Functional Programming for Machine Learning Panel
summary
Panel debates whether functional programming will dominate ML, citing concrete tools: Scala and Haskell, Rust, TensorFlow (now offering imperative mode), Spark/Frameless and Jupyter notebooks as current UX. Proponents point to type-safe feature pipelines and FRP used for feature engineering; opponents point to fast notebook iteration, deep learning black boxes, and startups that push notebooks to production. Audience votes shift during the session and ultimately favors FP for ML, so the claim wins this room despite admitted trade-offs in performance and developer ergonomics.
AI-written summary · about
chapters (23)
- 0:00Introduction and panel setup
- 1:21Team introductions
- 4:09Initial vote and proposition
- 5:11Opening arguments for functional programming
- 7:36Opening arguments against functional programming
- 9:41Counterarguments from the for team
- 14:19Passion and vision from Oscar
- 16:43Recap of opening statements
- 19:33Debate on data scientists and tools
- 21:22Audience perspective and skepticism
- 24:15Historical parallels and industry trends
- 26:05Spark and functional programming challenges
- 27:50Audience questions and responses
- 33:31Deja vu and rule-based systems
- 36:44Governing dynamics of adoption
- 40:16Short-term vs long-term value
- 44:16Business value and tooling
- 47:09Logistic regression and functional programming
- 49:00Closing statements against proposition
- 49:41Closing statements for proposition
- 50:44Final arguments and trade-offs
- 55:00Why hasn't functional programming won yet
- 57:10Oscar's closing and final vote
related talks
40mFunctional Programming for Machine Learning and Data Pipelines
39mFunctional Programming And the Real World
38mFunctional Programming - The Architecture of Choice
32mDoes FP still matter in the era of AI?
44mFunctional Programming You Already Know
01h 28mThe Many Paths Towards Functional Programming Language Adoption
01h 34mTeaching functional programming in the ivory tower ... and in the kitchen
44mFunctional Programming: Failed Successfully