GoogleThe 2023 Annual Advertising Security Report was released today. Last year, more than 5.5 billion bad ads were blocked and removed, slightly more than in 2022; Additionally, 12.7 million illegal advertiser accounts were blocked and removed, almost double the number in 2022.

Duncan Lennox, Google’s vice president and general manager of ad privacy and security, said in a post on its official blog that Google aims to block bad ads before they reach Google platforms and suspend fraudulent ad accounts, or fraudulent ad accounts. Have to be suspended even before their arrival. Posted on Google Platform. Remove immediately if detected. In these areas,artificial intelligence,Aye) are improving Google’s policy enforcement.

Lennox said Google’s security team has long been using AI-powered machine learning systems to detect and block billions of bad ads before people see them. Although these machine learning models are very mature, they have historically required extensive training, often relying on hundreds of thousands or even millions of cases of infringing content.

In contrast, large language models (LLMs) can quickly review and interpret large amounts of content while capturing important nuances in the content, Lennox said. These advanced reasoning capabilities have allowed Google to make more comprehensive and accurate implementation decisions for some of the more complex policies.

For example, larger language models are better at quickly identifying new trends in financial services, identifying patterns adopted by unscrupulous individuals who abuse these trends, and distinguishing legitimate business practices from get-rich-quick scams that Google helps teams respond more intelligently. Various emerging threats.

United States(TagstoTranslate)Google(T)AI(T)Artificial Intelligence