For technology companies competing to be king of the AI hill, there are few things more valuable than raw, original data.
To keep the large language models underlying our favorite AI chatbots up to date, tech companies need to feed them tons of fresh input. According to one study, the amount of data used to train AI has doubled every nine months since 2010. Exponential growth can quickly hit a wall with very little clean data stored.
Once there was no more original content to steal, companies began paying workers to generate new training data and offering low-quality contracts to train AI on very specific tasks, such as calculating the weekly payroll of Broadway musicians. Others have been hired to film themselves doing degrading chores such as folding laundry or engaging in activities that are clearly adult-oriented.
Predictably, the growing workforce behind the AI boom started cutting corners all at once. other The AI chatbot provides data to feed the AI chatbot. talk to new scientistmany insiders said that AI cannibalism (a method that experts have long warned could destabilize LLM) is surprisingly common.
“It’s very widespread,” said a worker who gave her name as Alice. news science. “I think they care because every company I’ve ever worked for has clear guidelines about it and they’re clearly trying to get people. But I don’t think they can stop it.”
In other words, AI companies are learning an ironic lesson. After plagiarizing the content of others to create products that threaten jobs across the economy, the new precariat they create uses the same technology to perform still-necessary human tasks in the laziest way possible.
While employees need to be careful not to be too obvious, Alice says it’s not hard to pass off AI-generated data as your own if you remove the annoying linguistic tics of chatbots like ChatGPT before sending it. “Only the sloppiest users get caught,” said the AI contractor. news science. “If you have some awareness of the characteristics of an AI, you can tell it not to use it in your output. At that point, what do you do?”
“If these companies want quality data, they should offer quality contracts,” Alice continued. “Instead, they hire people who are struggling, keep them on for as long as possible, and then dump them when the project ends without warning.”
other dealers said news science They use LLMs to avoid making mistakes and losing the job altogether.
“I was scared of not having a source of income, but then it became easier to run everything through the LLM,” one person explained. “A lot of the projects I do now are creating scenarios, so I use one LLM to create the scenarios, and then use another LLM to create the files that follow the scenarios. I feel guilty, but like I said, it was more about making sure I didn’t make mistakes in the beginning.”
Whatever the reason, it’s clear that employees are willing to give AI companies a taste of their failures. This is a situation that could have dramatic consequences for the entire AI race.
Learn more about AI: Police accused of using AI to falsify evidence
