AI for FE
! Disclaimer !
"You can use my idea because I don't know how to build it"
"I wish I did more AI related things in University, I feel like I missed out"
What I will go through
- Approaching problems with AI
- What algorithms to google
- What you can do with surface knowledge
- Famous APIs for most problems
What I will NOT go through
- How does a neural network work
- How to build your own AI
- How to beat the best Go player in the world
- How famous algorithms work (idk this)
- How to win a hackathon with AI (idk either)
Why even bother to lern AI as FE devs?
- Entire stack in same language
- No need to worry about scaling
- JS is not much worse than Python
(or any other language for that matter)
What is AI?
Artificial Intelligence
Simulation of human thought processes in machines
Machine Learning
Finding patterns without being explicitly programmed (subset of AI)
What can you do with AI
- Categorise
- Make Decisions
- Generate
Approaching problems
How likely is an MR "LGTM-ed" without the reviewer reading it?
Categorising Problem
Categorising Problem
Classification
In a game of tetris, which is the best move?
Simplified Tetris
- Score = how many lines you clear
- No time limit
- You only know the current piece
Decision Making
Decision Making
Decision Making
Decision Making
Decision Making
Decision Making
Decision Making
"Score" for each coice:
Generate a speech by a famous man
Gumshoos
Content Generation
Content Generation
Content Generation
Content Generation
Content Generation
Content Generation
Fantastic algorithms and how to use them
Markov Process
K nearest Neighbours
Random Forest Classification
Neural Networks
Convolutional Neural Networks
Recurrent Neural Networks
Transfer Learning
Enough theory! Show us some cool shit!
Title Text
Background info
Background info
Where to go from here?
Datasets
- Brown Corpus
- Quick draw
- Imagenet
- MS-COCO
- MNIST
- Free Spoken Digit
- Free Music Archive
Libraries
- ML.js
- Compromise.js
- Tensorflowjs
- Kerasjs
- ML5