If you use Google regularly, you may have noticed the new AI Overviews that provide concise answers to some of questions . If you use social media, you may have come across many examples of AI Overviews answers that are hilariously or even dangerously wrong.
See also: How to disable Google AI Overviews
Real bugs can occur in existing LLM chatbots, of course. However, the potential damage that can be caused by AI is multiplied when those bugs appear at the top of the extremely valuable web real estate of the Google search results page.
“The examples we’ve seen are generally very unusual queries and don’t represent experiences most people’s,” a Google spokesperson said. “The vast majority of AI insights provide high-quality information, with links to dig deeper on the web.”
See also: Google: Web search and an opt-out button from AI
After reviewing dozens of examples of errors in Google's AI Overviews, we noticed a few broad categories of errors that seemed to recur over and over again. These could be used as examples for the company to improve as the system continues to develop.
See also: Google Chrome: Fourth zero-day patched in a month
Google AI develops machine learning technologies that enable systems to learn and improve from data. These technologies include deep learning algorithms used in applications such as image recognition and natural language processing. Natural language processing (NLP) is another key technology developed by Google AI. This technology allows systems to understand and produce human language, improving the capabilities of digital assistants and search engines. Google AI also invests in the development of computer vision technologies. These technologies allow systems to recognize and interpret visual information from the world, such as images and videos, and are used in applications such as autonomous driving and medical diagnosis.
Source: arstechnica
