Snowflake cortex LLM functions

What is Snowflake Cortex?

Generative AI is increasing productivity, creating new opportunities for innovation, and gaining more value from data. It is imperative that generative AI be accessible to and used by everyone, not just AI experts, in order to fully realize this promise. Users of all skill levels can benefit from industry-leading AI models, LLMs, and vector search features with Snowflake Cortex, in addition to comprehensive LLM-powered experiences. Regardless of their level of technical proficiency, these developments allow all Snowflake users to safely harness the potential of generative AI to extract dynamic insights from their company data.Cortex offers pre-built ML functions for tasks like Getting stocks price and access to industry-leading large language models (LLMs) for working with unstructured data.

LLM Functions:

COMPLETE

Like other LLMs, COMPLETE simply accepts a prompt from the user (along with the model that will be used) and generates a response from the model. All of it is done, though, using SQL statements. Using the snowflake-arctic paradigm, the following is a syntax:

					SELECT SNOWFLAKE.CORTEX.COMPLETE( ‘llama2-70b-chat’, ‘what is cortex?’ )

				

Note: llama2-70b-chat is one of the models used for testing you can install of your choice.

EXTRACT_ANSWER

EXTRACT_ANSWER will answer a question based on a text document in plain English or as a string representation of JSON. Here is the syntax and example:

					SELECT SNOWFLAKE.CORTEX.EXTRACT_ANSWER(blogs_nash, ‘what is databricks’)

				

Note: Source document should be a string containing the plain-text or JSON document that contains the answer to the question.

EMBED_TEXT_768

Any unstructured data can be fed into EMBED_TEXT_768 to produce an embedded vector. After that, these vectors can be compared for commonalities with other applications. Once again, the vector is created using a simple SQL command. Here is the syntax and an example:

					SELECT SNOWFLAKE.CORTEX.EMBED_TEXT_768(‘llama2-70b-chat’, ‘Can KKR win the IPL?’)

				

SUMMARIZE

SUMMARIZE returns a summary of the given data. Here is the following syntax and an example:

					SELECT SNOWFLAKE.CORTEX.SUMMARIZE(blogs_nash) from blogs_nash
				

SENTIMENT

SENTIMENT provides a floating-point value ranging from -1 to 1, representing the degree of negativity in the text, 0 for being neutral, and 1 for the highest level of positivity. Here is the following syntax and an example:

					
SELECT SNOWFLAKE.CORTEX.SENTIMENT(blogs_nash)

				

TRANSLATE

At the most TRANSLATE will translate text from one language to another. Here is the following syntax and an example:

					SELECT SNOWFLAKE.CORTEX.TRANSLATE( ‘سعيد بلقائك’, ‘fr’, ‘en’)
				

Conclusion

Hence Snowflake Cortex empowers users of all technical levels to leverage generative AI for dynamic insights, enhancing productivity and innovation across organizations. By providing access to powerful AI models and user-friendly functionalities through simple SQL commands, Snowflake Cortex democratizes AI, making advanced data analysis and application development accessible to everyone.