Google: We are fixing strange AI summaries such as "eating glue", data gaps and user "pranks" are the main reasons

GoogleLiz Reid, head of the search department, published a long article on the 30th local time, further responding to the major factual errors given by the recent AI Overviews search.

Google: We are fixing strange AI summaries such as "eating glue", data gaps and user "pranks" are the main reasons

Liz Reid explained that strange suggestions such as "eating glue" and "eating rocks" were caused by data gaps and users' "trickery".There are also a large number of fake screenshots that have been widely circulated.Some of the fake results are obvious and silly. Others suggested we were providing dangerous results for topics such as leaving dogs in cars, smoking during pregnancy, and depression. These AI overviews never appeared. Therefore, we encourage anyone who comes across these screenshots to conduct searches themselves to verify them.”

For example, in the answer to "How many rocks should I eat?", Liz Reid said that almost no one had asked Google this question before these screenshots went viral. Some wrong suggestions came from "sarcastic or malicious" content in the forum. She said that the forum is usually an excellent source of real, first-hand information.But in some cases, it may give less helpful advice, such as gluing cheese to pizza..

Liz Reid announced that it will continue to improve its AI search function.

  • Built better detection mechanisms for meaningless queries that shouldn’t show AI overviews,and limits the inclusion of satire and humorous content.

  • Updated the system to limit the use of user-generated content in replies to avoid providing misleading suggestions.

  • Added trigger limits for queries that AI Overview cannot help effectively with.

  • Strict protection measures are set for topics such as news and healthFor example, for hard news topics where timeliness and authenticity are critical, the AI overview will not be displayed. In terms of health, Google has introduced additional trigger optimization measures to strengthen quality protection.

Liz Reid said that in addition to these improvements, they will also pay attention to feedback and external reports.And take action against a small number of AI overviews that violate content policies —— This means that the summary contains information that may be harmful, obscene or otherwise illegal.

Attached is a summary of the previous situation:

A netizen searched on Google to find a solution to the problem of "cheese and pizza not sticking together". AI gave a summary guide directly at the top, which seemed to be very practical. One of the steps was "You can also add 1/8 cup of non-toxic glue to the sauce to make it stickier."

Google: We are fixing strange AI summaries such as "eating glue", data gaps and user "pranks" are the main reasons

statement:The content is collected from various media platforms such as public websites. If the included content infringes on your rights, please contact us by email and we will deal with it as soon as possible.
Information

Apple may allow Siri to control all functions of the APP and has previously reached a cooperation with OpenAI

2024-6-1 9:37:50

Information

OpenAI has opened five major functions including networking and vision to ChatGPT free users!

2024-6-1 9:45:09

Search