Overview
Explore how machine learning is revolutionizing software development in this 59-minute Stanford HAI seminar. Delve into the concept of "Software 2.0" with Stanford associate professor Chris Re as he discusses the radical shift from conventional programming to systems that learn from high-level domain knowledge and vast amounts of data. Examine foundational challenges in weak supervision theory, self-supervised system guidance, and high-level abstractions for long-term system monitoring. Gain insights from real-world applications of systems like Snorkel, Overton, and Bootleg in major tech companies. Cover topics including AI engineering, quality monitoring, new challenges in deep learning, and the implications of Software 2.0 on bias and measuring quality in AI systems.
Syllabus
Introduction
Context
Models as a commodity
AI Engineering
New Modelitis
Monitoring Quality
Challenges
Potentially Controversial Claims
Overton Example
The Tail
New Challenges
Examples
DeepNets
Conclusion
Last Minute Questions
Software 20 Bias
Fire Yourself
Measuring Quality
AI Index Report
Taught by
Stanford HAI