Google advertises "enterprise-ready" artificial intelligence with more facts and less tall tales

Vertex AI, the Google Cloud development platform that enables companies to build services using machine learning and Google's big language models, is getting new features that will help prevent misleading information from appearing in apps and services. Following the launch of Vertex AI Grounding with Google Search in May, which allows models to retrieve actionable information from the web, Google announced that customers will now be able to improve the performance of their services with specialized third-party datasets.

Google says the service will use data from providers such as Moody's, MSCI, Thomson Reuters and ZoomInfo, and its support for third-party datasets will be available in the "third quarter of this year." It's one of several new features Google is developing to encourage organizations to adopt its "enterprise-ready" generative AI, reducing the frequency of models spitting out misleading or inaccurate information.

Another mode is "high-fidelity mode," which allows organizations to derive information for generated results from their own corporate datasets rather than from Gemini's extensive knowledge bank. The high-fidelity mode runs on a customized version of Gemini 1.5 Flash and is already available in a preview version through Vertex AI's Experiments tool.

Organizations can also allow Google's artificial intelligence models to extract information from the company's own datasets.

Vector search, which allows users to find images by linking to similar graphics, is also being extended to support hybrid search. The update is available in the public preview and allows vector search to be paired with text-based keyword searches to improve accuracy. Working with Google Search will soon also provide a "dynamic search" feature that automatically selects whether to pull information from Gemini's installed databases or from Google Search for queries that may require frequently updated resources.

The ability to control where Google's AI models get their information from could help improve the reputation Google's AI-powered search features have gained so far. After striking a deal to access Reddit data for AI training in February, Google's AI Reviews feature was mocked for making outlandish recommendations based on old Reddit posts - like adding Elmer's glue to pizza.