Vertex AI, a Google Cloud development platform that enables businesses to build services using Google's machine learning and large-scale language models, is introducing new features that help prevent apps and services from pushing inaccurate information. After making Vertex AI's Grounding with Google Search feature (which allows models to fetch live information from the internet) generally available in May, Google announced that it will also give customers the option to use specialized third-party datasets to improve the service's AI results.
Google said the service will leverage data from providers including Moody's, MSCI, Thomson Reuters and ZoomInfo, with integration with third-party datasets available “in the third quarter of this year.” This is one of several new features Google is developing to encourage organizations to adopt “enterprise-ready” generative AI experiences by reducing how often models spit out misleading or inaccurate information.
The other is a “high fidelity mode,” which allows organizations to inform the generated output from their own datasets rather than Gemini's extensive knowledge bank. High fidelity mode is powered by a special version of Gemini 1.5 Flash, and is currently available as a preview via Vertex AI's Experiments tool.
Vector search, which allows you to find images by referencing similar graphics, has also been expanded to support hybrid search. This update, available in public preview, allows you to combine vector-based search with text-based keyword search for improved accuracy. Integration with Google Search will also soon provide a “dynamic search” feature that will automatically choose between retrieving information from Gemini's established dataset or Google Search for prompts that require frequently updated resources.
