Google makes Gemini Pro API available to developers

 Less than a week after its unveiling, Google began making Gemini Pro available to developers and organizations on December 13, along with a host of other AI tools, models, and frameworks. . Gemini, the latest and greatest AI model, was unveiled on December 6, revealing technical details and sharing a roadmap of what's to come, including immediate availability for those interested in testing it on Bard. We also previewed what developers will be able to build with its cutting-edge multimodal capabilities.

Google's new announcements

On December 13, Google made a series of new announcements, here they are in brief:

• The Gemini Pro API is available to developers in Google AI Studio also to businesses through the Vertex AI Platform from Google Cloud.
• New templates in Vertex AI to help developers and businesses flexibly build and deploy apps:
  • an update to Image 2 text-to-image streaming tool.
  • a family of core models optimized for the healthcare industry, MedLM available (via greenlist) to Google Cloud customers in the United States.
• general availability of Duet AI for developers And Duet AI for security operations .

On the company blog, Thomas Kurian, CEO of Google Cloud, writes: “Throughout 2023, we have introduced new AI innovations for our customers and the broader community of developers and users , including: AI Hypercomputing to train and support generative AI models; support generative AI in Vertex, our enterprise AI platform; Duet AI for Google Workspace and Duet AI for Google Cloud. (..) Today we are introducing a number of important new features within our AI stack to support Gemini , our largest and most capable AI model. Gemini was designed from the ground up to be multimodal, meaning it can seamlessly generalize, understand, leverage, and combine different types of information, including text, images, audio, video, and code computing, in the same way as humans. seeing, hearing, reading, listening, and speaking many different types of information at the same time.

Google Cloud's unified AI stack

Gemini will be part of a vertically optimized and integrated AI technology stack, comprised of several core elements, all designed to work together:

• Super-Scalable AI Infrastructure : Google Cloud provides businesses with cutting-edge AI-optimized infrastructure – the same infrastructure Google uses – to train and support models. We offer this infrastructure in several ways: as a service in cloud regions, through Google Distributed Cloud for use in enterprise data centers and at the edge.
• World-class models : In late 2022, Google launched Pathways Language Model (PaLM), closely followed by PaLM 2, and now offers Gemini Pro. It also introduced industry-specific models, such as Med-PaLM and Sec-PaLM.
• Vertex AI – Leading enterprise AI platform for developers : To help developers create agents and integrate generative AI into their applications, Google has rapidly enhanced Vertex AI, the AI ​​development platform. Vertex AI helps customers discover, customize, power, deploy and manage agents built using Gemini APIs and a curated list of more than 130 open source and third-party AI models that meet rigorous security standards and quality of Google's sector of activity. Vertex AI uses Google Cloud's built-in data governance and privacy controls and provides tools to help developers use models responsibly and securely. Vertex AI also offers search and conversation tools that use a low-code approach to develop sophisticated search and conversation agents that can work across multiple channels.
• Duet AI – AI Support Agents for Workspace and Google Cloud : Duet AI is an AI-based collaborative support that provides assistance to users when using Google Workspace and Google Cloud. Duet AI in Google Workspace, for example, helps users write, create images, analyze spreadsheets, write and summarize emails, discussion messages, and summarize meeting content. Duet AI in Google Cloud can help users code, deploy, scale, and monitor applications, as well as identify and accelerate remediation of cybersecurity threats.

A new version of Imagen 2 and MedLM for the medical field

Google also unveiled an updated version of its Imagen 2 image model, its most advanced text-to-image synthesis technology. This latest version offers improved photorealism, text rendering and logo generation capabilities, allowing you to easily create images with text and logo overlays.

Furthermore, continuing the path undertaken with the creation of sectoral models with Med-PaLM, he announced MedLM, a suite of specific models for the medical field. MedLM offers its customers the power of Google's core models optimized for the medical field.

The Vertex AI platform powered by Gemini

Google announced that Gemini Pro is now available in preview on Vertex AI and will enable developers to create innovative and diverse agents currently capable of processing information from text, code, images and video. Vertex AI will help customers deploy and manage agents in production, automatically assess the quality and reliability of agent responses, as well as monitor and manage them.

Vertex AI provides comprehensive support for Gemini, with the ability to discover, customize, enhance, manage and deploy agents built with Gemini APIs, including:

• different ways to customize agents built with Gemini using your own data, such as designing prompts, tuning based on adapters such as low-rank adaptation (LoRA), reinforcement learning by human feedback (RLHF) and distillation.
• Augmentation tools, enabling agents to use built-in elements to retrieve, understand, and act on real-world information with configurable recovery augmented generation (RAG) blocks. Vertex AI also offers extensions to perform actions on behalf of users in third-party applications.
• Anchor to improve the quality of responses from Gemini and other AI models by comparing results to high-quality web and enterprise data sources.
• A wide range of controls to help customers use generative AI models, including Gemini, safely and responsibly.

More announcements

In addition to support for Gemini in Vertex AI, the following were announced:

• Auto Side-by-Side (Auto SxS ), an automated model comparison tool. Auto SxS is faster and more convenient than manually measured models, and is also customizable for several specific tasks to meet new generative AI use cases.
• With the addition of Mistral, ImageBind and DITO to Vertex AI's Model Garden continuing our commitment to an open model ecosystem.
• Gemini Pro will soon be integrated with Vertex AI Search and Conversation, helping customers quickly create engaging, production-ready applications.

Expanding Duet AI Capabilities

Google also announced that more than 25 code support and knowledge base partners will offer data sets specific to their platforms, allowing Duet AI for Developers users to receive AI support based on of their coding and data models, partner product documentation, best practices and more. Useful business resources.

Duet AI for Security Operations, Google Cloud's unified security operations platform, can help you more effectively protect organizations against cyberattacks. Security teams can hone their skills and help accelerate threat detection, investigation, and response through the power of generative AI. With Duet AI for Security Operations, Google brings AI assistance to Chronicle, where users can search large amounts of data in seconds with custom natural language queries, reduce time-consuming manual reviews, quickly surface critical context leveraging automatic data summaries and alerts and improve response times using next step suggestions to support incident resolution.

In addition to these new features on its vertically integrated AI technology stack, Google is increasing its indemnification guarantee to help customers protect themselves against copyright issues.