Thousands of ‘overworked, underpaid’ humans train Google’s AI to seem smart, report reveals

Thousands of ‘overworked, underpaid’ humans train Google’s AI to seem smart, report reveals
———————————
An intriguing report from The Guardian reveals the hidden “shadow workforce” of human contractors who are essential to the development and moderation of Google’s AI models, including its Gemini chatbot. These workers, who are often described as overworked and underpaid, are tasked with a variety of roles that are critical to the AI’s functionality and safety.

The article highlights the difficult conditions faced by these raters, who are contracted through firms like GlobalLogic. Many were hired for seemingly benign roles, only to discover their jobs involved content moderation and exposure to highly distressing or explicit material without prior warning or mental health support.

The report also details a growing disillusionment among these workers, who have expressed concerns over tightening deadlines and a perceived loosening of safety guidelines. For instance, the article notes a policy change that now allows AI to repeat hate speech, racial slurs, and explicit material if it is part of a user’s input, a development that many raters find deeply concerning. As a result, many of these essential workers have grown to distrust the very AI products they help build, viewing the technology not as “tech magic,” but as a product built on the backs of a largely invisible human workforce.