"Google Showcases ‘Enterprise-Ready’ AI: More Facts, Less Fantasy for Businesses"

Vertex AI, Google's Cloud development platform, is enhancing its capabilities to combat the spread of inaccurate information in apps and services. Following the launch of the Grounding with Google Search feature, which allows models to pull live information from the internet, Google has introduced an option for clients to enrich their AI outputs using specialized third-party datasets. This service will include data from reputable providers such as Moody's, MSCI, Thomson Reuters, and ZoomInfo, with availability expected in Q3.

These improvements are part of Google's efforts to promote its "enterprise-ready" generative AI solutions by minimizing the occurrence of misleading or erroneous information. A key feature, "high-fidelity mode," enables organizations to generate outputs based on their own corporate datasets instead of relying solely on Gemini’s broader knowledge. This mode, powered by a specialized version of Gemini 1.5 Flash, is currently available for preview through Vertex AI’s Experiments tool.

Organizations can also opt for Google’s AI models to extract information from their proprietary datasets. Additionally, Google is expanding its Vector Search capabilities to support hybrid searches. This public preview update allows vector-based searches to be combined with text-based keyword searches, enhancing accuracy.

Upcoming improvements will also include a “dynamic retrieval” feature through Grounding with Google Search, which will intelligently determine whether to source information from Gemini’s established datasets or real-time queries based on user prompts that might need the latest resources.

By providing organizations with greater control over their information sources, Google aims to enhance the credibility of its AI search features, especially in light of past criticisms regarding the accuracy and relevance of AI-generated content.