Developers can now ground their Gemini output with Google Search

Google is adding a new feature to the Gemini application programming interface (API) and AI Studio to help developers base the responses generated by artificial intelligence. The feature called Grounding with Google Search, which was announced Thursday, allows developers to compare the AI-generated responses with similar information available on the Internet. This way, developers can further refine their AI apps and provide their users with more accurate and up-to-date information. Google emphasized that such grounding methods are important for clues that generate real-time information from the Internet.

Google releases ‘Grounding With Google Search’ feature

Google AI for developers support page detailed the new feature that will be available on both the Gemini API and Google AI Studio. Both tools are largely used by developers building mobile and desktop apps with AI capabilities.

However, generating responses from AI models can often lead to hallucinations, which can negatively impact the credibility of the apps. The problem can be even worse when the app delves into current topics, requiring the latest information from the Internet. Although developers can manually fine-tune the AI ​​model, without a guiding data set, errors can still occur.

To solve this, Google offers a new way to verify the AI-generated output. This process is known as grounding and connects an AI model to verifiable sources of information. Such sources contain high quality information and add more context to the information. Some examples of these sources are documents, images, local databases and the Internet.

Grounding with Google Search uses the latest source to find verifiable information. Developers can now use the top results from Google Search to compare the information returned by the Gemini AI models. The Mountain View-based tech giant claims this exercise will improve the “accuracy, reliability and usability of AI outputs.”

The method also helps AI models exceed their knowledge cut-off date by extracting the information directly from the source. So in this case, Gemini models can obtain the latest information using the output of the search algorithm.

Google also shared an example of the difference between outputs that were grounded and outputs that were not grounded. An unfounded answer to the question “Who won the Super Bowl this year?” was “The Kansas City Chiefs won Super Bowl LVII this year (2023).”

However, after using the Grounding with Google Search feature, the refined response was: “The Kansas City Chiefs won Super Bowl LVIII this year, beating the San Francisco 49ers in overtime by a score of 25 to 22.” Notably, the feature only supports text-based output and cannot handle multimodal responses.