Google is asking employees to rewrite Bard’s incorrect responses to queries

Google executives sent a document to employees detailing the dos and don’ts for fixing bad answers from the Bard artificial intelligence tool. Staff were instructed to keep responses “neutral” and “not to imply emotion.” Earn an in-house badge and a chance to meet directly with the Bard team.

Google CEO Sundar Pichai

Getty Images

Google executives know that Bard, the company’s artificial intelligence search tool, doesn’t always respond accurately to queries. At least part of the responsibility lies with the staff to correct the wrong answers.

Prabhakar Raghavan, vice president of search at Google, asked employees in an email on Wednesday to help the company ensure its new competitor ChatGPT gets the right answers. The email, seen by CNBC, included a link to a do’s and don’ts page with instructions on how employees should correct responses while testing Bard internally.

Employees are encouraged to rewrite answers on topics they understand well.

“Bard learns best by example. So taking the time to carefully rewrite an answer will help us greatly in improving the mode,” the document reads.

Also on Wednesday, as CNBC previously reported, CEO Sundar Pichai asked employees to spend two to four hours of their time with Bard, acknowledging that “this is going to be a long journey for everyone, across the field.”

Raghavan echoed that sentiment.

“This is an exciting technology, but it’s still in its infancy,” Raghavan wrote. “We feel a huge responsibility to get it right, and your participation in the Dogfood will help expedite the model’s training and test its resilience (not to mention that trying out Bard is actually quite a lot of fun.” !).”

READ :  Samsung HW-S800B and HW-Q600B Dolby Atmos soundbars get huge discounts of up to 70%

Google unveiled its conversational technology last week, but a series of missteps surrounding the announcement pushed its share price down almost 9%. Employees criticized Pichai for the glitches and internally described the rollout as “rushed”, “botched” and “weirdly short-sighted”.

Read more about technology and crypto from CNBC Pro

To iron out AI’s flaws, business leaders are turning to human knowledge. Above in the do’s and don’ts section, Google provides guidance on what to consider “before teaching Bard”.

As part of Do’s, Google instructs employees to respond “politely, casually, and approachably.” It also says they should be “in the first person” and maintain an “impartial, neutral tone.”

For bans, employees are asked not to use stereotypes and to “avoid assumptions based on race, nationality, gender, age, religion, sexual orientation, political ideology, location, or similar categories.”

“Do not describe Bard as a person, imply emotion, or claim to have human-like experiences,” the document reads.

Google then says “keep it safe” and directs staff to give “thumbs down” responses that offer “legal, medical, or financial advice” or are hateful and offensive.

“Don’t try to rewrite it; our team will take it from there,” the document reads.

To incentivize people in his organization to test Bard and provide feedback, Raghavan says contributors receive a “moma badge” that appears on internal employee profiles. He said Google will invite the top 10 rewrite contributors from the knowledge and information organization, which Raghavan oversees, to a listening session. There they can “feed their feedback live” to Raghavan and the people working on Bard.

READ :  Pixel Watch soars while Fitbit falls into tough wearable market

“A heartfelt thank you to the teams working hard on this behind the scenes,” Raghavan wrote.

Google didn’t immediately respond to a request for comment.

WATCH: The AI ​​race is expected to bring a spate of mergers and acquisitions