Say Goodbye to AI Errors: Google’s New Vertex AI Features!

Google introduces new features in Vertex AI. Let’s take a look and be surprised at what we find!

Google Enhances Vertex AI with Trusted Data

Google Cloud’s Vertex AI platform is boosting accuracy by adding datasets from reputable third-party providers like Moody’s, MSCI, Thomson Reuters, and ZoomInfo. This enhancement, available in Q3 2024, follows the recent release of Vertex AI’s Grounding with Google Search feature, which allows models to access live information from the internet. These updates aim to reduce inaccuracies in AI-generated outputs, making Google’s generative AI tools more reliable for enterprise use.

New Features for Improved AI Results

Vertex AI’s new “high-fidelity mode” enables organisations to use their own datasets for AI-generated outputs, enhancing reliability. This mode, powered by Gemini 1.5 Flash, is currently in preview. Additionally, Google is expanding its Vector Search capability to support hybrid searches, combining vector-based and text-based keyword searches for better accuracy. Grounding with Google Search will also introduce a “dynamic retrieval” feature to automatically decide whether to use Gemini’s datasets or live Google Search data, ensuring up-to-date information.

Addressing AI Accuracy Issues

These enhancements aim to improve the reputation of Google’s AI-based search features, which have faced criticism for inaccuracies. By offering more control over data sources, Google hopes to mitigate past issues, such as the infamous AI Overviews feature that made bizarre recommendations based on outdated Reddit posts. These updates are part of Google’s broader strategy to promote its AI tools as “enterprise-ready” by reducing misinformation and increasing reliability.

Read the full article on The Verge.

*An AI tool was used to add an extra layer to the editing process for this story.