NVIDIA collaborates with Snowflake to develop large-scale language model AI.

2023-06-28

At the Snowflake 2023 User Conference, cloud database manufacturer Snowflake announced a partnership with chip giant Nvidia to jointly process so-called AI base models. According to the arrangement, Snowflake's customers will be able to rent cloud GPU capacity installed in Snowflake's data warehouse and utilize that capacity to optimize neural networks using Nvidia's NeMo framework, which was launched last year. Base models are very large neural networks, such as large language models, which are typically "pre-trained," meaning they have already reached a certain level of capability. Customers will use Snowflake's data warehouse to develop customized versions of the NeMo base model using their own data to meet their needs. Manuvir Das, Nvidia's Corporate Vice President of Enterprise Computing, said in a press briefing, "This is a very natural combination for both companies." Das continued, "The reason Nvidia and Snowflake are partnering is that if a company needs to create custom models for generative AI based on its data, which is stored in Snowflake's data cloud, why not introduce Nvidia's model building engine (NeMo) into Snowflake's data cloud, enabling enterprise customers to directly generate these models on their data cloud and then apply them to their business use cases." This announcement is part of the growing trend of using artificial intelligence, especially generative AI, as a business tool. On Monday, Apache Spark developer Databricks shocked the tech industry by acquiring startup MosaicML for $1.3 billion. MosaicML provides services for training and deploying base models. Snowflake will implement this service by purchasing Nvidia GPU instances from its existing partners. Das said, "Right now, we're just discussing expanding this partnership to include GPU instances." On Tuesday, Snowflake announced in another news release that it is expanding its Snowpark development platform to its Snowpark container service, which is currently in private preview. The company stated, "By expanding the scope of Snowpark, developers can unlock a wider range of infrastructure options, such as accelerated computing and AI software using Nvidia GPUs, to run more workloads within Snowflake's secure and governed platform without complex operations, including a broader range of AI and machine learning (ML) models, APIs, internally developed applications, and more." Das said, "Snowflake has a design structure that ensures when a customer chooses to compute on the Snowflake data cloud, it remains within the customer's boundaries, and then the NeMo engine just adapts to that model." Das added, "Of course, NeMo also has security responsibilities, which is why this is a joint engineering effort." This collaboration is the latest development following Nvidia's recent partnership with ServiceNow to use NeMo with ServiceNow's IT service customers. Das said that Snowflake's partnership is a "generic" partnership, while ServiceNow's partnership is more like an independent software vendor (ISV) model. ServiceNow uses NeMo code to train custom models for each customer, "so that when each of their customers does IT work and opens [fault] tickets, they get a response specific to that customer." Nvidia CEO Jensen Huang positions software as a key growth engine for the company, which has generated billions of dollars in revenue by selling GPU hardware for developing neural networks. NeMo is part of the company's promoted enterprise software stack, much of which is achieved through partnerships with cloud service providers. In March of this year, Nvidia CFO Colette Kress told investors at a Morgan Stanley conference, "Our software business revenue is now in the hundreds of millions of dollars, and we believe this is still an opportunity for us to move forward." Source: https://www.zdnet.com/article/nvidia-teams-up-with-snowflake-for-large-language-model-ai/

RECENT AI NEWS

RECENT AI TOOLS