Navigation Menu

Skip to content

Instantly share code, notes, and snippets.

View gcr's full-sized avatar

Kimmy gcr

  • Google AI
  • 03:21 (UTC -04:00)
View GitHub Profile
// cave.js
//
// contains ways of changing, generating, and displaying a cave.
// a cave is a 2d array.
//
// All these methods are side-effect-free builders.
//
var sys = require('sys');
function Cave(width, height, probability) {
pico-8 cartridge // http://www.pico-8.com
version 29
__lua__
-- kimmys-treasure-hunt-galaxy.p8
points={
0xfffe.ddfa,0x0009.d8e2,0xfffe.ace5,0x0001.0000,
0x0006.b93d,0xfffa.ab8c,0xfffa.dcf1,0x0001.0000,
0xfffa.5019,0x0007.7bd4,0x0003.69e8,0x0001.0000,
0xfffa.ca0e,0x0007.1a5f,0xfffb.4463,0x0001.0000,
@gcr
gcr / pi.c
Created January 1, 2012 01:31
/*
* Program to compute an approximation of pi
* by Brian Westley, 1988
* (requires pcc macro concatenation; try gcc -traditional-cpp)
*/
#define _ -F<00||--F-OO--;
int F=00,OO=00;
main(){F_OO();printf("%1.3f\n",4.*-F/OO/OO);}F_OO()
{
<!DOCTYPE HTML>
<html lang="en">
<head>
<meta charset="UTF-8">
<title>Bubble game</title>
<script type="text/javascript" src="bubbles.js"> </script>
</head>
<body>

ML in law symposium

Andreas’ idea: Given explainability / the ability to explain decisions, let’s maximize the performance we can get.

My three takeaways

  • Tech folks have a tendency to “fly in and fix everything.” That feels like a dangerous approach here. It’s far better to stand on the shoulders of existing legal precedent, which has studied fairness, discrimination, and bias for decades, even if that slows down progress.
  • Machine learning systems mirror and amplify bias by default. We cannot simply ignore sensitive attributes because the system averages loss over the majority. (Disparate mistreatment). Pithy corollary: this problem will only go away if we devote resources into making it go away.
  • Providing explanations for decisions is the only humane way to build automatic classification systems. Why? If I can’t test a result, I can’t contest it. If the decisions must be testable and explainable, they will be much more reliable as a result.
@gcr
gcr / keybase.md
Last active April 11, 2018 15:01

Keybase proof

I hereby claim:

  • I am gcr on github.
  • I am mwilber (https://keybase.io/mwilber) on keybase.
  • I have a public key whose fingerprint is 055A CA10 69B6 41A3 4AE1 C216 603D FE46 AFD3 DABD

To claim this, I am signing this object:

Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
@gcr
gcr / acat.png
Created August 28, 2012 05:01
Old CRT televisions for Racket!
acat.png
@gcr
gcr / rapid-image-viewer.rkt
Last active May 25, 2017 15:35
Tool for showing images
#lang racket
;;
;; Displays a directory full of images in rapid succession.
;; Handy for checking annotations or finding irregularities
;; in large-scale datasets.
;;
;; Usage:
;; - racket rapid-image-viewer.rkt [imglist]
;; Will show all of the '*.jpg' files given as lines in `imglist`,
Purchase WeekofPurchase StoreID PriceCH PriceMM DiscCH DiscMM SpecialCH SpecialMM LoyalCH SalePriceMM SalePriceCH PriceDiff Store7 PctDiscMM PctDiscCH ListPriceDiff STORE
1 CH 237 1 1.75 1.99 0 0 0 0 0.5 1.99 1.75 0.24 No 0 0 0.24 1
2 CH 239 1 1.75 1.99 0 0.3 0 1 0.6 1.69 1.75 -0.06 No 0.150754 0 0.24 1
3 CH 245 1 1.86 2.09 0.17 0 0 0 0.68 2.09 1.69 0.4 No 0 0.091398 0.23 1
4 MM 227 1 1.69 1.69 0 0 0 0 0.4 1.69 1.69 0 No 0 0 0 1
5 CH 228 7 1.69 1.69 0 0 0 0 0.956535 1.69 1.69 0 Yes 0 0 0 0
6 CH 230 7 1.69 1.99 0 0 0 1 0.965228 1.99 1.69 0.3 Yes 0 0 0.3 0
7 CH 232 7 1.69 1.99 0 0.4 1 1 0.972182 1.59 1.69 -0.1 Yes 0.201005 0 0.3 0
8 CH 234 7 1.75 1.99 0 0.4 1 0 0.977746 1.59 1.75 -0.16 Yes 0.201005 0 0.24 0
9 CH 235 7 1.75 1.99 0 0.4 0 0 0.982197 1.59 1.75 -0.16 Yes 0.201005 0 0.24 0