Exactly three months after OpenAI’s ChatGPT made a grand entry, Google introduced Bard. Since then, Google has been riding the generative AI wave with new features across various domains including Vertex AI, Model Garden and more. Over 70% of AI unicorns operate on Google Cloud infrastructure, exemplified by companies like Anthropic, AI21 Labs, and Character. AI, leveraging partnerships for global service delivery.
After Google I/O Connect, the big tech is back with Google Cloud Next, scheduled to be held in San Francisco from August 29th to 31st. Let’s take a look at major announcements and upgrades made at the event.
Vertex AI Gets Bigger & Better
Unveiled at Google I/O 2021, Vertex AI, the cornerstone suite of generative AI products for Google, gets new upgrades at Next. Now their proprietary LLM PaLM has four times more input token length and is available in 38 languages. Their AI-powered coding bot Codey’s new version has up to 25% quality improvement in code generation. Text-to-image diffusion model Imagen also has new updates with higher image quality, with a new feature called Style Tuning. It allows customers to align generated images with their brand guidelines using reference images.
Additionally, Meta’s Llama 2, Anthropic’s Claude 2, and TII’s Falcon join Model Garden, expanding model variety. Vertex AI Extensions aid model-data connectivity, Data Connectors grant read-only access to Google and third-party data, Vertex AI Search brings Google Search to business data, and Vertex AI Conversation enables easy chatbot deployment. Additionally, Colab Enterprise combines user-friendly Colab notebooks with enterprise-level security. Vertex AI Search is now generally available.
Duet AI for All
Good news for all as Duet AI in Google Workspace is now publicly available.
Now in Google Meet, Duet AI can make your appearance better and has improved sound quality. It will also take notes, and can even attend meetings on your behalf. Google Chat’s Duet AI answers queries, summarises content, and aids in recapping conversations, now featuring an improved UI and connectivity options.
Besides Workspace, Duet AI in Google Cloud is expanding its capabilities for various user groups, including developers, data practitioners, and cybersecurity professionals. It is being integrated into products like BigQuery, Looker, and the Database Migration Service to offer contextual assistance, auto-suggestions, and collaboration features. The launch of BigQuery Studio provides an all-in-one interface for data engineering, analytics, and machine learning. Moreover, Google Cloud is facilitating the integration of Vertex AI models into BigQuery and unveiling AlloyDB Omni, a downloadable edition that runs on multiple platforms.
Google is also introducing AlloyDB AI, enabling easy construction of enterprise AI applications, and announcing Cloud Spanner Data Boost for efficient operational data processing. Developer tools like Jump Start Solutions and Application Integration are becoming generally available, along with an extended partnership with GitLab for secure DevOps solutions. This expanded suite of Duet AI tools includes Cloud Workstations and Application Integration support.
Advancements in Google’s Supercomputing & Cloud Services
Bringing back the focus on supercomputing, Google revealed that their Cloud A3 Supercomputer, based on NVIDIA H100 designed for demanding generative AI workloads and LLMs, will be available generally from next month. The speciality of A3 is it leverages Google infrastructure processing units, claiming to provide three times faster training and better networking than the previous generation. Alongside, the Cross-Cloud Network is also introduced which is a new platform ensuring secure inter-cloud application connectivity.
Plus, now with Google Distributed Cloud (GDC), customers can have the flexibility to deploy Al and data workloads using select Vertex Al services and AlloyDB Omni. Additionally, the premium GKE Enterprise edition offers multi-cluster horizontal scaling, meeting the needs of today’s rigorous AI and ML workloads.
Amid updates focused on generative AI, Google Maps is introducing three Environment APIs – Solar, Air Quality, and Pollen – aimed at aiding businesses in reducing emissions and adjusting to environmental changes. Designed to cater to diverse health needs, it offers localised pollen count data, heatmap visuals, comprehensive plant allergen insights, and practical strategies for minimising pollen exposure.
The API computes daily pollen grain levels across a 1×1 km2 grid in over 65 countries, including a five-day forecast, encompassing three plant categories and 15 species. By integrating factors like land cover, climate data, and pollen production, the model generates dependable predictions of local pollen levels and associated risks.
Google Cloud is partnering with various customers and organisations, ranging from startups to enterprises, to enhance their generative AI. A121 Labs is integrating industry-specific generative AI into BigQuery, Bayer is using Vertex AI Search and Med-PaLM 2 for medical purposes and Culture Amp is using Vertex AI for employee feedback.
While MSCI is aiding investors with risk management, Runway is deploying AI for creative processes while Six Flags is implementing AI for park guidance. Google Cloud also expanded partnerships with SAP and Docusign for industry solutions and contract assistance. Furthermore, domain-specific datasets from partners like Acxiom, Bloomberg, CoreLogic, Equifax, NIQ, and Transunion are now available on Google Cloud for AI training.
Google Cloud’s always-on AI collaborator, Duet AI, offers generative AI-powered support for cloud users, extending to cybersecurity professionals through Duet AI’s integration in three core solution realms: Mandiant Threat Intelligence for streamlined threat assessment, Chronicle Security Operations for swift threat discovery, and Security Command Center for risk comprehension and remediation suggestions. These features are in Preview and set to be widely available later this year, with Mandiant Hunt for Chronicle enabling proactive search for hidden attacks using frontline intel. Further details are available in the blog post draft.
Read more: Google Turns AI ‘Bold & Responsible’