Google Cloud is launching Duet AI, which takes Google's foundational generative AI models to make building and deploying cloud applications easier. Developers can also optimize code with Generative AI Support in Vertex AI with three new foundation models.
The launch of Duet AI was announced at Google I/O. Google Cloud was a headliner during CEO Sundar Pichai's keynote, which focused on generative AI as well as examples of where the company was using AI across Google Workspace, Gmail, Docs, Slides and Meet. Google Cloud CEO Thomas Kurian's appearance at Google I/O highlights the unit's maturity. Google Cloud was profitable in the first quarter.
"All of these generative AI advancements are coming to business," said Kurian. He added that Google Cloud's AI technologies for business are secure and rightsized based on the compute needed.
Google Cloud's announcements land as Google outlined its Transformer architecture as well as its next-generation language model called PaLM 2. PaLM 2 has improved multilingual, reasoning, and coding capabilities. Competitively, Google Cloud’s Duet AI and Generative AI Support in Vertex AI will face off with Microsoft’s numerous OpenAI ChatGPT integrations, notably CoPilot.
Duet AI, which is available today for early adopters via Google Cloud's AI Trusted Tester Program, is designed to make developing cloud applications less complicated by offering the following across multiple services:
- Chat assistance that uses natural language to deliver answers on development and cloud-related questions. Users can engage with chat assistance, learn how to use services, and get implementation plans while providing architecture and coding best practices.
- Duet AI for AppSheet, which will allow Google Cloud users to create business applications, connect data and build workflows into Google Workspace in a natural language no-code format. Users will be able to describe an app guided by prompts.
Generative AI support in Vertex AI
Google Cloud also outlined Generative AI support in Vertex AI, which enables enterprises to fine-tune Codey models using their own code base. Codey models will be available directly in Vertex AI and ultimately connect to Duet AI. In March, Google Cloud outlined Gen App Builder to enable generative chat and search apps, Generative AI support in Vertex AI, Model Garden and Generative AI Studio.
At Google I/O, Google Cloud outlined three new foundation models including:
- Codey, a text-to-code foundation model to improve code generation and completion and quality. Codey suggests the next few lines based on the code entered, generates code based on natural language prompts, and features a bot to help with documentation and questions.
- Imagen, a text-to-image foundation model to generate and customize studio-grade images.
- Chirp, a speech-to-text foundation model to provide native language captioning and voice assistant. Chirp is a version of Google Cloud's 2 billion parameter speech model that supports more than 100 languages.
The company also added embedding API for text and images so developers can more easily build classifiers, question and answering systems and other applications and reinforcement learning with human feedback to extend Vertex AI's tuning. Human feedback is handy to generate accurate models for industries like healthcare.
Google Cloud also said Generative AI Studio, Model Garden, and PaLM 2 for Text and Chat are moving from trusted tester availability to preview. Google Cloud also announced A3 instances to run AI workloads.