Google has announced several cutting-edge features and software for major products like Google Workspace and Google Cloud as it sharpens its focus on generative AI and artificial intelligence. Google CEO Sundar Pichai stated the company’s ambition to dominate the AI and generative AI markets at Google I/O 2023. At this event, cutting-edge products and services were unveiled, highlighting Google’s committed pursuit of superior AI.
With a market cap of $280 billion, Google is making significant investments in integrating AI into its wide range of cloud computing, teamwork, and consumer services. Sundar Pichai emphasizes the importance of AI and sees a group effort to fully utilize its potential. According to CRN, notable developments at Google I/O 2023 include Duet AI, advancements to Vertex AI, and Google Bard.
These consist of: Google Workspace Duet AI
New Bard Capabilities in PaLM 2 For Google Cloud, Duet AI
New Foundation Models for Vertex AI
RLHF Vertex AI
Supercomputers powered by Google A3 and Nvidia H100
Embedded Vertex AI APIs
1. Google Workspace Duet AI
Headlines about Duet AI for Google Workspace Strong generative AI functionalities are combined at Google I/O for improved cooperation. To help with content generation, Gmail Mobile incorporates generative AI. Google Slides uses AI to automatically create visuals with little input. For data analysis, automated classification, and the development of custom tables, Sheets uses generative AI. Smart canvas features, similar to smart chips, improve Google Docs’ assisted authoring.
2. Google’s Next-Generation Language Model is PaLM
PaLM 2, an improved iteration of Google’s big language model, was released. PaLM 2 includes the Gecko, Otter, Bison, and Unicorn models, each of which has a unique size. Gecko includes interactive offline features and was created for mobile devices. As a result of their training in scientific, mathematical, and more than 100 other languages, CEO Pichai emphasized their aptitude for logic, reasoning, and multilingual understanding.
3. New Bard Features
Updates to Bard were presented at Google I/O, utilizing the company’s brand-new PaLM 2 big language model. Better visual cues and responses are part of Bard’s growth. Users may now ask for travel advice and receive not only text but also detailed graphics for more insightful responses. Additionally, users can include images in their prompts, which fosters creativity. By utilizing Google Lens’s capabilities, this integration makes Bard’s capabilities more cutting-edge.
4. Google Cloud Duet AI
Duet AI for Google Cloud, an AI-powered collaborator that can accommodate users of all skill levels, is now available from Google. This skilled pair programmer helps cloud customers by assisting with code review, contextual code completion, targeted suggestions, and real-time function development. With its seamless integration with IDEs, the Google Cloud Console, and chat, Duet AI significantly changes how users create experiences across Google Cloud, offering increased productivity and innovation.
5. New Foundation Models for Vertex AI
Three new foundation models were added by Google to Vertex AI and are now available via the API, the UI in Generative AI Studio, or data science notebooks. These three new foundational models are Chirp, Codey, and Imagen. With real-time code completion and generation, Google’s Codey is a text-to-code foundation paradigm that is incorporated in SDKs or apps to increase developer productivity. For a variety of commercial applications, Imagen, the text-to-image paradigm, creates and customizes studio-quality images at scale. Through voice help and native language captioning, Chirp, their speech-to-text foundation paradigm, promotes inclusive involvement with clients, boosting relationships.
6. Vertex AI New RLHF
By incorporating human feedback for customized model advancement, Vertex AI is introducing New Reinforcement Learning from Human Feedback (RLHF), which improves tuning and prompt design. As the first cloud hyperscale provider to offer RLHF as a managed service within Vertex AI, Google sets the bar high by supporting model performance that is economical, consistent, and reliable as well as safer and more advantageous model deployment to production.
7. Supercomputers Google A3 Nvidia H100 GPUs are used
The private preview launch of the A3 GPU supercomputer, which combines Nvidia H100 Tensor Core GPUs with cutting-edge networking, expands Google Cloud’s compute options. By extending computing possibilities for machine learning workloads including large language models, generative AI, and diffusion models, this expansion caters to a range of customer sizes.
8. Embedded Vertex AI APIs
Google incorporates text and image APIs into Vertex AI, enabling programmers to create sophisticated applications like classifiers and recommendation engines. Text and visual data are converted into numerical vectors using embeddings while maintaining semantic links. huge models can process these vectors effectively, and they are useful for huge inputs like texts with many thousands of tokens.
