Google today added a raft of capabilities intended to make its platforms more attractive to organizations looking to deploy artificial intelligence (AI) applications across multicloud computing environments.
Announced at the Google Cloud Next ’24 conference, the latest extensions to a Google Distributed Cloud (GDC) platform designed to be deployed anywhere add support graphical processor units (GPUs) from NVIDIA and the Google Kubernetes Engine (GKE) to enable applications to be deployed in containers.
Thomas Kurian, CEO of Google Cloud, told conference attendees these additions to a set of platforms and services that Google describes as an AI Hypercomputer provide access to both open source and commercial AI models that can be deployed in any cloud.
Google first made GDC available in 2023 and is now looking to extend the reach of a platform that, in addition to the Google Cloud Platform (GCP), is designed to also be deployed in any other cloud computing environment. The goal has been to streamline the building and deployment of applications across a connected hybrid IT environment and air-gapped data centers.
In the case of AI applications, Google is extending the reach of GDC to now include multiple AI models, including its own Gemma large language model (LLM), in addition to making available AlloyDB Omni for Vector Search to make it simpler to access data.
Google is also making it simpler to build AI applications via an update to its Gemini tools for generating code that is based on an update to CodeGemma, an instance of Gemma specifically trained to generate code. Available in public preview, version 1.5 of Gemini Pro comes in two forms to provide consistent access to 128,000 tokens and one million tokens, respectively. Gemini 1.5 Pro in Gemini Code Assist, available in private preview, can now take advantage of up to one million tokens to provide deeper, more accurate insights into code bases.
In addition, Gemini now supports the processing of audio files. Organizations can process in a single stream one hour of video, 11 hours of audio, codebases with over 30,000 lines of code, or more than 700,000 words.
Google is also extending the scope of tools it makes available via its Vertex AI, a platform for managing machine learning operations (MLOps) that now provides access to the Claude 3 LLM from Anthropic along with additional tools for evaluating AI models and managing the workflows used to build them using libraries of prompts curated by Google.
Gemini is also being embedded within Google AlloyDB database and the BigQuery data lakes to enable IT teams to apply generative AI to streamline the management of both platforms. Google is also adding vector indexing capabilities to generate more real time outputs when invoking LLMs.
At the same time, Google is expanding it networking and storage services to provide significantly higher throughput for AI models.
Google is extending the AI tools and capabilities provided in Google Workspace to make it easier to create videos, craft emails using voice prompts, manage spreadsheets and secure sensitive data in addition to adding generative AI capabilities to its Looker business intelligence application. An Imagen tool also makes it possible to now create animated videos and apply digital watermarks to any content created.
Finally, Google is also injecting more generative AI capabilities across its Mandiant portfolio of cybersecurity services.
There are, of course, plenty of options when it comes to building and deploying AI applications. The issue now is determining to what degree organizations will want to manage the underlying platforms themselves versus relying on a managed service provided by a cloud service provider.
