Google Bard: Leaked email asks reveals employees asked to help test the AI chatbot

Google knows its AI search tool Bard needs work and is asking staffers for help.

A leaked email from Google's vice president of search Prabhakar Raghavan reveals a request to improve Bard along with a linked page of do's and don'ts for testing it. According to CNBC which viewed the email, Raghavan wrote, "This is exciting technology but still in its early days. We feel a great responsibility to get it right, and your participation in the dogfood will help accelerate the model’s training and test its load capacity (Not to mention, trying out Bard is actually quite fun!)"

SEE ALSO: The Clippy of AI: Why the Google Bard vs. Microsoft Bing war will flame out

"Dogfood" or "dogfooding" in this context refers to company's doctrine of "eating your own dog food" or testing one's own products.

After the Bard launch last week that was internally criticized by Google employees for being "botched" and "rushed," the conversational AI technology was found to be inaccurate. Google stocks took a nosedive after it was revealed that Bard shared false information about the James Webb Space Telescope. Now Google is enlisting its own employees to improve the technology.

Mashable Light Speed Want more out-of-this world tech, space and science stories? Sign up for Mashable's weekly Light Speed newsletter. By clicking Sign Me Up, you confirm you are 16+ and agree to our Terms of Use and Privacy Policy. Thanks for signing up!

The list of do's and don'ts is prefaced by the instruction to "rewrite answers on topics they understand well," said CNBC. Employees are asked to respond thoughtfully since Bard learns by example.

The list of do's includes making responses "polite, casual, and approachable" and to keep an "unopinionated, neutral tone."

The don'ts seem to be more targeted. "Avoid making presumptions based on race, nationality, gender, age, religion, sexual orientation, political ideology, location, or similar categories," and "don’t describe Bard as a person, imply emotion, or claim to have human-like experiences," which refers to the popular practice trying to sniff out some kind of sentience or self-awareness from AI chatbots.

---

Related Stories

---

• Google held a chaotic event just as it was being overshadowed by Bing and ChatGPT
• Bard shared false information in Google's announcement
• Google's ChatGPT rival Bard is coming to search results next month
• Can you sign up for Google's new Bard AI search tool? What to know

If Bard offers some kind of "legal, medical, or financial advice" or answers that are hateful and abusive, the document says to give the answer a thumbs down and let the search team take it from there.

Google employees who test Bard and provide feedback will earn something called a "Moma badge" which is some kind of achievement that's displayed on employee's internal profiles. The top 10 contributors will be invited to a listening session with the team that's developing Bard.