Why I read: Between AI, history and languages

By: Mohammed Sahir

The recent buzz about AI companies buying books in bulk and using them to feed their LLMs has generated a debate regarding the excessive dependency of humans on stored information in AI models. What is more concerning about this is not that books are used as a feeder for AI models, but that, in the process, books are being destroyed or are not remaining in such a shape that they can be used for further studies, research and academics, particularly some rare niche books. Thus, in the long run, making humans excessively dependent on AI.

LLM and destructive scanning

At its basics, a Large Language Model (LLM) works on the principle of pattern matching, which means that such engines are engineered in such a way that they have to predict the next word or piece of a word given the nature of the preceding text; they can achieve human-level coherence at advanced levels.

To ensure that such models are expanded with an abundant level of information, it becomes necessary that LLM models be fed with nearly every piece of information known to the human race.

Since books cannot be quickly flipped through and scanned tightly by high-speed scanners, it becomes necessary to break the spine of a book, loosen the pages, and drop them into high-speed Optical Character Recognition (OCR), thus rendering a book futile for future use.

Paradox of information and interpretation

It seems a triumph of modern engineering that the sum of human knowledge is available at a keystroke. Yet therein lies the trap: an abundance of easily summoned facts does not guarantee discernment.

Because LLMs are turned to provide concise, authoritative answers tailored to frequently asked queries by users, they strip away the layered nuance, historical friction and interpretative doubts essential for genuine understanding.

If such conditions persist and no mechanism is implemented to manage and regulate AI models, soon we might be living in the nightmare of 1984. The only difference will be that we will be volunteering for thought policing, believing that the information on our mobile screens is the absolute truth.

To avoid such a situation from arising, the way out lies in an international regulatory mechanism and the strengthening of the fields of Social Sciences. The more we see the expansion of AI, the more we need good Historians, Sociologists, Anthropologists, Economists, etc. AI might provide us with information, but to comprehend it, interpret it, and use it in the right way only comes from books written by humans.

Writer is a student of Economics.