Artificial Intelligence Chris Piech CSBridge 2019 CSBridge 17 A - - PowerPoint PPT Presentation

▶
artificial intelligence
SMART_READER_LITE
LIVE PREVIEW

Artificial Intelligence Chris Piech CSBridge 2019 CSBridge 17 A - - PowerPoint PPT Presentation

Nov 25, 2022 255 likes •1.5k views

Artificial Intelligence Chris Piech CSBridge 2019 CSBridge 17 A Little AI CSBridge 17 Something big is happening in the world of AI CSBridge 17 Where is my robot? CSBridge 17 Sci-Fi Has Promised Me Robots CSBridge 17

slide-1
SLIDE 1

CSBridge ‘17

Artificial Intelligence

Chris Piech CSBridge 2019

slide-2
SLIDE 2

CSBridge ‘17

A Little AI

slide-3
SLIDE 3

CSBridge ‘17

Something big is happening in the world of AI…

slide-4
SLIDE 4

CSBridge ‘17

Where is my robot?

slide-5
SLIDE 5

CSBridge ‘17

Sci-Fi Has Promised Me Robots

slide-6
SLIDE 6

CSBridge ‘17

House Cleaning Robot

slide-7
SLIDE 7

CSBridge ‘17

House Cleaning Robot

slide-8
SLIDE 8

CSBridge ‘17

Robots? Body Mind

slide-9
SLIDE 9

CSBridge ‘17

Robots? Body Mind

slide-10
SLIDE 10

CSBridge ‘17

What is AI?

slide-11
SLIDE 11

CSBridge ‘17

[suspense]

slide-12
SLIDE 12

CSBridge ‘17

AI: The study and design of intelligent agents

slide-13
SLIDE 13

CSBridge ‘17

AI: The study and design of intelligent agents

Computer programs Better than chance As well as humans

slide-14
SLIDE 14

CSBridge ‘17

Narrow Intelligence

Play Chess Translate Turkish Drive a Car Play Breakout

slide-15
SLIDE 15

CSBridge ‘17

General Intelligence

Play Chess Translate Turkish Drive a Car Play Breakout

slide-16
SLIDE 16

CSBridge ‘17

Brief History

slide-17
SLIDE 17

CSBridge ‘17

1952

Early Optimism 1950s

slide-18
SLIDE 18

CSBridge ‘17

Early Optimism 1950s “Machines will be capable, within twenty years, of doing any work a man can do.” –Herbert Simon, 1952

slide-19
SLIDE 19

CSBridge ‘17

The world is too complex

Underwhelming Results 1950s to 1980s

slide-20
SLIDE 20

CSBridge ‘17

slide-21
SLIDE 21

CSBridge ‘17

Big Milestones

slide-22
SLIDE 22

CSBridge ‘17

Almost perfect… Told Speech Was 30 Years Out

slide-23
SLIDE 23

CSBridge ‘17

The Last Remaining Board Game

slide-24
SLIDE 24

CSBridge ‘17

Computers Making Art

slide-25
SLIDE 25

CSBridge ‘17

Self Driving Cars

slide-26
SLIDE 26

CSBridge ‘17

What is going on?

slide-27
SLIDE 27

CSBridge ‘17

[more suspense]

slide-28
SLIDE 28

CSBridge ‘17

Story of Modern AI

slide-29
SLIDE 29

CSBridge ‘17

Focus on one problem

slide-30
SLIDE 30

CSBridge ‘17

Make a Harry Potter Sorting Hat

slide-31
SLIDE 31

CSBridge ‘17 Logistic Regression is like the Harry Pottery Sorting Hat

Classification

That is a picture

  • f a one
slide-32
SLIDE 32

CSBridge ‘17 Logistic Regression is like the Harry Pottery Sorting Hat

Classification

That is a picture

  • f a zero
slide-33
SLIDE 33

CSBridge ‘17

Classification

That is a picture

  • f an zero

* It doesn’t have to be correct all of the time

slide-34
SLIDE 34

CSBridge ‘17

Can you do it?

slide-35
SLIDE 35

CSBridge ‘17

slide-36
SLIDE 36

CSBridge ‘17

slide-37
SLIDE 37

CSBridge ‘17 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

How about now?

What a computer sees What a human sees

slide-38
SLIDE 38

CSBridge ‘17

Why is it easy for Humans?

About 30% of your cortex is used from vision 3% is used to process hearing

slide-39
SLIDE 39

CSBridge ‘17

Very hard to Program

public class HarryHat extends ConsoleProgram { public void run() { println(“Todo: Write program”); } }

??

slide-40
SLIDE 40

CSBridge ‘17

Perhaps there is an insight?

slide-41
SLIDE 41

CSBridge ‘17

Two Great Ideas

  • 1. Artificial Neurons
  • 2. Learn by Example
slide-42
SLIDE 42

CSBridge ‘17

Two Great Ideas

  • 1. Artificial Neurons
  • 2. Learn by Example
slide-43
SLIDE 43

CSBridge ‘17

  • 1. Artificial Neurons
slide-44
SLIDE 44

CSBridge ‘17

Neuron

slide-45
SLIDE 45

CSBridge ‘17

Neuron

slide-46
SLIDE 46

CSBridge ‘17

Neuron

slide-47
SLIDE 47

CSBridge ‘17

Neuron

slide-48
SLIDE 48

CSBridge ‘17

Neuron

slide-49
SLIDE 49

CSBridge ‘17

Some Inputs are More Important

slide-50
SLIDE 50

CSBridge ‘17

Artificial Neuron

+

slide-51
SLIDE 51

CSBridge ‘17

+

Inputs

Numbers between 0 and 1

slide-52
SLIDE 52

CSBridge ‘17

+

Inputs

Numbers between 0 and 1

slide-53
SLIDE 53

CSBridge ‘17

+

Inputs

Numbers between 0 and 1

slide-54
SLIDE 54

CSBridge ‘17

+

Weights

Negative or positive numbers

slide-55
SLIDE 55

CSBridge ‘17

+

Weights

slide-56
SLIDE 56

CSBridge ‘17

+

Weights

slide-57
SLIDE 57

CSBridge ‘17

+

Weighted Sum

double weightedSum = 0; weightedSum = weightedSum + input0 * weight0; weightedSum = weightedSum + input1 * weight1; weightedSum = weightedSum + input2 * weight2; weightedSum = weightedSum + input3 * weight3;

slide-58
SLIDE 58

CSBridge ‘17

+

Filter and Output

slide-59
SLIDE 59

CSBridge ‘17

Java Demo

slide-60
SLIDE 60

CSBridge ‘17

  • A neuron
  • Your brain

Actually, it’s probably someone else’s brain

Biological Basis for Neural Networks

+

???

slide-61
SLIDE 61

CSBridge ‘17

+ + + +

Put Many Together

slide-62
SLIDE 62

CSBridge ‘17

Put Many Together

Input Neurons Hidden Neurons Output Neurons

slide-63
SLIDE 63

CSBridge ‘17

Making a Prediction

… Input Neurons Hidden Neurons Output Neurons

slide-64
SLIDE 64

CSBridge ‘17 … Input Neurons

Making a Prediction

Hidden Neurons Output Neurons

slide-65
SLIDE 65

CSBridge ‘17 …

Making a Prediction

Input Neurons Hidden Neurons Output Neurons

slide-66
SLIDE 66

CSBridge ‘17 …

Making a Prediction

Input Neurons Hidden Neurons Output Neurons I think that is a picture of a

  • ne!
slide-67
SLIDE 67

CSBridge ‘17

Demonstration

http://scs.ryerson.ca/~aharley/vis/conv/

slide-68
SLIDE 68

CSBridge ‘17

+

Great Idea: Artificial Neurons

slide-69
SLIDE 69

CSBridge ‘17

Neural Networks get their intelligence from their sliders (parameters)

slide-70
SLIDE 70

CSBridge ‘17

Two Great Ideas

  • 1. Artificial Neurons
  • 2. Learn by Example
slide-71
SLIDE 71

CSBridge ‘17

Two Great Ideas

  • 1. Artificial Neurons
  • 2. Learn by Example
slide-72
SLIDE 72

CSBridge ‘17

  • 2. Learn From Experience
slide-73
SLIDE 73

CSBridge ‘17

Learn by Example

slide-74
SLIDE 74

CSBridge ‘17

+

slide-75
SLIDE 75

CSBridge ‘17

+

slide-76
SLIDE 76

CSBridge ‘17

+

I think that is a picture of a One! What do you mean it’ s actually a Zero? I’ll adjust my sliders so that I do a better job in the future

slide-77
SLIDE 77

CSBridge ‘17

+

I think that is a picture of a One! What do you mean it’ s actually a Zero? I’ll adjust my sliders so that I do a better job in the future

slide-78
SLIDE 78

CSBridge ‘17

+

slide-79
SLIDE 79

CSBridge ‘17

+

slide-80
SLIDE 80

CSBridge ‘17

+

slide-81
SLIDE 81

CSBridge ‘17

+

I think that is a picture of a One! Wahoo I got it right!

slide-82
SLIDE 82

CSBridge ‘17

+

slide-83
SLIDE 83

CSBridge ‘17

+

slide-84
SLIDE 84

CSBridge ‘17

+

I think that is a picture of a zero! What do you mean it’ s actually a one? I’ll adjust my sliders so that I do a better job in the future

slide-85
SLIDE 85

CSBridge ‘17

+

I think that is a picture of a zero! What do you mean it’ s actually a one? I’ll adjust my sliders so that I do a better job in the future

slide-86
SLIDE 86

CSBridge ‘17

Study Hard!

slide-87
SLIDE 87

CSBridge ‘17

pixels edges

  • bject parts

(combination

  • f edges)
  • bject models

[Honglak Lee]

Training set: Aligned images of faces.

Visualize the Sliders

slide-88
SLIDE 88

CSBridge ‘17

Woah… that’s like a brain…

slide-89
SLIDE 89

CSBridge ‘17

True.

slide-90
SLIDE 90

CSBridge ‘17

Decomposition

Mountain Karel Climb Up Mountain Climb Down Mountain Pick Beeper Step Up Step Down

slide-91
SLIDE 91

CSBridge ‘17

… smoothhound, smoothhound shark, Mustelus mustelus American smooth dogfish, Mustelus canis Florida smoothhound, Mustelus norrisi whitetip shark, reef whitetip shark, Triaenodon obseus Atlantic spiny dogfish, Squalus acanthias Pacific spiny dogfish, Squalus suckleyi hammerhead, hammerhead shark smooth hammerhead, Sphyrna zygaena smalleye hammerhead, Sphyrna tudes shovelhead, bonnethead, bonnet shark, Sphyrna tiburo angel shark, angelfish, Squatina squatina, monkfish electric ray, crampfish, numbfish, torpedo smalltooth sawfish, Pristis pectinatus guitarfish roughtail stingray, Dasyatis centroura butterfly ray eagle ray spotted eagle ray, spotted ray, Aetobatus narinari cownose ray, cow-nosed ray, Rhinoptera bonasus manta, manta ray, devilfish Atlantic manta, Manta birostris devil ray, Mobula hypostoma grey skate, gray skate, Raja batis little skate, Raja erinacea …

Stingray Mantaray

Image Net Classification

slide-92
SLIDE 92

CSBridge ‘17

0.005%

Random guess

1.5% ?

Le, et al., Building high-level features using large-scale unsupervised learning. ICML 2012

Pre Neural Networks GoogLeNet

slide-93
SLIDE 93

CSBridge ‘17

0.005%

Random guess

1.5%

Pre Neural Networks 43.9% GoogLeNet

Szegedy et al, Going Deeper With Convolutions, CVPR 2015

slide-94
SLIDE 94

CSBridge ‘17

0.005%

Random guess

1.5%

Pre Neural Networks 66.3% 2016

http://image-net.org/challenges/LSVRC/2016/results

slide-95
SLIDE 95

CSBridge ‘17

slide-96
SLIDE 96

CSBridge ‘17

Google Brain

slide-97
SLIDE 97

CSBridge ‘17

1 Trillion Artificial Neurons

Google Brain

slide-98
SLIDE 98

CSBridge ‘17 Optimal stimulus by numerical optimization Le, et al., Building high-level features using large-scale unsupervised learning. ICML 2012 Top stimuli from the test set

A Neuron That Fires When It Sees Cats

slide-99
SLIDE 99

CSBridge ‘17

slide-100
SLIDE 100

CSBridge ‘17 Neuron 1 Neuron 2 Neuron 3 Neuron 4 Neuron 5 Le, et al., Building high-level features using large-scale unsupervised learning. ICML 2012

Other Neurons

slide-101
SLIDE 101

CSBridge ‘17

An algorithm learned to detect skin cancer from photo, better than the worlds top expert. Developed last year

Esteva et al., Nature 2017. Esteva, Andre, et al. "Dermatologist-level classification of skin cancer with deep neural networks." Nature 542.7639 (2017): 115-118.

It can be useful

slide-102
SLIDE 102

CSBridge ‘17 run cond body putBeeper putBeeper move

// User defined method private void run() { while(isClear()){ putBeeper(); move(); } putBeeper(); }

while It looks like you have a fencepost error!

Helping Students Learn to Program

slide-103
SLIDE 103

CSBridge ‘17

Beyond Harry Potter Hats

http://cs.stanford.edu/people/karpathy/convnetjs/demo/rldemo.html

slide-104
SLIDE 104

CSBridge ‘17

más allá de la clasificación

slide-105
SLIDE 105

CSBridge ‘17 Performance Neural Networks 2012 Time Now Human Intelligence Complex Programs

The Future of AI

slide-106
SLIDE 106

CSBridge ‘17 Performance Complex Programs Neural Networks 2012 Time Now Human Intelligence

The Future of AI

slide-107
SLIDE 107

CSBridge ‘17 2012 Performance Human Intelligence Neural Networks Now Time Complex Programs

The Future of AI

slide-108
SLIDE 108

CSBridge ‘17 2012 Performance Human Intelligence Einstein Neural Networks Now Time Complex Programs

The Future of AI

slide-109
SLIDE 109

CSBridge ‘17

Open Problems?

slide-110
SLIDE 110

CSBridge ‘17

Machine Learning Uses a Lot of Data

slide-111
SLIDE 111

CSBridge ‘17

One Shot Learning

Single training example: Test set:

slide-112
SLIDE 112

CSBridge ‘17

One Shot Learning

Single training example:

slide-113
SLIDE 113

CSBridge ‘17

Where is my robot?

slide-114
SLIDE 114

CSBridge ‘17

… coming soon

slide-115
SLIDE 115

CSBridge ‘17

Should I study AI?

slide-116
SLIDE 116

CSBridge ‘17

Powerful technology

slide-117
SLIDE 117

CSBridge ‘17

Now is an amazing time

slide-118
SLIDE 118

CSBridge ‘17

Know It So You Can Beat It

Little math

slide-119
SLIDE 119

CSBridge ‘17

Todos son bienvenidos

slide-120
SLIDE 120

CSBridge ‘17

Road towards AI

Libraries to use AI Mathematics to invent AI

slide-121
SLIDE 121

CSBridge ‘17

Next Step:

…

slide-122
SLIDE 122

CSBridge ‘17

slide-123
SLIDE 123

CSBridge ‘17

The End?