Google Cloud showcased innovations in various stages of development over the three days of its Google Cloud Next 2023 conference held in San Francisco last week. The predominant focus of the event was generative AI, and with more than 160 features announced it is difficult to select highlights, but here are a few standout offerings.
AI optimised infrastructure
CEO Thomas Kurian told participants “Our AI-optimised infrastructure is a leading choice for training and serving gen AI models. In fact, more than 70% of gen AI unicorns are Google Cloud customers, including AI21, Anthropic, Cohere, Jasper, MosaicML, Replit, Runway and Typeface; and more than half of all funded gen AI startups are Google Cloud customers”.
Google Cloud’s global network of data centres are optimised for training and running AI applications. At Google I/O in May, It announced the development of A3 virtual machines (A3 VMs) which deliver high-performance training for machine learning workloads with greater bandwidth and speed than the previous A2 VMs. Last week Google confirmed the A3 VMs will be generally available from September and powered by Nvidia’s industry leading H100 GPU.
Additionally, the latest version of Google’s own machine learning specific Cloud TPU v5e will run large scale AI training and inference with more than twice the performance and cost efficiency of previous models.
A swath of hardware and software enhancements will together enable businesses to fully integrate and optimise on- premise, data centre and edge computing workloads. The end result is reduced operational costs, greater functionality and improved business outcomes.
Vertex AI platform
The Vertex AI platform enables Google Cloud customers to build, deploy and scale machine learning (ML) models using their own data. In his keynote Thomas Kurian revealed the number of generative AI customer projects running on the platform had increased 150 times between April and July this year. The platform offers more than 100 foundation models many of which are optimised for different tasks and sizes, including text, chat, images, speech and software code. Industry specific models are also accessible such as Sec-PaLM 2 for cybersecurity and Med-PaLM 2 for healthcare and life sciences.
When a business utilises a model on the Vertex AI platform, it receives a snapshot of the model allowing the business to train it in a private configuration using private documents and data. The prompts, data, user inputs and outputs are not accessible to other customers and are not used to improve models.
Duet AI in Workspace and Google Cloud
Duet AI, a natural language conversational assistant is Google’s equivalent to Microsoft’s 365 Copilot, was launched at Google I/O in May. Its original Workspace features included helping to write and refine content in Gmail and Google Docs, create original images in Google Slides, identify data insights in Google Sheets and improve connections in Google Meet. The assistant will now take notes, send meeting summaries and translate captions in Google Meet. In Google Chat users can chat directly with Duet AI and use it to keep up to date with Chat events they have missed.
Duet AI in Google Cloud, which will be generally available later this year, will assist with every aspect of cloud computing including coding, software development, hardware configurations, data analytics, networking and security management.
Sunil Potti, VP/GM of Google Cloud Security explained how Google Cloud Security AI Workbench, powered by Sec-PaLM 2 Security AI, will be enhanced with Duet AI. Duet AI in Security Command Centre will provide near-instant analysis of security findings and possible attack paths to simplify complex issues for non-specialist IT staff to help ensure they do not overlook critical events.
In Mandiant Threat Intelligence Duet AI will help identify techniques and procedures used by cyber criminals by instantaneously summarising real time threat intelligence data for technicians to counteract.
In Chronicle Security Operations Duet AI will simplify search, complex data analysis and threat detection engineering to reduce workload by automatically summarising events and cases, giving context and guidance on important threats, and recommending responses. Duet AI also powers Chronicle’s new natural language search so technicians can enter questions in natural language and Chronicle will generate a query from their statement and present a fully mapped search syntax.
Over the three day event Google Cloud clearly demonstrated its focus on harnessing the capacity of AI to enrich customer outcomes.
About the Author
The Education of a Value Investor by Guy Spier
The Digitisation of Everything - Retail
The Digitisation of Everything - Entertainment
The Digitisation of Everything - Automotive
The Digitisation of Everything - Payments