Leveraging Data to Unlock AI's True Potential in Enterprise Projects

Discover how to maximize AI's potential through effective data leveraging in this 30-minute talk by Snorkel AI CEO Alex Ratner. Explore the critical role of data development in enterprise AI projects and learn why fine-tuning large language models (LLMs) is essential for achieving production-grade performance on complex, domain-specific challenges. Gain insights into Snorkel Flow, a tool that empowers data science teams to rapidly develop high-quality datasets by amplifying the impact of subject matter experts. Examine two compelling case studies demonstrating Snorkel Flow's effectiveness: one showcasing the creation of a smaller, more accurate model using Google's PaLM 2 as a baseline, and another highlighting a novel training approach that achieved a top ranking on the AlpacaEval leaderboard. Delve into topics such as foundation models, customization, specialization, accuracy, use cases, data-centric development, and the data comp stack. Access accompanying slides and additional resources to further enhance your understanding of leveraging data for unlocking AI's true potential in enterprise settings.

Intro
Welcome
Outline
My POV
Foundation models
Customization
Specialization
Accuracy
Use cases
DataCentric development
Data development
Snorkel
Data comp
Stack
Snorkel Flow
Palm 2 Guess
Example
Tuning multiple components