5 Key Vertex AI Innovations Unveiled at Google Cloud Next

Today, Google Cloud launched its Cloud Next conference at the Mandalay Bay Convention Center in Las Vegas, unveiling significant AI-focused advancements across its cloud product portfolio.

During his keynote, CEO Thomas Kurian highlighted various innovations, encompassing new AI hardware, expanded integration of Gemini, and model upgrades.

Key Announcements from Vertex AI

As part of the discourse, Vertex AI, Google’s flagship platform for developing, training, and deploying machine learning (ML) projects, received a notable upgrade. Here are the most prominent enhancements coming to Vertex AI:

1. Extended Context Windows and Live Image Generation

Vertex AI now supports over 130 models, including Gemini, Claude 3, Gemma, Llama 2, and Mistral. Google announced the public preview of Gemini 1.5 Pro, which features context windows of up to 1 million tokens and the capability to process audio streams for cross-modal analysis. Additionally, Imagen 2 will enable the creation of four-second animated images and introduce advanced photo editing tools.

2. Search-Based Grounding for Enhanced Accuracy

To improve response accuracy, Google is introducing a new Search-based grounding feature in public preview. This technology merges foundation model outputs with up-to-date, high-quality information from Search. If this method doesn't yield satisfactory results, users can harness Retrieval Augmented Generation (RAG) to ground models with data from enterprise applications, such as Salesforce or Workday.

3. Advanced MLOps Tools for Performance Enhancement

With numerous models available, selecting the right one can be challenging for teams. Google Cloud is expanding its MLOps tools to facilitate prompt management and evaluation processes. Teams can benefit from a collaborative library of prompts that allows for side-by-side comparisons and insights into how prompt modifications influence model performance.

4. No-Code Vertex AI Agent Builder

The new Vertex AI Agent Builder allows enterprises to create and deploy generative AI agents tailored to various use cases. This tool caters to developers of all skill levels, providing a no-code console for creating agents using natural language prompts or the option to utilize open-source frameworks like LlangChain. It also supports grounding through search and proprietary enterprise data.

5. Expanded Data Residency Options

In response to increasing concerns about data sovereignty and regulatory compliance, Google announced new data residency regions. Enterprises can now store data securely in 11 additional countries, including Australia, Brazil, and Switzerland. This expansion complements the previous offerings from North America, Europe, and Asia, affording users greater control over their data's storage and accessibility.

Mark your calendars for Google Cloud Next 2024, running from April 9 to April 11.