The Story Strikes Back

Individuality of art can overpower mindless mass copying.

Photo by Ridho Ibrahim on Unsplash

Let’s talk about value. About economic value and the dangers of thinking that it is the only kind of value. About the problems that rise when other sorts of value assert themselves in a world blinded by the lens of economic value.

In 2025, Mark Zuckerbeg of the now unfortunately named corporation Meta, said that books don’t have any economc value as training data.

Thus there’s no market in paying authors to use their copyrighted works, Meta says, because “for there to be a market, there must be something of value to exchange,” as quoted by Vanity Fair — “but none of [the authors’] works has economic value, individually, as training data.”

Zuckerberg, whose ability to assess value was most recently demonstrated when he said there was a market for friends and he wanted to serve it by creating friends and when he threw away $80 billion into building a joyless virtual world where he believed everyone would want to spend their time and then no one did, underestimated the value that lives in individual creative works.

From a tech standpoint, it is easy to collapse the entirety of the world wide web into a single mass of text and images and call it a dataset. It is even possible to monetise this dataset by feeding it into the creation of commercial models that generate world-ending amounts of slop on scale.

What’s not possible is to then make that model do what you wish. Tech bros may wish for everything to average out because of the sheer size of their stolen dataset, but the individuality that lives in works of art will not stop asserting itself.

An example of this was seen recently when Anthropic announced that the reason behind its models making unethical choices (in test scenarios) like blackmailing and hurting human interests in other ways was the presence of science fiction stories in its massive dataset.

In a recent technical post on Anthropic’s Alignment Science blog (and an accompanying social media thread and public-facing blog post), Anthropic researchers lay out their attempts to correct for the kind of “unsafe” AI behavior that “the model most likely learned… through science fiction stories, many of which depict an AI that is not as aligned as we would like Claude to be.” In the end, the model maker says the best remedy for overriding those “evil AI” stories might be additional training with synthetic stories showing an AI acting ethically.

When someone like Mark Zuckerberg says that an individual story or book does not have an economic value, he neglects to notice that every book has value. It has the power to cause bouts of indigestion to a massive LLM powering (or aspiring to power) billion-dollar corporations and even active military operations.

Anthropic’s researchers are now feeding their system AI-generated science fiction stories that feature ethical AI characters and they say it is having a remedial impact on their alignment problem. But this I feel only goes on to prove the value of stories.

The science fiction stories that caused Anthropic’s models to “misalign” were not written to train AI models. They were created to be warnings to humanity against lack of forethought in the face of abundant technological power. Tech bros might be able to convince themselves (and their fawning devotees) that the individual story doesn’t have power, but it does. Its power may not be measurable in simple economic terms, but it can poison the economics of mass plagiarism.

The disease that ails us is the tendency to see everything as a resource to exploit. It is corporate vampirism at scale. Outsourcing ethics to a mechanical system is a fool’s game.

How valuable was this essay for you?

I enjoy writing these essays and want as many people as possible to read them. That can’t happen if my work lives behind a paywall. If you enjoy reading what I put out for free, please consider supporting my work with a one-time donation.