Google releases new generative AI products and features for Google Cloud and Vertex AI

Google today announced a variety of new products and features that give customers access to new ways of generating AI and expanded ways to use and optimize custom models, at its annual developer conference, Google I/O 2023.

Among the new offerings are three new base models available in Vertex AI, Google Cloud’s end-to-end machine learning platform. These templates are Codey, a text-to-code template that can help developers with code completion, generation, and chat; Imagen, a text-to-image template that can help clients generate and edit high-quality images for any business need; and Chirp, a speech-to-text model that can help organizations engage with customers and constituents more inclusively in their native languages.

New tools for generative AI

Google’s new offerings, which in total include three brand-new core templates, an embedding API, and a unique optimization feature, aim to equip developers and data scientists with more capabilities to build generative AI applications faster.

The first of the new core templates released today, Codey, aims to accelerate software development by providing real-time code completion and generation. Perhaps best of all, it can be customized to a user’s code base. The model supports more than 20 coding languages ​​and is capable of simplifying a wide variety of coding tasks. It essentially helps developers ship products faster by generating code based on natural language prompts and offers code chat for assistance with debugging and documentation.

Imagen, the second basic model, helps organizations generate and edit high-quality images for a wide variety of use cases. This text-to-image model simplifies the creation and modification of images at scale, providing enterprise-grade, low-latency data governance capabilities.

In one of the most exciting features launched today, unmask editing allows users to make changes to an image generated through natural language processing. This essentially means that you can have a conversation with the UI about how to generate the perfect photo, continuously repeating the output. The template also offers image upscaling and captioning in over 300 languages. Users can quickly generate production-ready images, while built-in content moderation ensures safety.

The third basic model, Chirp, focuses on improving customer engagement through text-to-speech. Trained on millions of hours of audio, Chirp supports more than 100 languages, with more languages ​​and dialects added today. Chirp is a new version of Google Speech model with 2 billion parameters now boasting 98% accuracy in English and up to 300% relative improvement in languages ​​with fewer than 10 million speakers.

Finding new relationships in data

To complement its new base templates, Google has introduced the Embeddings API for text and images, now also available in Vertex AI. This API converts text and image data into multidimensional numeric vectors that map to semantic relationships, enabling developers to create more engaging apps and user experiences. Applications range from powerful semantic search and text classification functionality to Q&A chatbots based on an organization’s data.

Another standout feature of the Vertex AI update is human feedback reinforcement learning (RLHF), which Google says makes Vertex AI the first end-to-end machine learning platform among hyperscalers to offer RLHF as a managed service . This feature allows organizations to incorporate human feedback to train a reward model for fine-tuning baseline models, making it particularly useful in industries where accuracy and customer satisfaction are critical.

Google’s new Generative AI advances are poised to revolutionize the development landscape, giving developers and data scientists an increasingly sophisticated toolset to leverage AI in the cloud. With these new models and foundational tools, the possibilities for innovation and responsible development of AI are virtually limitless.