Why That Chatbot Is So Good at Imitating Bart Simpson
This is Atlantic Intelligence, a newsletter in which our writers help you wrap your mind around artificial intelligence and a new machine age. Did someone forward you this newsletter? Sign up here.
Earlier this week, The Atlantic published a new investigation by Alex Reisner into the data that are being used without permission to train generative-AI programs. In this case, dialogue from tens of thousands of movies and TV shows has been harvested by companies such as Apple, Anthropic, Meta, and Nvidia to develop large language models (or LLMs).
The data have a strange provenance: Rather than being pulled from scripts or books, the dialogue is taken from subtitle files that have been extracted from DVDs, Blu-ray discs, and internet streams. “Though this may seem like a strange source for AI-training data, subtitles are valuable because they’re a raw form of written dialogue,” Reisner writes. “They contain the rhythms and styles of spoken conversation and allow tech companies to expand generative AI’s repertoire beyond academic texts, journalism, and novels, all of which have also been used to train these programs.”
Perhaps it no longer comes as a major shock that creative humans are having their work ripped off to train machines that threaten to replace them. But evidence demonstrating exactly what data have been used, and for what purposes, is hard to come by, thanks to the secretive nature of these tech companies. “Now, at least, we know a bit more about who is caught in the machinery,” Reisner writes. “What will the world decide they are owed?”
There’s No Longer Any Doubt That Hollywood Writing Is Powering AI
By Alex Reisner
For as long as generative-AI chatbots have been on the internet, Hollywood writers have wondered if their work has been used to train them. The chatbots are remarkably fluent with movie references, and companies seem to be training them on all available sources. One screenwriter recently told me he’s seen generative AI reproduce close imitations of The Godfather and the 1980s TV show Alf, but he had no way to prove that a program had been trained on such material.
I can now say with absolute confidence that many AI systems have been trained on TV and film writers’ work. Not just on The Godfather and Alf, but on more than 53,000 other movies and 85,000 other TV episodes: Dialogue from all of it is included in an AI-training data set that has been used by Apple, Anthropic, Meta, Nvidia, Salesforce, Bloomberg, and other companies. I recently downloaded this data set, which I saw referenced in papers about the development of various large language models (or LLMs). It includes writing from every film nominated for Best Picture from 1950 to 2016, at least 616 episodes of The Simpsons, 170 episodes of Seinfeld, 45 episodes of Twin Peaks, and every episode of The Wire, The Sopranos, and Breaking Bad. It even includes prewritten “live” dialogue from Golden Globes and Academy Awards broadcasts. If a chatbot can mimic a crime-show mobster or a sitcom alien—or, more pressingly, if it can piece together whole shows that might otherwise require a room of writers—data like this are part of the reason why.
What to Read Next
- “What I found in a database Meta uses to train generative AI”: “Nobel-winning authors, Dungeons and Dragons, Christian literature, and erotica all serve as datapoints for the machine,” Alex Reisner wrote in an earlier investigation for The Atlantic.
- AI’s fingerprints were all over the election: “But deepfakes and disinformation weren’t the main issues,” Matteo Wong writes.