Google is adding a new feature to the Gemini Application Programming Interface (API) and AI Studio to help developers identify AI-generated responses. The feature announced Thursday, called Grounding with Google Search, will allow developers to check AI-generated responses against similar information available online. This way, developers will be able to improve their AI applications and provide their users with more accurate and up-to-date information. Google has highlighted that these grounding methods are important for claims that generate real-time information from the web.
Google launches “Grounding with Google Search” feature.
Google Artificial Intelligence for Developers Support page He detailed the new feature that will be available on both the Gemini API as well as Google AI Studio. These two tools are largely used by developers who build mobile and desktop applications with AI capabilities.
However, generating responses from AI models can often lead to hallucinations, which can negatively impact the credibility of applications. The issue can become more significant when the application delves into current affairs topics, where the latest information from the web is required. While developers can fine-tune the AI model manually, without a guiding data set, errors may still exist.
To solve this problem, Google is introducing a new way to verify AI-generated output. This process, known as grounding, links the AI model to verifiable sources of information. These sources contain high-quality information and add more context to the information. Some examples of these sources include documents, images, local databases, and the Internet.
Grounding with Google search uses the last source to find verifiable information. Developers can now use the top results from a Google search to compare the information displayed by Gemini AI models. The Mountain View-based tech giant claims that this exercise will improve “the accuracy, reliability and usefulness of AI output.”
This method also helps AI models go beyond the final history of knowledge by obtaining information directly from the primary source. Therefore, in this case, Gemini models can obtain the latest information using the outputs of the search algorithm.
Google also shared an example of the difference in outputs that are grounded versus those that are not. A baseless response to the question “Who won the Super Bowl this year?” It was “The Kansas City Chiefs won Super Bowl LVII this year (2023).”
However, after using the Grounding with Google Search feature, the revised response was “The Kansas City Chiefs won Super Bowl LVIII this year, defeating the San Francisco 49ers in overtime by a score of 25 to 22.” It is worth noting that the feature only supports text output and cannot handle multimedia responses.
