This year's Google Cloud Next event, which took place on August 29-31, featured many exciting innovations and services related to Google Cloud technologies. We've summarized the highlights for you below.

‍

Google Cloud's New Infrastructure and Tools Announced

● Cloud TPU v5e: The most cost-effective, versatile and scalable purpose-built AI accelerator. Cloud TPU v5e is scalable to tens of thousands of chips and is optimized for efficiency by isolation. Compared to Cloud TPU v4, it provides a 2-fold increase in training performance and up to 2.5x in inference performance.

● A3 VMs met NVIDIA H100 GPU: It is purpose-built with high-performance networking and other enhancements to enable the most demanding gene-AI and large language model (LLM) innovations. This allows it to achieve three times better training performance than the previous generation A2.

● GKE Enterprise: It enables multi-cluster horizontal scaling, which is essential for difficult and important AI/ML workloads. GKE offers benefits such as auto-scaling, workload editing, and auto-upgrade now available in Cloud TPU v5e.

● Cross-Cloud Network: Designed to give customers easier access to Google services from any cloud, reducing network latency by up to 35%.

● Google Distributed Cloud: It is designed to meet the demands of organizations that want to run workloads in data centers. In addition to next-generation hardware and new security features, Vertex offers AI integrations and a new managed AlloyDB Omni offering on GDC Hosted.

‍

Explore Advanced Features of Vertex AI Platform

The most comprehensive AI platform that allows customers to build, deploy and scale machine learning (ML) models. Between April and July of this year, there was tremendous use, with the number of AI client projects increasing by more than 150 times. Customers have access to more than 100 basic models, including third-party and popular open source versions, in our Model Gardens. These are all optimized for different tasks and different sizes, including text, chat, images, speech, software code, and more.

● PalM 2, Imagen and Codey Upgrades: PalM 2 has been updated to 32k context windows so that customers can easily process longer form documents such as research articles and books. Imagen is also being developed and support for new languages is being expanded in Codey.

● Adjustment tools: For PalM 2 and Codey, adapter tuning is available for general use and preview, respectively, which can help improve LLM performance with as few as 100 examples. It also introduces a new tuning method called Style Tuning for Imagen, so businesses can create visuals that fit their own custom brand guidelines or other creative needs with a small amount of reference images.

● New models: The launch of Llama 2 and Code Llama from Meta and the Falcon LLM from Technology Innovative Institute, a popular open source model, is announced, as well as a pre-announcement of Claude 2 from Anthropic. In the case of Llama 2, it will be the only cloud provider offering both adapter tuning and RLHF.

● Vertex AI extensions: Developers can access, create, and manage extensions that provide real-time information, consolidate company data, and act on behalf of the user. This opens up endless new possibilities for Gen-AI applications that can work as an extension of your organization, providing the ability to access private information and transact on third-party platforms such as your CRM system or email.

● Grounding: Vertex AI foundation models, an enterprise grounding service that works on Search and Conversation and allows customers to base responses on their own enterprise data to give more accurate answers to customers is announced. It is also working with several initial customers to test grounding with technology that supports Google Search.

● Digital Watermarking in Vertex AI: Powered by Google DeepMind SynthID, this theme embeds the watermark directly into the pixel image, making it invisible to the human eye and making it difficult to tamper with. Digital watermarking provides customers with a scalable approach to responsibly creating and identifying AI-generated images. Google Cloud is the first hyperscale cloud provider to offer this technology for images generated by artificial intelligence.

● Colab Enterprise: This managed service combines the ease of use of Google's Colab notebooks with enterprise-grade security and compatibility. Data scientists can use Colab Enterprise to access the full range of Vertex AI platform capabilities, integrate with BigQuery, and even accelerate AI workflows (workflows) collaboratively with code completion and creation.

Equally important to discover and train the right model is to check your data. From the beginning, we designed Vertex AI to provide complete control and segregation over your data, code, and IP with zero data leakage. When you customize and train your model with Vertex AI (with custom documents and data from your SaaS applications, databases, or other proprietary sources), you do not expose that data to the underlying model. Taking a snapshot of the model allows you to train and encapsulate it together in a custom configuration, giving you full control over your data. Your prompts and data, as well as user inputs during inference, are not used to develop models and are not accessible to other customers.

‍

What are Duet AI Features in Google Workspace and Google Cloud?

In May, Duet AI at I/O was introduced at I/O; powerful new features were introduced in Workspace, and developer features such as code and chat assistance were showcased in Google Cloud. Since then, trusted testers around the world have experienced the power of Duet AI; however, efforts have been made to expand Duet AI's capabilities and integrate it into a wide range of products and services in Workspace and Google Cloud.

Speaking of Workspace, a few months ago, with the launch of Duet AI, a number of features were introduced that will make the teams that use it more productive, such as allowing you to write and refine content in Gmail and Google Docs, create original images in Google Slides, turn ideas into action and data into analytics with Google Sheets, create more meaningful connections in Google Meet, and more allows you to do it. Since then, thousands of companies and more than a million trusted testers have helped ensure that every user and organization has control over their data while using Duet AI as a strong collaboration partner.

New enhancements:

● Duet AI on Google Meet: Duet AI will take notes during video calls, send meeting summaries, and even automatically translate subtitles into 18 languages. Also, Duet AI in Meet announced the studio look, studio lighting, and studio sound to ensure that each meeting attendee can be seen, heard, and understood clearly.

● Duet AI in Google Chat: You'll be able to chat directly with Duet AI to ask questions about your content, get a summary of documents shared in an area, and track missed conversations. It also featured a revamped user interface to keep you up to date with chats, new shortcuts and improved search, as well as group meetings in Chat, which allows teams to start meetings from where they're already working together.

Duet AI is now available for general use in Google Workspace, and Duet AI's preview capabilities in Google Cloud are being expanded and will be available to the general public later this year.

Duet AI is now able to provide AI support across a wide range of Google Cloud products and services, such as a coding assistant to help developers write code faster, an expert consultant to help operators quickly troubleshoot application and infrastructure issues, providing faster and better insights as a data analyst, and recommending best practices to help prevent cyber threats as a security consultant.

Duet AI Announcements on Google Cloud

● Software development: Duet AI helps developers be more productive by providing expert assistance throughout your entire software development lifecycle. In addition to code completion and code generation, it can help you modernize applications faster by helping you with code rearrangement; any developer using Duet AI in APIGEE can now easily create APIs and integrations using simple natural language prompts.

● Application and infrastructure operations: Operators can chat with Duet AI in natural language on various services directly in Google Cloud Console to quickly get “how-to” information on infrastructure configuration, deployment best practices, and expert advice on cost and performance optimization.

● Data analysis: Duet AI in BigQuery provides contextual help for writing SQL queries as well as Python code, creates full functions and code blocks, automatically suggests code completions, and explains SQL statements in natural language, and can generate suggestions based on your schema and metadata. These capabilities can allow data teams to focus more on business outcomes.

● Accelerating and moderating databases: Duet AI in Cloud Spanner, AlloyDB, and Cloud SQL help generate code to configure, modify, or query data using natural language. Duet AI also brings the power of Duet AI to the Database Migration Service (DMS), which helps automate the conversion of database codes that cannot be converted with traditional translation technologies such as stored procedures, functions, triggers, and packages.

● Security operations: Duet AI is brought to security products including Chronicle Security Operations, Mandiant Threat Intelligence and Security Command Center, enabling security professionals to more efficiently prevent threats, reduce hassle in security workflows, and maximize security capabilities.

With Duet AI, your data is your data. Your code, inputs to Duet AI, and suggestions generated by Duet AI will not be used to train shared models or develop any products.

Simplify Analytics at Scale with Google Cloud

Data is at the heart of Genai, which is why new features are being brought to Google's Data and Artificial Intelligence Cloud to help unlock new analytics and increase the productivity of data teams. In addition to the launch of Duet AI, which helps data engineers and data analysts with our BigQuery, Looker, Spanner, Dataplex, and database migration tools, there are a few other important announcements in the data and analytics space:

● BigQuery Studio: As a single interface for data engineering, analysis, and predictive analysis, BigQuery Studio helps improve the efficiency of data teams. In addition, new integrations into Vertex AI core models help organizations enable AI-enabled data pools with AI through innovations in cross-cloud analytics, governance, and secure data sharing.

● AlloyDB AI: AlloyDB AI offers an integrated set of capabilities for easily building GenAI applications, including high-performance vector queries that are up to 10x faster than Standard PostgreSQL. You can also run AlloyDB almost anywhere with AlloyDB Omni. This includes on-premise, through Google Cloud, AWS, Azure, or Google Distributed Cloud.

● Data Cloud Partners: The open data ecosystem is a value for customers' GenAI strategies, and the scope of partner solutions and datasets available in Google Cloud continues to expand. Our partners like Confluent, DataRobot, Dataiku, Datastax, Elastic, MongoDB, Neo4j, Redis, SingleStore, and Starburst all offer new capabilities to help customers accelerate and enhance AI development with data. Our partners are also adding more datasets to the Analytics Hub that customers can use to build and train GenAI models.

These innovations help organizations harness the full potential of data and artificial intelligence through a unified data base. With Google Cloud, companies can now run their data anywhere and move AI and machine learning tools directly to their data, reducing the risk and cost of data migration.

‍