Snowflake is bringing Artificial Intelligence (AI) to its cloud-based data platform to help enterprises make the most of the huge amounts of unstructured data they hold.
At its annual user conference, Snowflake Summit 2023 in Las Vegas, the company announced new features for its unified data platform, including a Large Language Model (LLM) that it hopes customers will use to make the most of their information.
In his keynote speech to introduce the event, Frank Slootman, chairman and CEO at Snowflake, conceded that the hype around AI is strong, but argued that his own company must put a stake in the ground for AI and LLMs:
With all the talk of AI, it’s almost like they’re the only two letters left in the alphabet, so I will be sparing with my use of them. But we also like to say that you need a data strategy to have an AI strategy. No surprise, then, that we see our Data Cloud as essential to enabling AI and LLMs and deriving real value from it.
Snowflake’s LLM is built using Applica’s generative AI technology. Snowflake purchased Applica in September 2022 and is using the AI specialist’s multimodal LLM in some of its new products, including Document AI, which is currently in private preview and uses AI to help customers find insight in unstructured data, explained Slootman:
Suffice to say that we are super-excited about the potential because it lets us mobilize data in ways we could only dream of not too long ago.
IDC research suggests about 90% of global data created during the next five years will be unstructured in the form of documents, images, video and audio. Snowflake’s platform will include Document AI, which is pitched at making it easier for organizations to extract value from documents using natural language processing.
Native Application Framework
It was also announced at the event that the Snowflake Native Application Framework is now available for developers to build and test their own apps. Snowflake Native Apps run directly within customers’ Snowflake accounts. Slootman said the innovation is a big step forwards – and likened the development to the creation of an app store for Snowflake users:
The highest-level analogy that I can give you for this is the Apple iPhone and the App Store. There is now such a thing as a Snowflake application. We're super-excited about that because they're using all our common services – they’re using the governance frameworks and our database engine, which is the centrepiece of the entire platform.
Apps can run inside an enterprise user’s existing Snowflake account, which means customers no longer have to export or provide external access to data. The company hopes its customers will use these Apps to unlock new revenue streams, without having to move or expose their data, explained Slootman:
The great thing here, on our Data Cloud, is you build one application and it can run unmodified. So if you're a software developer, you can address the entire environment with a single implementation and we think that’s incredibly cool.
More than 25 new Snowflake Native Apps are now available for customers to install from Snowflake Marketplace, which is a collaboration space that allows companies to exploit a variety of third-party datasets. In addition, over 100 providers are developing apps:
There's going to be hundreds of apps. And, if we're successful, there’s going to be thousands. This is a core part of our strategy. Apps are not just limited to application developers and software companies. Everybody these days is a software developer; everybody is building customer-facing applications and internal applications.
Snowpark Container Services
Snowflake also announced the launch of Snowpark Container Services, which expands the tech company’s compute infrastructure to run a variety of workloads, including full-stack applications and the hosting of LLMs. Slootman said the move to containers is a huge expansion of Snowpark, which will give users more options:
When we first introduced Snowpark, a lot of people were saying that it's great for Spark jobs. But a lot of people were saying, ‘Wait a second. I have engines and I have applications and I don't even know what to do with these things.’ That's why we developed containerization. We can take whole applications and services and we can run them inside the Snowflake governance parameter.
Using Snowpark, developers can now unlock broader infrastructure options, including accelerated computing with NVIDIA GPUs and AI software. Snowflake has partnered with companies such as NVIDIA, Alteryx, Astronomer, Dataiku, Hex, and SAS to give customers secure access to their products using Snowpark Container Services:
It's a great strategy because we don't have to port things – instead, they’re going to go wholesale into the container and then they can be addressed by the applications themselves. It’s a great idea because we are anticipating very rapid development in these areas and we need a strategy to adopt that shift and deliver it to the world.
More on data
Finally, Snowflake announced updates to Iceberg Tables. The tech company says increasing number of companies are using Apache Iceberg as the industry standard for open table formats. Slootman said Snowflake wants to make it easier for enterprises to use the Data Cloud to extract value from Iceberg data:
The goal here is that we want you to be able to reference any and all data. We're putting the infrastructure in the Data Cloud to make sure there are no limits to your ability to reference and trust data, especially given all the developments in data science. That’s super important.
With Iceberg Tables, organisations can work with data in their own storage in the Apache Iceberg format, whether that data is managed by Snowflake or managed externally. Slootman pitched that this work on Iceberg Tables – as well as the company’s other announcements – will help the company keep delivering data-led benefits to major enterprises:
We’re pushing the limits back on applications, we’re pushing the limits back on the use of data, and we’re pushing the limits back on Snowflake. The whole strategy here is for the Data Cloud to not have any limitations – and you have our word for it.