Skynet for Beginners
summary
Builds a Ruby Twitter bot that replies with calming manatee images from manate dop, trains a small neural model (perceptron layers with perceptron and backprop learning), and measures outcomes: training needs thousands of iterations, public tweet corpus is noisy and produces poor sentiment labels, the learned classifier is unreliable, and a slip deploy floods a personal account (the infamous manatecalypse). Final result: simple rule-based replies work better than the undertrained network.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 1:37Neural Networks
- 3:54What is a Neural Network
- 5:02Threshold
- 6:59Neurons
- 8:54Threshold and Weight
- 9:40Network
- 12:46Weights
- 19:38Dont Panic
- 20:32The Story
- 21:41Manatee
- 21:57Manatee Dog Picks
- 22:26Twitter Bot
- 23:38Problems
- 24:31Learn your lesson
- 25:21Use the right tool
- 26:22Cute animals
related talks
49m"Small Data" Machine Learning
13mUsing JavaScript to Teach Machines Basic Shit Things
39mMachine learning failures - for art!
11mTalking to my past self (without introducing temporal paradoxes!)
01h 03mIt's All About AI
50mBuilding Continuous Learning Systems
48mKeynote – Sameer Singh
42mDōmo arigatō, Mr. Roboto: Machine Learning with Ruby