AI Assisted Development
summary
Defines trained vs direct functions, uses Lambda calculus to visualize the space of algorithms, and implements isPalNum three ways: direct s.reverse, a 100k-entry lookup of all 10‑digit palindromes, and a neural network with ~897 weights that converges in ~100 training hops and passes all 100k palindromes but not exhaustively tested against 9.9 billion non‑palindromes. Shows text→code (ChatGPT), GitHub Copilot for test bodies, and Diffblue’s RL test generation, and warns of the oracle problem for correctness.
AI-written summary · about
chapters (25)
- 0:00Introduction
- 0:32Definition
- 0:51Church Turing Thesis
- 2:00Lambda Calculus
- 3:44Space of All Possible Functions
- 12:19Requirement Spec
- 13:06Helper Function
- 13:50Actual Data Model
- 14:36Linear Regression
- 15:16Data vs Model
- 16:02Data Approach
- 17:32Neural Network Architecture
- 21:30Machine Learning Algorithm
- 23:22Test Driven Development
- 25:05Neural Network Demo
- 26:41Net Weights
- 27:28Bugs
- 27:40Star Journey
- 27:57Star Space
- 28:49When to use
- 31:09Text to Code
- 33:11Code Completion
- 34:06Text to Text
- 37:36Oracle Problem
- 39:22What do you think
related talks
43mBecome a 10x developer by harnessing the power of Generative AI
48mDyalog and AI
01h 16mAI in Software Development: Boosting Efficiency
35m300: The Battle of Thermopylae or Generative AI vs Developers
26mThe present and future of programming with AI (Keynote)
45mProgrammer’s Apprentice Season 2: Future Directions in AI-assisted Coding
50mIdentifying and fixing Issues in Code using AI based tools
58mAI, Determinism, Code Generation and Refactoring