Natural Language-Guided Programming
summary
Proposes enriching code with plain-English intent to drive completion and implements a prototype for Python data science that fine-tunes GPT-2 on ~300,000 Jupyter notebooks with three comment strategies (natural, docstring-inserted, none). Extracts 200 hand-validated test cases and gets human ratings showing the natural and docstring models produce about two thirds useful predictions while the no-comments model fails, notes Codex/Copilot outperform but at far greater scale and cost, and releases models, benchmarks and annotations.
AI-written summary · about
chapters (10)
- 0:00Intro
- 0:39Motivation
- 2:04Natural Language-guided Programming (NLGP)
- 3:20Language models
- 5:11Prototyping an NLGP assistant for Python
- 8:37How good are the models?
- 9:21Creating a benchmark
- 10:36Evaluating the models
- 12:32Recent related work
- 13:58Conclusion
related talks
17mNatural Language Processing in Automation
24mNatural Language Coding Revolutionising Programming with GenAI
45mBringing the Power of AI to Your Application
36mUnleashing Generative AI: Improving Developer Productivity in SRE
01h 00mGoing beyond ChatGPT: introduction to prompt engineering & LLMs
43mBecome a 10x developer by harnessing the power of Generative AI
01h 01mC++hatGPT: How AI Tools Impact Your Life as a Systems Programmer
36mGoing Beyond Copilot with AI Agents
