Google's Response to AI Search Errors Improving and Taking actions
TapTechNews on May 31st reports that Liz Reid, head of Google's search department, released a long article on the 30th local time to further respond to the significant factual errors in the AIOverviews search that occurred recently.
Liz Reid explained that strange suggestions such as 'eating glue' and 'eating stones' are caused by data vacancies and users' 'pranks'. 'Also, a large number of forged screenshots have been widely spread, some of which are obviously and very stupid forged results. Others imply that we have provided dangerous results for topics such as leaving dogs in the car, pregnant women smoking and depression. These AI overviews have never occurred. Therefore, we encourage anyone who encounters these screenshots to conduct a search in person for verification.'
For answers like 'how many stones should I eat', Liz Reid said that before these screenshots became widespread, almost no one had asked Google this question. Some wrong suggestions come from'satirical or malicious' content in the forum. She said that forums are usually an excellent source of real, first-hand information, but in some cases, may give less useful advice, such as gluing cheese to pizza with glue.
Liz Reid announced that it will continue to improve its AI search function.
Built a better detection mechanism for meaningless queries for which artificial intelligence overviews should not be displayed and restricted the inclusion of satirical and humorous content.
Updated the system and restricted the use of user-generated content in responses to avoid giving misleading advice.
Increased the triggering restrictions for queries for which AI overviews cannot provide effective help.
Set strict safeguards for topics such as news and health. For example, for hard news topics where timeliness and authenticity are crucial, artificial intelligence overviews will not be displayed. In terms of health, Google has also launched additional triggering optimization measures to enhance quality protection.
Liz Reid said that in addition to these improvements, it will always pay attention to feedback and external reports, and take actions against a small number of AI overviews that violate the content policy - which means that the overview contains information that may be harmful, obscene or other violations.
TapTechNews attached the previous background:
A netizen searched on Google to find a solution to the problem that 'cheese and pizza don't stick together'. The AI directly gave a summary guide at the top, seemingly decent, and one of the steps was 'You can also add 1/8 cup of non-toxic glue to the sauce to make it stickier.'
Related reading:
'Google Responds to AI Search Suggesting Netizens to Eat Stones, Poisonous Mushrooms and Other Major Factual Error Contents'
'Google's AI Search Causes Big Trouble: Suggesting Netizens to Eat Stones, Poisonous Mushrooms, and Add Glue to Pizza'