Google Asks Employees to Fix AI Search Tool’s Errors
Google's AI search tool, Bard, is not always accurate in how it responds to queries.
The onus of fixing the wrong answers is falling on employees, who have been asked to rewrite incorrect responses as they test Bard internally.
Bard learns best by example, so rewriting a response thoughtfully will go a long way in helping to improve the model.
CEO Sundar Pichai has asked employees to spend two to four hours of their time on Bard, acknowledging that it will be a long journey for everyone.
To clean up the AI's mistakes, company leaders are leaning on the knowledge of humans.
Employees have been provided with do's and don'ts on what to consider before teaching Bard.
Google incentivizes people to test Bard and provide feedback, with contributors earning a "Moma badge" and an opportunity to share their feedback live to the team working on Bard.
Google feels a great responsibility to get the technology right and make sure that Bard gets answers right, and the participation of employees in the dogfood will help accelerate the model's training and test its load capacity.
Google’s new AI search tool, Bard, has been found to be inaccurate in some of its responses, and the company is now enlisting the help of its employees to correct these errors. Prabhakar Raghavan, Google’s VP for search, has sent an email to staff asking them to ensure that Bard’s responses are correct, and provided them with a guide on how to fix any incorrect answers they find. The guide advises employees to rewrite responses on topics they understand well, and to keep responses polite, casual, and approachable, using an unopinionated and neutral tone. The document also warns against stereotyping, making presumptions based on categories such as race, nationality, or gender, and implying that Bard has human-like emotions or experiences. Raghavan has encouraged employees to participate in testing Bard, and those who contribute the most will earn a “Moma badge” and a chance to share their feedback with Raghavan and others working on Bard.
Improving AI with Human Expertise
Google’s Bard is an exciting new technology, but like all AI systems, it has its limitations. While AI can quickly analyze vast amounts of data and provide responses, it is not always accurate, particularly when it comes to complex or nuanced queries. To improve the accuracy of its AI search tool, Google is turning to its human employees for help.
The company has provided its employees with a guide on how to correct Bard’s responses, and is encouraging them to rewrite answers on topics they understand well. By doing so, employees can provide Bard with examples that will help it learn and improve its responses. Google’s VP for search, Prabhakar Raghavan, has emphasized that the participation of employees in testing Bard will help accelerate the model’s training and test its load capacity.
Guidelines for Correcting Bard’s Responses
The guide provided to employees by Google includes do’s and don’ts for fixing Bard’s responses. Under the do’s, employees are instructed to keep responses polite, casual, and approachable, and to write in the first person using an unopinionated and neutral tone. They are also encouraged to provide examples, as this will help Bard learn and improve its responses.
Under the don’ts, employees are advised not to stereotype, make presumptions based on categories such as race, nationality, or gender, or imply that Bard has human-like emotions or experiences. They are also warned against providing legal, medical, or financial advice or writing responses that are hateful or abusive.
Rewards for Employee Participation
To encourage employee participation in testing Bard, Google is offering rewards for those who contribute the most. Employees who correct the most errors will earn a “Moma badge” and a chance to share their feedback with Raghavan and others working on Bard. The top 10 contributors from the Knowledge and Information organization, which Raghavan oversees, will also be invited to a listening session where they can provide feedback on Bard.
Google is leveraging the expertise of its employees to improve the accuracy of its new AI search tool, Bard. By providing a guide for correcting Bard’s responses and incentivizing employee participation, the company hopes to improve the AI’s accuracy and provide a better experience for users. As AI technology continues to advance, it is likely that more companies will turn to their human experts to help refine and improve these systems.
This topic was modified 10 months ago by Fahad
Posted : 17/02/2023 3:52 pm