As these tech big shots make their way to Capitol Hill to chat with lawmakers about potential AI regulations, they’re also getting grilled about the work conditions of the folks behind ChatGPT, Bard, and Bing. US lawmakers are taking a closer look at Microsoft, OpenAI, Anthropic, Meta, Alphabet, Amazon, Inflection AI, Scale AI, and IBM to shed light on the working conditions of data labelers.
These data labelers have the crucial task of labeling training data and evaluating chatbot responses to ensure that AI systems are safe and dependable. But here’s the kicker: according to a group of lawmakers including Senator Edward Markey, Elizabeth Warren, and Bernard Sanders, these millions of data workers worldwide are subjected to constant surveillance, low wages, and zero benefits. And that’s not cool. Not only does it harm the workers, but it also puts the quality of AI systems at risk. Accuracy, fairness, and data protection are all on the line here.
But it doesn’t stop there. The letter from the lawmakers also puts the spotlight on rising AI startups like Inflection AI, Scale AI, and Anthropic. These are the players shaping the AI landscape. So, the pressure is on for tech companies to step up. It’s their responsibility to ensure that these workers have safe working conditions, fair pay, protection against unfair disciplinary actions, and more transparency regarding their role in AI companies.
When it comes to data labeling, tech companies often outsource the work to staffing firms that hire workers outside of the US. Countries like Kenya, India, and the Philippines serve as hubs for these operations. The job of these workers is to train AI algorithms to “see” things. Think about self-driving cars, for example. The algorithm needs to be able to recognize pedestrians from stop signs. And who helps teach it that? Data labelers who spend hours upon hours watching videos and identifying objects in each frame. According to The Financial Times, annotating one hour of video takes a whopping eight hours. That’s dedication.
But here’s the unfortunate truth: many of these workers endure harsh conditions. In order to make ChatGPT safer, Kenyan laborers, who earn less than $2 per hour, have to label images containing explicit content, hate speech, and violence. A Time investigation revealed that workers were expected to read and label between 150 and 250 snippets of text, ranging from 100 words to over 1,000 words, during a nine-hour shift. The mental toll is real, and wellness sessions offered to these workers often don’t make much of a difference. To make matters worse, data labelers don’t receive the same benefits as the tech company employees they work for.
And guess what? Data labeling is only becoming more prevalent with the rise of generative AI products. The global data annotation and labeling market skyrocketed to $800 million last year, and experts predict it’ll hit a whopping $3.6 billion by the end of 2027. The demand for labeling services is growing as the volume of data to be labeled balloons. And as a result, data labelers are becoming specialists in different types of data, whether it’s related to driving or medical information.
In the world of AI, it’s clear that data labelers are the unsung heroes behind the scenes. It’s high time they get the respect, fair treatment, and recognition they deserve.