A heist for the ages ...

Image by Drew Angerer via Getty / Futurism

This pix of Zuck, depicted as a borrower of books extreme, kinda indirectly shows how the third luckiest man on earth does business. It's never enough, it's akin to cutting down all the trees in a forest in order to build houses nobody will live in because the potential for money overrides all, in this case, data, as LLMs needs this now rapidly dwindling resource in order to evolve. This need for data also shows, in indirect fashion, how inefficient the physical construct of AI truly is as digital is used to crunch analog, the data stream used by neural nets to interact with the real world in real time as researchers seem to be fixated on that model to enable AI to evolve as needs warrant. This is not sustainable but you already knew that, right?

The training compute of notable large models in FLOPs vs publication date over the period 2010-2024. For overall notable models (top left), frontier models (top right), top language models (bottom left) and top models within leading companies (bottom right). The majority of these models are language models.

For all the revolutionary change artificial intelligence promises, it also makes lofty demands. For starters, AI is extraordinarily power hungry. Generating all the electricity that AI datacenters consume takes forest-loads of energy, not to mention hardware and cooling infrastructure. That stuff all costs a lot, making AI a huge money pit. That's had a big effect on our economy, as the tiniest bit of AI hype can send huge shockwaves through Wall Street and beyond.

But AI's also greedy in less noticeable ways: namely, for your data.

The large language models (LLMs) that underpin products like OpenAI's ChatGPT, for instance, need to devour enormous datasets of written words to fine tune an algorithm to follow the rules of language. They're so hungry for raw data, in fact, that original material for these algorithms to gobble up is becoming hard to come by.

"We’re literally running out of text in the universe to train these systems on," said computer science scholar Stuart Russell back in 2023. Now in 2025, the well is all but drying up.

Meta, the parent company of Facebook and Instagram, has inadvertently pulled the curtain back on what it looks like to ingest all that data.

In January, Meta lost a huge fight with a group of authors who sued the company for using their books to train its AI. The case uncovered the fact that Meta had illegally downloaded an infamous pirate library, LibGen, to procure millions of legally protected texts. Those books were then fed to Meta's LLM, Llama, after software engineers got approval from the Zuck himself. In other words, one of the largest companies in the world didn't even bother to pay for a single copy of each book it used to build its AI.

This week, The Atlantic compiled a search engine that could trawl the LibGen files and uncover which books, exactly, were scraped by Meta. The scope of Meta's data harvesting operation is extensive, spanning over 7.5 million books and some 81 million academic papers, on top of work published by museums, architects, and artists.

It's never enough is it.

Aperture