Contextual AI, the company building
AI that works for work, today announced a strategic partnership with Google
Cloud as its preferred cloud provider to build, run, and scale its growing
business and to train its large language models (LLMs) for the enterprise.
Contextual AI came out of stealth
mode in June 2023 to build the next generation of foundation models that
provide fully customizable, trustworthy, privacy-aware AI that lets companies
focus on the work that matters. The company selected Google Cloud for its
leadership and open approach to generative AI, as well as the comprehensiveness
of its compute infrastructure, purpose-built for AI/ML. AI workloads require large amounts of
time-consuming computation, both to train the underlying machine learning
models and to serve those models once they are trained. As part of the partnership,
Contextual AI will build and train its LLMs with the choice and flexibility
offered through Google portfolio of GPU VMs, specifically A3 VMs and A2
VMs, which are based on the NVIDIA H100 and A100 Tensor Core GPUs,
respectively. Contextual AI will also leverage Google Cloud's custom AI
accelerators, Tensor Processor Units (TPUs), to build its next generation of
LLMs.
Contextual AI enables
enterprises to unlock the true potential of AI by grounding language models in
their internal knowledge bases and data sources. Built on Google Cloud,
Contextual Language Models (CLMs) will craft responses that are tailored to an
enterprise's data and institutional knowledge, which results in higher
accuracy, better compliance, less hallucination, and the ability to trace
answers back to source documents. For example, a customer service agent can
leverage CLMs to answer a user's questions with greater precision by relying
only on approved data sources such as the user's account history, company
policies, and prior tickets that are similar or a financial advisors can
automate reporting workflows to provide personalized recommendations based on a
client's unique portfolio and history, proprietary market insights, and other
private data assets.
"Building a large language model to solve some of the
most challenging enterprise use cases requires advanced performance and global
infrastructure,"
said Douwe Kiela,
chief executive officer, Contextual AI. "As
an AI-first company, Google has unparalleled experience operating AI-optimized
infrastructure at high performance and at global scale which they are able to
pass along to us as a Cloud customer."
Contextual AI is helping
its customers—many of which are Fortune 500 companies—solve shared pain points
when it comes to AI, including hallucinations, attribution, compliance,
latency, and data privacy. Contextual AI's LLMs take into consideration data
privacy while providing customization and efficiency. Co-founder Douwe Kiela helped
pioneer the retrieval augmented generation (RAG) technique that underpins Contextual
AI's text-generating AI technology. RAG allows enterprise customers to build
custom LLMs on top of their data, ensuring that data remains secure, using
external sources to generate responses that take context into consideration.
"At Google Cloud, we believe that enabling the next
generation of generative AI services requires a purpose-built, AI-optimized
infrastructure stack, spanning hardware, software, and services," added Mark Lohmeyer,
VP/GM, Compute and ML Infrastructure, Google Cloud. "We're proud to offer
customers unparalleled flexibility and performance, and excited to support
Contextual AI's world-class team of AI innovators as they build next generation
LLMs for the enterprise on Google Cloud."