Snowflake Inc. is dropping a lot of big names as it leans into the hype for artificial-intelligence capabilities in cloud software.
Snowflake
SNOW,
announced a partnership with Nvidia Corp.
NVDA,
on Monday night to customize AI models for businesses that executives say will keep proprietary data safe. Earlier in the day, Snowflake revealed a partnership with Microsoft Corp.
MSFT,
to bring generative AI to the company’s Data Cloud platform and link the company to OpenAI.
Nvidia, Microsoft and OpenAI are good names to drop in sales pitches, as hype for generative AI revs up, and Snowflake Chief Executive Frank Slootman added them at Snowflake’s Summit 2023 event in Las Vegas. Nvidia Chief Executive Jensen Huang joined Slootman in a “fireside chat” Monday evening ahead of a Tuesday morning keynote.
“A large language model turns any data knowledge base into an application,” Huang told the assembled audience.
“The intelligence is in the data,” Slootman added.
Snowflake’s partnership with Nvidia may provide an alternative of sorts to companies concerned about ChatGPT and proprietary information. While ChatGPT soared in popularity when it was launched late last year, one complaint about the AI service is that once you put data in, that data is not “forgotten” — it stays in the AI as a part of its own learning.
Read: Snowflake expands Microsoft partnership to bring generative AI to its data cloud
Engineers at Samsung Electronics Co.
005930,
found out the hard way back in April, when they pasted lines of confidential code into ChatGPT to try to get the AI to fix a coding problem — but once put into ChatGPT, that data becomes the “property” of ChatGPT’s “brain.” Back in May, Samsung banned use of ChatGPT as a result.
Using Nvidia’s NeMo framework for AI developers, which was introduced back in September, Snowflake said its customers will be able to use the data in their Snowflake accounts to make custom large-language models, or LLMs, “for advanced generative AI services, including chatbots, search and summarization,” and have “the ability to customize LLMs without moving data enables proprietary information to remain fully secured and governed within the Snowflake platform.”
With Snowflake’s use of NeMo, customers are not sending data into a growing central AI, Manuvir Das, the head of Nvidia’s enterprise computing business, told MarketWatch at an earlier press conference. He likened the process to copy and paste: Once a customer starts putting in data, the customer is not using the original AI model but a copy, which they are now altering by adding data.
“Effectively, there’s now a copy of the model, and that copy only belongs to that customer,” Das explained. “Nothing that is happening in that copy is going back into the original foundation model.”
Back in March, ServiceNow Inc.
NOW,
teamed up with Nvidia on generative AI, and Das said that the company is already using NeMo “to train up custom models for each of their customers.” Nvidia’s Huang also showed in person at ServiceNow’s event for the announcement.
Read: Nvidia is the only near-term ‘beat and raise’ chip maker when it comes to AI, analyst says
Snowflake’s initial public offering was one of the biggest of 2020, as an investment from Warren Buffet’s Berkshire Hathaway
BRK.B,
and large sales growth wooed investors and drove its valuation higher than $100 billion. Snowflake’s market cap was less than $60 billion as of Monday’s close.
For the year, Nvidia shares are up 178%, while Snowflake’s have gained 18.3% and Microsoft’s are up 37%. The Dow Jones Industrial Average
DJIA,
which counts Microsoft among its 30 components, is up 1.7%, the S&P 500 index
SPX,
has advanced 12.7% and the tech-heavy Nasdaq Composite
COMP,
is up 27.4%.