Let’s start with a feature with one of the highest impact: LLM functions. We’ll introduce it with a practical example:
Imagine you are selling a product on Amazon with a few 1 ⭐ ratings, and you want to understand why customers are disappointed.
You start reading the reviews, and some of them are nonsense, like this one:
How will you know if it is an outlier, or you’re target of a silly joke?
To understand what your users think, the only way is to analyze the reviews, but doing so with traditional methods would be quite a task.
With Cortex LLM Functions you can identify patterns automatically, and it can be as simple as this:
While we won’t go in the details of what each LLM function does (SUMMARIZE is one of them, but there are more), the gist of it is that they enable the use of natural language to interact with and understand unstructured data.
Other things that can be done are, for instance:
- Translate text
- Perform sentiment analysis (is this comment positive or negative?)
- Answer questions
- … and more
This greatly expands what a data warehouse can do.
It turns it from a tool mainly for analyzing structured data, into a versatile solution that meets many more analytical needs than before.
In the following chapters, we will keep exploring other features of Cortex.
Next in line — Snowflake Copilot.
