Google Inc. (GOOGL) announced yesterday that it has launched a new feature, Grounding with Google Search, in its Google AI Studio and Gemini API. This feature allows users to verify AI generated answer content through Google search, thereby solving the problem of information lag in large language models (LLMs) when processing the latest events.
Including OpenAI, Anthropic, and Google's own large language models, they all face a common challenge: due to the time constraints of training datasets, these models perform poorly in answering questions related to recent events. The launch of Grounding with Google Search function is precisely to make up for this deficiency.
The Grounding with Google Search feature aims to address the issue of insufficient response in large language models (LLMs) when dealing with the latest events. By combining real-time search, developers can obtain more accurate and timely information, enhancing the reliability and practicality of AI applications.
Google's official press release pointed out that this feature can reduce the illusion generated by AI, which is to provide reality based information and ensure that users receive accurate facts. At the same time, it can also help the model obtain the latest information, making it more relevant in various scenarios.
This feature also enhances the transparency of AI applications by providing support links. Users can click on links to access more relevant information, thereby increasing the credibility and richness of the information.
The advantages of this feature are:
Reduce Illusion: The new feature ensures that AI applications provide users with more accurate facts by providing reality based information.
Get the latest information: Combined with Google search, the model can access information in real-time, making it more relevant in various scenarios.
Enhancing credibility: By providing support links, the transparency of AI applications has been enhanced, encouraging users to click to view more information.
Rich information content: The information extracted from Google search can provide more detailed background for many queries.
However, although the "integration with Google search" feature can significantly improve the accuracy of information, its usage cost is relatively high. Developers need to pay a fee of $35 for every 1000 basic queries.
The field of artificial intelligence search engines is heating up, with AI startups, chatgpt producers OpenAI, AI startups Perplexity, Apple (AAPL), Microsoft (MSFT), and Google all vying for dominance in this rapidly growing market.
Overall, the launch of Grounding with Google Search is an important innovation for Google in the field of AI. It not only addresses the shortcomings of large language models in handling the latest events, but also provides developers with more powerful and flexible tools to build and optimize their AI applications. At the same time, this feature also helps Google resist the challenge of other companies' artificial intelligence search engines to its dominant position in the search market.