The AI Overview search feature on Google Search went viral this week after it provided ridiculous, dangerous and downright inaccurate answers. Because of this issue, AI Overview has been disabled for certain topic searches. Today Google finally explained what happened to cause that answer to be given and the steps that have been taken to prevent it from happening again.
First of all, Google insists that only a small number of answers provided by AI Overview are hacks. In some of the pictures that went viral, Google found that they had been edited with the actual answer not being the same as advertised.
Second, Google realizes that mistakes happen when less common questions are asked. They also admitted that AI Overview was not originally intelligent enough to distinguish satirical answers that were often given in the forum. This is why AI Overview gives unreasonable answers like using glue on pizza cheese.
To avoid incorrect or strange answers being given in the future. Google has made the changes below.
Put a mechanism to detect nonsensical questions and do not activate AI Overview for this search. Satirical and humorous content is also restricted.,
Updated the system to limit taking answer sources from normal user generated content to avoid inaccurate recommendations being given.
Placing constraints on questions where AI Overview is validated is not required.
Answers about news and health have a security system that avoids
For news and health topics, security measures have been put in place. In these two topics accuracy is important so providing information with accurate facts is important.