Google Cloud Next ’24 Keynote Highlights: Generative AI Takes Center Stage

Google Cloud Next ’24 kicked off with a keynote address that showcased the company’s continued focus on generative AI (GenAI) and its applications across various industries. Here’s a summary of the key announcements:

1. AI Hypercomputer Enhancements:

  • A3 Mega VMs: General availability of A3 Mega VMs, offering double the network bandwidth per GPU compared to A3 instances.
  • NVIDIA HGX B200 and NVIDIA GB200 NVL72: Next-generation chips with enhanced networking capabilities, slated for release in early 2025.
  • TPU v5p: General availability of TPU v5p, the most powerful AI accelerator for training and inference, with 4x the compute capacity of the previous generation.
  • Hyperdisk ML: Preview of Hyperdisk ML, a next-generation block storage service optimized for AI workloads, reducing model load times by up to 11.9x and increasing per-volume throughput by over 100x compared to traditional storage solutions.
  • Dynamic Workload Scheduler: Introduction of Dynamic Workload Scheduler, a resource management and job scheduling platform, offering calendar mode for guaranteed start times and FlexStart for cost optimization, enabling efficient resource management for complex training and inference tasks.

2. Google Distributed Cloud Enhancements:

  • Expanded Functionality: NVIDIA GPUs, GKE Enterprise, OpenAI models (including Gemma and Llama), and Vector Search are now available on GDC.

3. Google Axion:

  • Custom ARM-Based CPU: Unveiling of Google Axion, the company’s first custom ARM-based CPU, offering up to 50% better performance and 60% better energy efficiency compared to current-gen x86 VMs.

4. Vertex AI Enhancements:

  • Expanded Model Portfolio: Addition of new models to Vertex AI, including Gemini 1.5 Pro, Claude 3 (Sonnet, Haiku, and Opus), and CodeGemma.
  • Grounded AI: Enhanced grounding capabilities for Vertex AI, allowing Gemini models to be grounded in Google Search results, customer data, and external databases.
  • MLOps Features: Introduction of new MLOps features, including Prompt Management, Automatic Side by Side (AutoSxS), and Rapid Evaluation.

5. Customer Agents:

  • Vertex AI Agent Builder: Introduction of Vertex AI Agent Builder, a tool for rapid creation of multi-modal AI agents that can interact with customers across various channels.
  • AI Meetings and Messaging Add-on: Preview of AI Meetings and Messaging Add-on, a Google Workspace add-on that leverages Gemini to provide real-time translation for audio and chat in Google Meet.
  • AI Security Add-on: Preview of AI Security Add-on, a tool that automatically labels Google Workspace documents based on their security level.
  • Gemini in Google Chat: Introduction of Gemini in Google Chat, a feature that summarizes conversations in Google Chat threads.

6. Creative Agents:

  • Google Vids: Unveiling of Google Vids, a generative AI tool for creating videos, including automatic video generation, script creation, and integration with personal photos and videos.
  • Imagen 2.0: General availability of Imagen 2.0, a text-to-image model.
  • Text-to-Live Image: Introduction of Text-to-Live Image, a feature that generates animated images from text descriptions.

7. Creative Agents:

  • Gemini in BigQuery: Integration of Gemini with BigQuery to support data preparation, discovery, analysis, and governance.
  • BigQuery Data Canvas: Introduction of BigQuery Data Canvas, a new notebook-like experience for BigQuery that includes natural language support and Embedding Visualization.
  • Vertex Indexing in BigQuery and AlloyDB: Introduction of Vertex Indexing in BigQuery and AlloyDB, enabling direct querying of vector indexes stored on BigQuery and AlloyDB.
  • Direct Access to Vertex AI from BigQuery: Introduction of Direct Access to Vertex AI from BigQuery, allowing direct access to Vertex AI models from BigQuery.
  • Gemini in Looker: Integration of Gemini with Looker to enable data-driven chat and Google Workspace integration.

8. Code Agents:

  • Gemini Code Assist: Introduction of Gemini Code Assist, an evolution of Duet AI for Developer, providing enhanced code assistance capabilities, including large-scale changes, private codebase adaptation, and support for multiple source code repositories.

9. Security Agents:

  • Gemini in Threat Intelligence: Integration of Gemini with Mandient’s threat intelligence data to enable chat-based threat analysis and large-scale code