Google touts “enterprise-ready” AI with more facts and less make-believe

Vertex AI, the Google Cloud development platform that allows companies to build services using Google’s machine learning and large language models, is getting new capabilities to help prevent apps and services from pushing inaccurate information. After rolling out general availability for Vertex AI’s Grounding with Google Search feature in May — which enables models to retrieve live information from the internet — Google has now announced that customers will also have the option to improve their services’ AI results with specialized third-party datasets.

Google says the service will utilize data from providers like Moody’s, MSCI, Thomson Reuters, and ZoomInfo and that grounding with third-party datasets will be available in “Q3 this year.” This is one of several new features that Google is developing to encourage organizations to adopt its “enterprise-ready” generative AI experiences by reducing how often models spit out misleading or inaccurate information.

Another is “high-fidelity mode,” which enables organizations to source information for generated outputs from their own corporate datasets instead of Gemini’s wider knowledge bank. High-fidelity mode is powered by a specialized version of Gemini 1.5 Flash and is available now in preview via Vertex AI’s Experiments tool.

Vector Search, which allows users to find images by referencing similar graphics, is also being expanded to support hybrid search. The update is available in public preview and allows those vector-based searches to be paired with text-based keyword searches to improve accuracy. Grounding with Google Search will soon also provide a “dynamic retrieval” feature that automatically selects if information should be sourced from Gemini’s established datasets or Google Search for prompts that may require frequently updated resources.