"Manuscripts don't burn"

We somehow made it to August but the universe is not taking a break from scandal and outrageousness because it has no sense of decency. The latest scandal has to do with language, knowledge, culture, and history — pretty much all things that make us human. It has to do with books.

The famous Bulgakov quote I’ve used in the headline (for those who haven’t read “The Master and Margarita”), refers to the enduring power of thoughts and ideas. It is certainly an impactful line. Alas, Bulgakov could not have predicted AI.

The news, as you may have seen it online: Anthropic bought and destroyed millions of books, many rare, after scanning their contents to train its large language models. But that’s actually misleading. The news is that a judge this month approved a $1.3-billion settlement between Anthropic and several thousand authors “after using pirated copies of their works to train its Claude chatbot.”

That news report raised a major stink and the internet went up in arms in defence of books. But in fact, the information that Anthropic — and more, of course — were buying, scanning, and destroying books was reported as early as January by the Washington Post, which proves that nothing is black and white. In its report, the Washington Post detailed Anthropic’s Project Panama, which ““is our effort to destructively scan all the books in the world,” an internal planning document unsealed in legal filings last week said. “We don’t want it to be known that we are working on this.”” I’m afraid this will not be my usual Friday merrymaking at the expense of other people who deserve it.

The WaPo report further states that “In a January 2023 document, one Anthropic co-founder theorized that training AI models on books could teach them “how to write well” instead of mimicking “low quality internet speak.” A 2024 email inside Meta described accessing a digital trove of books as “essential” to being competitive with its AI rivals.”

The excerpt above suggests, to me at least, two things. The first is that software companies are run by people with questionable morality and, believe me, these are not the words I wanted to use. In fact, the words I would very much like to use to describe said people make up the foundation for this here neat little swearword dictionary. The second thing is that said people (biologically speaking only) would stop at nothing to win a race whose only point is itself.

Now, in order to overcome the nausea that’s been with me since I first read the WaPo report, let me tell you about a book I translated a few years ago. It was the author’s first work and it was written as something that aimed to check boxes: racism, alternative sexuality, more racism, personal growth in the face of racist and homophobic adversity, etc. You know the sort.

That book was objectively rubbish. As a translator I got two copies. For a moment I considered leaving them on a bench in a park to get rid of them. For another moment I considered just throwing them in the thrash because why would I do this to any innocent reader? In the end I simply left the two copies on the floor in my room. I could not bring myself to throw them away. And that is because like hundreds of millions of people I was raised to respect books. This doesn’t mean I haven’t thrown books at walls out of frustration with the crap contained between the covers. But destruction? This is not an option. And, notably, it is not the scum of fiction that Big Tech is destroying physically.

Following from the above, I was immensely pleased to see the reaction of people on social media to the Anthropic news. People were furious. People were outraged. And the media jumped on the wave of outrage. This is a good thing given big news outlets’ ability to amplify stories. This power, as we’ve come to know, can be used for evil but it can also be used for good, and we are seeing the latter now.

There is even, unbelievably, a humourous twist to the story, because, per the WaPo article, “In their lawsuit, the authors alleged that Meta higher-ups considered paying for books to train their AI models but opted to instead download millions of books free from “torrent” platforms that facilitate online piracy.” In other words, the psychopathic little turds were, allegedly, too cheap to pay for the books they wanted and took to piracy instead. But at least that didn’t involve any shredding.

What that book story suggests is that, whether deliberately or not, Big Tech wants to take over ownership of the cultural heritage and collective knowledge of humanity. To what end they would then use this heritage and knowledge I wouldn’t speculate because I feel a headache coming on but it’s probably safe to say they’d try to make money out of it.

Perhaps the worst part of the story comes from this Telegraph report:

The books didn’t have to be destroyed for the scanners to read them, but their destruction strengthened Anthropic’s legal case. The judge ruled that stripping books of their contents through digital scanning, then destroying them entirely, was not a breach of copyright because the digital version “replaced” the physical one: it was seen as an act of transformation rather than duplication, and thus not against copyright law.

Woland must be laughing his head off. Me, I wish this judge a long and very interesting life.

To end on a somewhat positive note, while Anthropic and Meta were scanning books, one creative writing professor from the U.S. wrote, yes, a book. And it won an award. Not just any award, either. “Hum” won the Climate Fiction Prize, so wow and clap, please.

The plot, per Bloomberg: “Imagine this: a climate-damaged world where advertising-funded robots known as hums provide all the manual labor and constant surveillance over society. May Webb has just lost her job to AI and is now using proceeds from an experimental operation to take her family on a holiday to one of the last bits of nature left: a botanical garden.” Why would the protagonist use much needed money to take her kids to a botanical garden? Because, due to climate change, the air quality is very, very, very poor.

Meanwhile in reality, deaths from outdoor air pollution fell from 180.3 per 100,000 in 1990 to 91.8 in 2023. Particulate matter deaths? Down from 71.6 per 100,000 to 56 per 100,000 between 1990 and 2023. But sure, climate change makes air quality poor. And once upon a time fiction had to be plausible. I would strongly urge Big Tech to get their hands on that work of climate anxiety (the author said she drew inspiration from her climate anxiety, yes) and scan it, without then shredding it. Then it can use that to produce more award-winning climate fiction.

I know a lot of you approve of AI developments, noting legitimate benefits such as making things faster and more accurate than they would be if they were left to humans. However, I fail to see any benefit to AI’s takeover of arts and science. It could sure be a useful tool in science, but training LLMs to write fiction or compose music (and they do, there’s AI country on Spotify) can only result in one thing: AI slop. Soulless garbage. Botswill. Because LLMs can certainly replicate words and sounds. Creation, however, remains exclusive to humans. May you choke on your chips.

Share