Google Search "AI Mode" Goes Live: Taking Aim at Perplexity and ChatGPT

Recently, Google Search launched an experimental feature in its generative experience (SGE) called "AI Mode" (AI Mode), which is powered by the Gemini 2.0 Large model-driven, capable of handling complex queries with powerful reasoning capabilities, and directly benchmarked against Perplexity AI and OpenAI's ChatGPT Search and other similar products.

What is the "AI model"?

"AI Mode" is an experimental feature from Google Search Labs that allows users to ask complex, multi-step questions directly in Google Search and explore specific topics in depth through multiple rounds of conversations with AI.

This feature is currently available to Google One AI Premium subscribers. Users can experience it by joining Search Labs, typing a question in the search bar and clicking on the "AI Mode" tab, or by going directly to a specific link (https://www.google.com/search?udm=50) to use. Please note, however, that this feature may not be available in some areas (e.g., domestic) at this time.

The experiment is currently only available to U.S. users over the age of 18, and users are required to register via the Labs to register. If you join the waiting list for the experiment, we will notify you when the experiment is open to you.

Gemini 2.0 Drivers for Deep Exploration and Comparison

Google says "AI Mode" uses a customized version of the Gemini 2.0 model, which has advanced reasoning, thinking and multimodal capabilities, making it particularly well-suited for queries that require deep exploration and comparison. Compared to previous versions, Gemini 2.0 provides significant improvements in handling complex questions and long text, better understanding user intent and providing more accurate and comprehensive answers.

For example, you could ask "What is the difference between the sleep tracking features of a smart ring and a smart watch?". The "AI Mode" will compare the two products in detail and provide links to articles with relevant information. You can continue to ask "What happens to my heart rate during deep sleep?" to learn more. to learn more.

In the past, to compare detailed options or explore new concepts, users often had to perform multiple searches. The emergence of "AI mode" has changed this. Not only can it access web content, but it can also provide more comprehensive, in-depth answers using real-time information sources such as the knowledge graph, real-world information, and shopping data containing billions of products.

In an interview with TechCrunch, Robby Stein, VP of Search Products at Google, said, "In our testing, we found that users ask questions twice as long as traditional search and follow up about a quarter of the time.AI mode opens up more possibilities for Google search, and that's what we're excited about."

"Fan-out" technology and factual accuracy guarantees

Stein noted that with Google's introduction of AI Overviews, users want to be able to get these AI-driven responses for more searches, which is one of the reasons Google introduced "AI Mode". "AI Mode" works through a technology called "query fan-out" (query fan-out).

How does the "query fan-out" technique work?

The core of this technology is its ability to break down a complex question posed by a user into multiple related sub-questions and search multiple data sources simultaneously. It then integrates the results of these searches to form a comprehensive answer that is easy to understand.

Stein emphasizes, "The model has been trained to prioritize factual accuracy and to support its claims with verifiable information, which is very important. It also pays particular attention to sensitive areas, such as health. If the model is not confident in the response, it may simply provide a list of web links, as this is most helpful in the moment. It will do its best to provide the most useful information."

Of course, Stein is also honest: "But that doesn't mean it will never make mistakes. As with all newly launched cutting-edge AI technologies, it's likely to make mistakes."

AI Overviews Feature Upgrade

Google also announced the U.S. launch of the Gemini 2.0 model for AI Overviews, which can now assist with more difficult problems, starting with code, higher-order math, and multimodal queries. Additionally, users can access AI summaries without having to log in.

Future outlook: smarter, more personalized search experience

As an early experiment, "AI Mode" still has a lot of room for improvement, and Google says it will continue to refine the user experience and expand functionality in the future, for example, by making the experience more visual and pulling information from a variety of different sources (e.g., user-generated content) Google is training models to determine when to add hyperlinks to responses (e.g., booking tickets) or when to prioritize images or videos (e.g., for how-to queries). responses to add hyperlinks (e.g., booking tickets) or when to prioritize images or videos (e.g., how-to queries).

Overall, the launch of "AI mode" marks another important step in the direction of Google search intelligence. It not only improves the depth and breadth of search, but also brings users a more convenient and personalized search experience. With the continuous development of technology, we can expect Google Search to bring more surprises in the future.