Practical AI Getting in the Flow with Snorkel AI
Dec 21, 2020
Braden Hancock, Co-founder and Head of Technology at Snorkel AI and former Stanford PhD student, dives into innovative AI solutions. He discusses how noisier data sources can improve scalability and efficiency in labeling. Braden shares his transformative journey from mechanical engineering to founding Snorkel, emphasizing the practical application of machine learning. The conversation also highlights Snorkel Flow's advantages for managing AI pipelines and explores the ethical implications of AI, including auditability and privacy in training data.
AI Snips
Chapters
Transcript
Episode notes
Data Bottleneck
- Training data is the biggest bottleneck for machine learning applications.
- Deep learning models require vast amounts of labeled data.
Spam Detection Example
- Braden Hancock explains how Snorkel AI can be used for spam detection.
- Users can write labeling functions to identify spam indicators like certain keywords.
Scaling Data Labeling
- Snorkel's open-source library helps scale data labeling efforts.
- This allows for faster model creation by overcoming the data bottleneck.