It's Alive! Machine Learning Writes Your Code! 🧠🔮
summary
Claim: a model trained on all open-source TypeScript from GitHub learns to predict function bodies from signatures. Pipeline uses BigQuery to collect code, AST-based tokenization and placeholdering for names/strings, a TensorFlow model predicting function bodies (sequence-to-sequence), and one-hot outputs. Result: a demoed VS Code extension built with TensorFlowJS that autocompletes bodies; sometimes it generates sensible implementations, other times buggy or nonsensical output, but the system works end-to-end.
AI-written summary · about
chapters (8)
- 0:00Intro
- 6:38Neural Networks
- 10:00Machine Learning Approach
- 13:50Big Query
- 14:40Cleaning the Data
- 19:05Tokenization
- 20:30Translation
- 27:30Tensorflow
related talks
34mIt's Alive! Machine Learning Writes Your Code! 🧠
35mIt's Alive! Machine Learning Writes Your Code!
37mMachine Learning on Source Code
01h 05mSelf-Writing Code
39mMachine Learning Magic for Your JavaScript Application
36mLet Your Angular Code Write Itself
44mA Deep Dive Into Advanced TypeScript: A Live-Coding Expedition
22mIncreasing developer productivity with Generative AI