Google announced an expansion of its partnership with Anthropic, committing to the highest standards of AI security and safety, and that Anthropic will leverage the latest generation Cloud TPU v5e chips for AI inference.
The two companies have worked closely since Anthropic’s founding in 2021, as the AI safety and research startup has built one of the largest Google Kubernetes Engine (GKE) clusters in the industry. Anthropic also currently uses AlloyDB, Google’s fully managed PostgreSQL-compatible database, to handle transactional data with high performance and reliability; and Google’s popular BigQuery data warehouse to analyze vast datasets and extract insights for its employees. Working Together to Advance AI Safety Standards
announcement follows the attendance of both companies at the inaugural AI Safety Summit (AISS) hosted by the UK government at Bletchley Park last week. Attended by senior leaders from Google and Anthropic, the historic event gathered government officials, technology leaders, and experts from academia and civil society to share concerns around frontier AI and determine measures that could be taken to address them. This builds on Google and Anthropic’s respective collaborations with the Frontier Model Forum, an initiative to support the development of better measures for AI safety. Both companies are also working with MLCommons, a non-profit group which works to create trustworthy, third-party data about AI systems. Committing to the Highest Standards of AI Security
To protect organizations who deploy Anthropic’s models on Google Cloud —and Anthropic’s own employees and contractors—Anthropic is also now using Google Cloud’s security services, including Chronicle Security Operations, Secure Enterprise Browsing, and Security Command Center.
Security Command Center helps Anthropic gain visibility into its Google Cloud environment and detect misconfigurations, while Secure Enterprise Browsing allows the company to secure access for managed and unmanaged devices for both employees and contractors. Chronicle Security Operations provides Anthropic with AI-powered remediation with frontline threat intelligence through automation, as well as the ability to create automated security controls, policies, and configurations.
Scaling up with Cloud TPU v5e
Anthropic is now one of the first companies to deploy at scale TPU v5e—Google Cloud’s most versatile, efficient, and scalable AI accelerator to date. Now generally available, TPU v5e enables Anthropic to serve its Claude large language model (LLM) in a performant and efficient manner.
Cloud TPU v5e is purpose-built to deliver the cost-efficiency and performance required for medium- and large-scale training and inference. Results of the MLPerf 3.1 Training benchmark, released, found that TPU v5e delivers 2.3x higher training performance per dollar relative to the prior generation; this follows the MLPerf 3.1 Inferencing benchmark announced in September that found 2.7x serving performance per dollar compared to Cloud TPU v4. With Multislice large-scale training technology, Anthropic can scale its LLMs beyond the physical boundaries of a single TPU pod, up to tens of thousands of interconnected chips.
“Anthropic and Google Cloud share the same values when it comes to developing AI–it needs to be done in both a bold and responsible way,” said Thomas Kurian, CEO of Google Cloud. “This expanded partnership with Anthropic, built on years of working together, will bring AI to more people safely and securely, and provides another example of how the most innovative and fastest growing AI startups are building on Google Cloud.”
“Our longstanding partnership with Google is founded on a shared commitment to develop AI responsibly and deploy it in a way that benefits society,” said Dario Amodei, co-founder and CEO of Anthropic. “We look forward to our continued collaboration as we work to make steerable, reliable and interpretable AI systems available to more businesses around the world.”
SOURCE: PRNewswire