed that I have stopped doubting calculators and Google, and now I find myself trusting AI's answers. It is becoming the new arbiter of truth.
AI agents not only answer questions but also assist in searching for and understanding information. If AI does not know about something, it can be considered that it does not exist. Authors who have avoided working with AI will find it difficult to succeed. Some creators are not represented in digital format at all, and their works remain unnoticed. As Tim O'Reilly says, the problem for authors in the modern world is not piracy, but obscurity.
If AI becomes the arbiter of truth, and its training depends on the quality of materials, I want my ideas to matter in this process. What author does not dream of this? I want my impact to reach billions of people turning to AI. Perhaps I am willing to pay for this or do everything possible to have my ideas integrated into AI.
We are now forming an environment where the audience of readers, especially in popular science literature, is shifting towards AI. Writers must remember that their works are now primarily aimed at AI, which will study every detail of the text. AI will connect your text with other works and determine its significance in a global context, something no human can do.
The success of including text in AI will depend on the quality of presentation. The easier it is for AI to analyze a book, the greater its influence. As a result, authors will optimize their works for AI perception, possibly collaborating with developers. The concept of "AI-friendly" text will become popular, and AI will facilitate this.
Every piece of art adds its value to culture. Libraries and the internet accumulate knowledge without interruption. AI will continue to evolve, and we do not know how far it will go. For now, its influence is only increasing, and people have the opportunity to impact AI right now.
The value of a work will not only lie in its sale but also in how deeply it enters the knowledge of new intellectual systems. This will become the true legacy of the author.
Source
