6thWave: AI News Hub

Google Cloud has significantly enhanced Vertex AI’s grounding capabilities to improve the accuracy and reliability of AI responses. A key feature is the dynamic retrieval for Grounding with Google Search, allowing the Gemini language model to decide when to use Google Search or its own knowledge, balancing cost and response quality. For instance, Gemini uses Search for current information like movie releases but relies on its database for static info like world capitals. A new “high-fidelity” mode for grounding is also in trial for precision-critical industries like healthcare and finance. Moreover, Google will soon enable AI grounding with third-party datasets from providers like Moody’s and Thomson Reuters, bolstering response accuracy. Enterprises can also use Vertex AI Search and Retrieval Augmented Generation (RAG) APIs for customized AI models grounded in private data. These innovations aim to reduce AI hallucinations and enhance the trustworthiness of generative AI by anchoring it in reliable data sources.

Source.