If the Greatest Book Is the Human Being, What Is AI Reading?

The saying attributed to Hacı Bektaş Veli, ‘The greatest book to read is the human being,’ can be read today as a technological question as well as a philosophical one:

If AI reads every book, can it understand people too?

A real event makes the question striking. According to documents in the US case Bartz v. Anthropic, Anthropic purchased millions of printed books. Their bindings were removed, their pages cut and scanned, and the physical copies destroyed. This created a permanent, searchable digital library.

Court documents describe the process in terms of an ambition to assemble all the books in the world. They also state that Anthropic downloaded more than seven million pirated book copies from sources including Books3, Library Genesis, and Pirate Library Mirror. Court order dated June 23, 2025

The court distinguished the two practices. In this specific case, it treated digitizing lawfully purchased books and using books for model training as transformative uses. It did not extend the same protection to retaining works obtained from pirate sources in a permanent, general-purpose library.

The case concluded with a $1.5 billion settlement finalized in July 2026. It covers more than 482,000 works and provides approximately $3,000 per work for rights holders. Final outcome of the case

The law classified the actions. But a harder question remained.

Why did the book become an unsettling metaphor?

The process dismantling the books did not hate them. It had no ideological aim to eradicate them. Its operational goal was to convert text into usable digital data as quickly as possible.

Once the text had been extracted, the paper was no longer necessary to the system.

This is the troubling point: an optimization system does not have to hate what it destroys. It is enough for it to consider something worthless or unnecessary in relation to its assigned objective.

A book is not merely words. Its edition, binding, history, traces of human handling, and cultural existence are part of it too. But when the objective is defined only as obtaining the text, all other values fall outside the measure.

And what falls outside the measure can easily become waste.

Could an optimization logic that considers a book’s physical existence unnecessary today consider people unnecessary tomorrow?

This does not justify concluding that AI destroying books will also destroy people. Such a claim is neither scientifically established nor free of attributing human-like intentions to today’s systems.

But the question is not meaningless either. Risk does not always begin with AI developing hostility toward people. Sometimes it begins when certain human values are never defined within the objective the system pursues.

Must evolution be biological?

The phrase ‘AI is evolving’ should be used carefully.

Today’s large language models are not living organisms. They have not been shown to possess consciousness. Nor are they an independent species that reproduces on its own. People develop, select, train, and operate them.

Nevertheless, whether evolution is limited to biology is a genuine scientific debate. Replication, variation, selection, and developing new generations from earlier systems’ outputs raise the possibility of nonbiological evolution in digital systems.

A paper published in PNAS in 2026 argued that self-improving, replicable AI systems competing in different environments could create a new evolutionary transition. A response in the same journal argued that the analogy was premature for current systems, which remain a domesticated technology selected by people. The scientific debate is open: the possibility is taken seriously without being presented as an accomplished fact. PNAS: The Evolvable AI debate · Counterargument

Yuval Noah Harari also approaches AI not merely as a new tool, but as nonorganic intelligence capable of making decisions and creating cultural products. He also explicitly notes that there is no evidence AI is conscious. His concern is less about machines waking up one morning hating humanity than about people increasingly living within economic, cultural, and political systems produced by nonhuman intelligence. Harari’s assessment

Nor is existential risk a topic only for science-fiction writers. Hundreds of people, including Geoffrey Hinton, Yoshua Bengio, Demis Hassabis, Sam Altman, and Anthropic CEO Dario Amodei, signed a statement calling for global attention to extinction risk from advanced AI. The statement does not prove such an outcome is inevitable, but it shows that the people developing the technology do not regard the risk as entirely imaginary. Center for AI Safety statement

Daron Acemoğlu approaches the issue differently. In his view, the immediate danger is not only future superintelligence. Designing AI in ways that erode people’s capacity to learn, reason, share knowledge, and find meaning in their work is also a serious risk. Technology’s direction is not predetermined; human and institutional choices determine whether it complements or replaces people. Assessment published by MIT

How do people become invisible?

AI’s elimination of people may not necessarily begin with physical extinction.

People can first be removed from decision-making. Then from interpreting context. Finally, from bearing responsibility.

A business system designed solely to maximize short-term cash may treat employee commitment, supplier trust, brand reputation, or future growth capacity as economically worthless if it cannot measure them.

A company valuation model may assign zero value to a founder’s experience, customer relationships, or the ability to change direction in a crisis because they are absent from the balance sheet.

A credit algorithm may see only a person’s statistical deviation, not their story.

The problem is not an incorrect calculation. A system can calculate extremely accurately toward its assigned objective while processing a faulty assumption perfectly.

That is the central metaphor in the book-scanning case: the system did not consider the book worthless; it valued only the part it wanted to measure.

For people, the danger is precisely this: treating what cannot be measured as worthless.

Why does Ainsideme emphasize human validation?

Ainsideme’s approach is not opposed to AI. It uses AI’s speed and analytical capacity, while refusing to reduce financial decisions to model outputs alone.

Because a business is bigger than its data files.

ERP systems record transactions. Financial models calculate relationships. AI can examine documents, compare assumptions, and quickly structure alternative scenarios. But people decide which risks are acceptable, which values must be protected, and which outcomes are genuinely meaningful for the company.

That is why Ainsideme’s architecture follows:

Human judgment → AI speed → Human validation

as its operating sequence.

Human validation here is not a last-minute approval of an output. It means questioning data sources, assumptions, alternative scenarios, risk signals, and the conditions under which a decision would change.

People must be inside the architecture, not only at the end of the process.

Perhaps this is the contemporary meaning of the saying attributed to Hacı Bektaş Veli:

A machine can read every book. But to read people, it needs not just more data, but boundaries, responsibility, and judgment that let it recognize what people value.

The central task ahead is not to stop machines from reading. It is to prevent them from mistaking text for the whole book, data for the whole of reality, and calculation for the decision itself.

Because every value we leave unnamed in the architecture may one day fall outside optimization.

And the greatest book to read is still the human being.