Google’s internal documents reveal that it plans to evaluate AI response by crowdsourcing workers within 3 minutes.
Google announced plans to launch a new feature called “Bard” Google Docs will employ human auditors who will review and evaluate the accuracy provided by the artificial intelligence (AI). The auditors will be required to review AI generated responses and make any necessary edits or changes within a time limit of three minutes. This initiative aims to ensure the AI system delivers reliable and trusted information to users in situations where accuracy is critical.
Bard represents Google’s commitment towards maintaining high-quality experiences for users and ensuring the accuracy of data provided by its AI technologies. Google wants to strike a balanced between the efficiency and speed of AI systems and the thorough scrutiny and understanding that humans can provide. The auditors’ three-minute deadline forces them to make quick evaluations and identify any errors or inaccuracies. This could improve the credibility and reliability of the information provided, which will ultimately benefit users who depend on Google Docs to provide accurate and useful answers.
- Bloomberg reports that Google’s Bard has been trained by contractors to quickly review answers.
- The documents indicating the deadlines given to contractors was as low as 3 minutes.
- Bloomberg said that the workers are under constant pressure and receive minimal training.
Google’s Bard The AI chatbot is said to be trained by thousands contractors who are under pressure to review the answers provided by it in just three minutes.
ChatGPT, Google’s competitor to OpenAI, relies on contractors from companies like Appen and Accenture who receive minimal training and are paid as little as $14 per hour. Bloomberg reported, citing several contractors. The publication said that the workers had requested anonymity.
Google first announced Bard in February after the launch of ChatGPT The company was put on high alert. OpenAI’s chatbot attracted 100 million users a month and was a direct competitor to Google Search. Microsoft had invested billions in OpenAI.
Chatbots such as Bard and ChatGPT rely on the large language models Humans are involved in the review of responses to make sure they are accurate and reliable.
It is not possible to review all the documents manually. responses for Bard Bloomberg, citing six contractors and internal documents, reported that the project has grown larger and more complex.
The instructions were published by Bloomberg and reviewed by Insider.
“As it stands right now, people are scared, stressed, underpaid, don’t know what’s going on,” One contractor told Bloomberg. “And that culture of fear is not conducive to getting the quality and the teamwork that you want out of all of us.”
The report sheds light on how seriously Google is taking the threat from OpenAIThe AI arms race is intensifying between two companies who are hoping to lead the AI rollout in the world.
Rating responses is often part of the contractor’s task. “helpfulness” Scales range from “not at all helpful” You can also find out more about the following: “extremely helpful” You can gauge the relevance of the response by looking at how recent it is.
Insider received a response from a Google spokesperson: “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 added that “”Ratings don’t have a direct impact on the output of our model and are not the only way to promote accuracy.”
“Teams across Google with specialized skill sets – from engineering, to user experience, to trust and safety experts – use a range of techniques to build these products and continuously improve their quality and accuracy,” The spokesperson said.
Insider’s requests for comment to Appen and Accenture were not immediately responded.