In a video ad Google posted to Twitter, its yet-to-launch AI chat board Bard confidently spread misinformation about the James Webb Space Telescope. “JWST has captured the first-ever images of a planet outside of our own solar system,” the chatbot replied, which is obviously false. (It was the European Southern Observatory’s Very Large Telescope that first captured images of exoplanets.) Now the tech giant is trying to improve Bard’s accuracy and is asking its staff for help, according to CNBC.
Prabhakar Raghavan, vice president of search at Google, reportedly emailed staff asking them to rewrite Bard Answers on topics they know well. The chatbot “learns best by example,” Raghavan said, and training it with factual answers will help improve its accuracy. Raghavan also included a list of “dos” and “don’ts” when it comes to correcting Bard’s responses, based on the email CNBC saw.
Responses should be first-person, impartial and neutral, and have a polite, easy-going, and approachable tone. Employees are also instructed to “avoid assumptions based on race, nationality, gender, age, religion, sexual orientation, political ideology, location or similar categories.” You are asked not to describe Bard as a person, imply that she has emotions, or claim that she has human-like experiences. They are also instructed to reject any responses from the chatbot that contain “legal, medical or financial advice” or are hateful and offensive.
Raghavan’s memo came after Google CEO Sundar Pichai emailed employees asking them to spend a few hours each week testing the AI chatbot. Google employees reportedly criticized Pichai for a “rushed” and “botched” launch of Bard. The CEO is now giving employees the opportunity to “help shape things [the chatbot] and make a contribution” by testing the company’s new product. He also recalled that some of Google’s “most successful products weren’t first to market” and that they “gained momentum because they met key user needs and were based on deep technical insights.”
People have been expecting Google’s answer to ChatGPT since the OpenAI chatbot launched late last year. The Microsoft-backed technology has surged in popularity over the past few months, enough to unsettle Alphabet and its investors. Google tried to allay investor concerns during its quarterly conference call in early February by discussing its own chatbot and its work on developing AI-powered search to compete with next-gen Bing.
All products recommended by Engadget are selected by our editorial team independently from our parent company. Some of our stories contain affiliate links. If you buy something through one of these links, we may receive an affiliate commission. All prices are correct at time of publication.