Google is adding a new feature to the Gemini application programming interface (API) and AI Studio to help developers ground the responses generated by artificial intelligence. Announced on Thursday, the feature dubbed Grounding with Google Search will allow developers to check the AI-generated responses against similar information available on the Internet. This way developers will be able to further fine-tune their AI apps and offer their users more accurate and up-to-date information. Google highlighted that such grounding methods are important for prompts that generate real-time information from the web.
Google Releases ‘Grounding With Google Search’ Feature
Google AI for Developers support page detailed the new feature which will be available on both the Gemini API as well as the Google AI Studio. Both of these tools are largely used by developers who are building mobile and desktop apps with AI capabilities.
However, generating responses from AI models can often result in hallucinations, which can negatively impact the credibility of the apps. The problem can be even more significant when the app delves into topics of current affairs, where the latest information from the web is required. While developers can manually fine-tune the AI model, without a guiding dataset, errors can still exist.
To solve this, Google is offering a new way to verify the output generated by AI. Known as grounding, this process 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 include documents, images, local databases, and the Internet.
Grounding with Google Search uses the last source to find verifiable information. Developers can now use top results from Google Search to compare the information returned by the Gemini AI models. The Mountain View-based tech giant claims that this exercise will improve the “accuracy, reliability, and usefulness of AI outputs.”
The method also helps AI models surpass their knowledge cut-off date by sourcing the information directly from the grounding source. So, in this case, Gemini models can get the latest information using the Search algorithm’s output.
Google also shared an example of the difference in outputs that were grounded vs those which were not grounded. An ungrounded response to the query “Who won the Super Bowl this year?” was “The Kansas City Chiefs won Super Bowl LVII this year (2023).”
However, after the Grounding with Google Search feature was used, the refined response was, “The Kansas City Chiefs won Super Bowl LVIII this year, defeating the San Francisco 49ers in overtime with a score of 25 to 22.” Notably, the feature only supports text-based outputs and cannot process multimodal responses.
