Snowflake, a leader in data, is of the most important enterprise software companies of our generation. What role does it have to play in the world of

Snowflake CEO Sridhar Ramaswamy on Using Data to Create Simple, Reliable AI for Businesses

submited by

Style Pass

2024-10-08 16:00:04

Snowflake, a leader in data, is of the most important enterprise software companies of our generation. What role does it have to play in the world of AI? Snowflake is integrating AI into its core to transform unstructured data, democratize data access and improve functionality, and create reliable AI applications. In this episode CEO Sridhar Ramaswamy describes how the company has been able to build reliable talk-to-your-data business applications with 90%+ accuracy, whereas even frontier models achieve ~45% off the shelf.

Sridhar Ramaswamy, CEO of Snowflake and former head of Google’s advertising business, on why he thinks data is key to creating reliable AI for business use cases.

Reliability and precision are critical for enterprise AI applications. Even with advanced models like GPT-4, out-of-the-box solutions often struggle with accuracy, achieving only about 45% reliability for tasks like answering questions about company data. Snowflake has pushed this to over 90% for their talk-to-your-data applications by treating it as a software engineering problem rather than purely an AI model problem. This approach involves carefully constraining the problem space and applying rigorous software development practices.

AI founders should focus on making complex tasks simple for end users. Ramaswamy emphasizes that most enterprise customers want to solve business problems, not grapple with complex technology. By turning multi-week software engineering projects into simple commands that analysts can execute in hours, founders can create significant value. This approach of “making the hard simple” can be a powerful differentiator.