Google's latest AI announcements at the Google Cloud Next event highlight a clear strategy to upgrade their AI infrastructure and capabilities, aiming to compete strongly with OpenAI and Microsoft. Here's a quick rundown of the key points from the event:
AI Infrastructure and Models:
Gemini 1.5 Pro: Google introduced the impressive Gemini 1.5 Pro model, previously available only in private preview. This model stands out with its advanced performance and ability to handle long-context understanding, processing up to 1 million tokens of information. It also supports multiple formats like audio, video, text, and code.
AI Hypercomputer: Google launched a cutting-edge AI hypercomputer designed for top efficiency in AI training and application, using Google-designed TPUs and NVIDIA GPUs.
Hardware Upgrades:
TPU v5p: The new TPU v5p pods boast four times the computing power of their predecessors, making large-scale machine learning training and applications easier.
A3 Mega Instances: These instances, powered by NVIDIA H100 Tensor Core GPUs, offer double the network bandwidth per GPU compared to older versions.
Custom Arm-Based CPU:
Google Axion Processor: Introducing Google's first custom ARM-based CPU for data centers, delivering a performance boost of up to 50% and improving energy efficiency by 60% compared to current x86-based VMs.
Vertex AI Upgrades:
Vertex AI Model Garden: This platform now hosts over 130 models, including the latest Gemini models, partner models like Claude from Anthropic, and popular open models like Llama, Gemma, and Mistral.
Gemini for Google Workspace: Google Workspace products such as Gmail, Docs, and Sheets now feature AI-powered tools to boost productivity and teamwork.
AI-Enhanced Creative Tools:
Google Vids: A brand-new AI-powered app designed for video creation, allowing users to effortlessly create and edit videos using text prompts and information from existing documents.
Imagine 2.0: An advanced text-to-image model, featuring digital watermarking for AI-generated images and new editing options.
Code and Cloud Assistance:
Gemini Code Assist: With Gemini 1.5 Pro integrated into code assist tools, there's a 1 million token context window for substantial code transformations.
Gemini Cloud Assist: This tool supports users throughout the application lifecycle, from design to operation and optimization.
These developments demonstrate Google's dedication to leading the AI revolution by offering robust tools and infrastructure tailored to various business needs. Integrating these AI capabilities into Google Cloud and Workspace products aims to boost productivity, streamline processes, and drive innovation across different sectors.