Google contract workers training ChatGPT rival Bard say they are underpaid and overworked

By Divyanshi SharmaIn January this year, reports had surfaced that tech giant Google was losing sleep over OpenAI’s ChatGPT, the AI chatbot that went viral a couple of days after its launch in November 2022. And just next month, Google introduced its own AI chatbot, Bard, with ChatGPT-like features. However, the world didn’t get to experience Bard until May this year, after the Google I/O event. Even though Bard was available to select users before May, most of us got to use Google’s AI chatbot in May. And a lot of people began sharing their experiences on social media. More recently, Bard got some major updates and it looks like with time, Google’s AI chatbot will get even better, as promised by CEO Sundar Pichai.

However, the people who are responsible for training Bard aren’t really happy with their working conditions. Several contract workers came forward and told Bloomberg that they are ‘overworked, underpaid and stressed’ while reviewing Bard’s answers.

Google contract workers claim to be underpaid, stressed

A Bloomberg report reveals that generative AI tools like Bard are trained using LLMs. However, their answers are reviewed by actual human beings whose primary responsibility is to provide feedback on the AI chatbot’s answers, check them for biases, flag any mistakes, etc. Google is using contract workers to carry out this task and thousands of workers are working to check and improve Bard’s response. But, these workers aren’t really happy with the working conditions.

The report cites six contract workers who say that since the company entered into a competitive race with OpenAI, the ‘size of their workload and the complexity of their tasks increased’. Without any proper training, the contract workers were asked to check answers in subjects ranging from medicine to law. The publication also reviewed documents in which ‘convoluted instructions’ were given to the workers. The report also says that workers were given extremely tight deadlines, as little as three minutes, to review Bard’s responses.

“As it stands right now, people are scared, stressed, underpaid and don’t know what’s going on,” one of the contractors told the publication and added, “And that culture of fear is not conducive to getting the quality and the teamwork that you want out of all of us.”

In addition to this, a Google contract worker also warned Congress via a letter in May that Bard could become a ‘faulty and dangerous product’ as the contractors are asked to review content under such tight deadlines.

The report also added that these contract workers as paid as little as USD 14 an hour for their labour.

Google spokesperson’s statement

Meanwhile, a Google spokesperson told Business Insider that contract workers are just one of the ways that the company tests Bard’s responses. They rely on other ways too.

“Connecting people to high-quality information is core to our mission. We undertake extensive work to build our AI products responsibly, including rigorous testing, training, and feedback processes we’ve honed for years to emphasize factuality and reduce biases. Human evaluation – from individuals internal and external to Google – is one of many approaches we use to improve our products,” the spokesperson said and added, “Ratings don’t directly impact the output of our models and they are by no means the only way we promote accuracy.”