74
Talk nerdy to me How the future of UX is conversation and bots Brian Lee Yung Rowe November 5, 2016 Founder + CEO // Pez.AI Incorporated Adjunct Professor // CUNY SPS Master of Data Analytics Adjunct Professor // Baruch Master of Financial Engineering

Talk nerdy to me: how the future of UX is conversation and bots by Brian Rowe

Embed Size (px)

Citation preview

Talk nerdy to me

How the future of UX is conversation and bots

Brian Lee Yung Rowe

November 5, 2016

Founder + CEO // Pez.AI IncorporatedAdjunct Professor // CUNY SPS Master of Data AnalyticsAdjunct Professor // Baruch Master of Financial Engineering

ABOUT BRIAN LEE YUNG ROWE

Founder and CEO of Pez.AI

Adj. Professor, CUNY Master’s in Data Analytics

Adj. Professor, Baruch Master’s in Financial Engineering

Blog at http://cartesianfaith.com @cartesianfaith

Brian Lee Yung Rowe 2

INTERFACE

What is an interface?

Brian Lee Yung Rowe 3

INTERFACE: A DEFINITION

User interfaces dictate howwe interact with the world

around us

Brian Lee Yung Rowe 4

INTERACTION

What is interaction?

Brian Lee Yung Rowe 5

INTERACTION: LITERARY CONFLICT

Human vs Nature

Human vs Human Human vs Self

Brian Lee Yung Rowe 6

INTERACTION: LITERARY CONFLICT

Human vs Nature Human vs Human

Human vs Self

Brian Lee Yung Rowe 7

INTERACTION: LITERARY CONFLICT

Human vs Nature Human vs Human Human vs Self

Brian Lee Yung Rowe 8

INTERACTION: THE PLOT THICKENS

Brian Lee Yung Rowe 9

INTERACTION: NOT JUST CONFLICT

Human <3 Nature Human <3 Human

Human <3 Self Human <3 Machine

Brian Lee Yung Rowe 10

INTERACTION: NOT JUST CONFLICT

Human <3 Nature Human <3 Human

Human <3 Self Human <3 Machine

Brian Lee Yung Rowe 11

HUMAN INTERACTION

What is the interface forhuman

interaction/communication?

Brian Lee Yung Rowe 12

HUMAN INTERACTION: LANGUAGE AS AN INTERFACE

Language is the ultimateinterface

More than 7 billion users

Brian Lee Yung Rowe 13

HUMAN INTERACTION: LANGUAGE AS AN INTERFACE

Language is the ultimateinterface

More than 7 billion users

Brian Lee Yung Rowe 14

HUMAN INTERACTION: LANGUAGE AS AN INTERFACE

Spoken

Written

w − η∑n

i=1∇Qi(w)Symbolic Emoji

Brian Lee Yung Rowe 15

HUMAN INTERACTION: LANGUAGE AS AN INTERFACE

Spoken Written

w − η∑n

i=1∇Qi(w)Symbolic Emoji

Brian Lee Yung Rowe 16

HUMAN INTERACTION: LANGUAGE AS AN INTERFACE

Spoken Written

w − η∑n

i=1∇Qi(w)Symbolic

Emoji

Brian Lee Yung Rowe 17

HUMAN INTERACTION: LANGUAGE AS AN INTERFACE

Spoken Written

w − η∑n

i=1∇Qi(w)Symbolic Emoji

Brian Lee Yung Rowe 18

LANGUAGE

Why is language popular?

Brian Lee Yung Rowe 19

LANGUAGE: FEATURES

• Scales to billions of users

• Extends to multiple domains

• Natural personalization

• Intuitive and easy to use

Brian Lee Yung Rowe 20

LANGUAGE: PERSONALIZATION

Say things your way

Brian Lee Yung Rowe 21

LANGUAGE: PERSONALIZATION

Say things your way

Brian Lee Yung Rowe 22

LANGUAGE: PERSONALIZATION

Say things your way

Brian Lee Yung Rowe 23

LANGUAGE: EASE OF USE

Change your mind in 0 taps/clicks

Brian Lee Yung Rowe 24

LANGUAGE: EASE OF USE

Change your mind in 0 taps/clicks

Brian Lee Yung Rowe 25

LANGUAGE: EASE OF USE

Change your mind in 0 taps/clicks

Brian Lee Yung Rowe 26

MACHINE INTERACTION

How do we interact withmachines (software)?

Brian Lee Yung Rowe 27

MACHINE INTERACTION: GUIS

Brian Lee Yung Rowe 28

MACHINE INTERACTION: GUIS

Brian Lee Yung Rowe 29

MACHINE INTERACTION: GUIS

Brian Lee Yung Rowe 30

GUIS: COMPLICATED

Brian Lee Yung Rowe 31

GUIS: COMPLICATED

Brian Lee Yung Rowe 32

GUIS: UNNATURAL

Gestures are often unintuitive

and require training

Brian Lee Yung Rowe 33

GUIS: UNNATURAL

Gestures are often unintuitiveand require training

Brian Lee Yung Rowe 34

GUIS: INTERFACE OPTIMIZATION

GUIs are rigid

and takeaway choice

Brian Lee Yung Rowe 35

GUIS: INTERFACE OPTIMIZATION

GUIs are rigid

and takeaway choice

Brian Lee Yung Rowe 36

MESSAGING

Can we use languageinstead?

YES NO

Brian Lee Yung Rowe 37

MESSAGING

Can we use languageinstead?

YES

NO

Brian Lee Yung Rowe 38

MESSAGING

...via text messaging

Brian Lee Yung Rowe 39

MESSAGING: BENEFITS

• Text-messaging is pervasive

• Abstracts counterpart at other end of the line

• More flexible than GUIs in small space

• Minimizes data costs for users

Brian Lee Yung Rowe 40

MESSAGING: PERVASIVENESS

App Monthly Active Users

WhatsApp > 1 billionFacebook Messenger > 1 billionWeChat > 800 millionViber > 780 millionLine > 220 million

Brian Lee Yung Rowe 41

MESSAGING: ABSTRACTION

Messaging (disinter)mediates human interaction

Brian Lee Yung Rowe 42

MESSAGING: ABSTRACTION

Counterpart can be human and/or bot (machine)

Brian Lee Yung Rowe 43

MESSAGING: ABSTRACTION

Turing Test Which use humans?

Pez.AI X.ai Facebook M Magic

Brian Lee Yung Rowe 44

MESSAGING: THE BOT ADVANTAGE

Emulate the best aspects of human interaction

Brian Lee Yung Rowe 45

MESSAGING: THE BOT ADVANTAGE

...without the drawbacks

happy angry sad tired hungry

human

bot

Brian Lee Yung Rowe 46

MESSAGING: THE BOT ADVANTAGE

...without the drawbacks

happy angry sad tired hungry

human

bot

Brian Lee Yung Rowe 47

MESSAGING: THE BOT ADVANTAGE

Tirelessly does tedious, repetitive work

so humanscan do interesting, creative, challenging work

Brian Lee Yung Rowe 48

MESSAGING: THE BOT ADVANTAGE

Tirelessly does tedious, repetitive work so humanscan do interesting, creative, challenging work

Brian Lee Yung Rowe 49

CONVERSATION DESIGN

What makes goodconversation?

Brian Lee Yung Rowe 50

CONVERSATION DESIGN: BEST PRACTICES

Make machines more human-like

but not too much

Brian Lee Yung Rowe 51

CONVERSATION DESIGN: BEST PRACTICES

Make machines more human-like but not too much

Brian Lee Yung Rowe 52

CONVERSATION DESIGN: BEST PRACTICES

Be useful

by being relevant and helpful

Brian Lee Yung Rowe 53

CONVERSATION DESIGN: BEST PRACTICES

Be useful by being relevant

and helpful

Brian Lee Yung Rowe 54

CONVERSATION DESIGN: BEST PRACTICES

Be useful by being relevant and helpful

Brian Lee Yung Rowe 55

CONVERSATION DESIGN: BEST PRACTICES

Anticipate edge cases

to be awesome

Brian Lee Yung Rowe 56

CONVERSATION DESIGN: BEST PRACTICES

Anticipate edge cases to be awesome

Brian Lee Yung Rowe 57

CONVERSATION DESIGN: BEST PRACTICES

Support free form text

since it’s text messaging

Brian Lee Yung Rowe 58

CONVERSATION DESIGN: BEST PRACTICES

Support free form text since it’s text messaging

Brian Lee Yung Rowe 59

CONVERSATION DESIGN: BEST PRACTICES

Guide users back on track

to avoid dead ends

Brian Lee Yung Rowe 60

CONVERSATION DESIGN: BEST PRACTICES

Guide users back on track to avoid dead ends

Brian Lee Yung Rowe 61

CONVERSATION DESIGN: BEST PRACTICES

Be efficient

and remember context

Brian Lee Yung Rowe 62

CONVERSATION DESIGN: BEST PRACTICES

Be efficient and remember context

Brian Lee Yung Rowe 63

CONVERSATION DESIGN: MECHANICAL TURK

Who controls our robotoverlords?

Brian Lee Yung Rowe 64

CONVERSATION DESIGN: MECHANICAL TURK

human training bot human escalation

machine learning bot human escalation

undisclosed human + bot

human coding bot

human

Brian Lee Yung Rowe 65

CONVERSATION DESIGN: MECHANICAL TURK

All bots are trained by humans

Brian Lee Yung Rowe 66

DEMO

Demo time

Brian Lee Yung Rowe 67

DEMO: FOOD DELIVERY

Brian Lee Yung Rowe 68

DEMO: FOOD DELIVERY

Brian Lee Yung Rowe 69

DEMO: FOOD DELIVERY

Brian Lee Yung Rowe 70

DEMO: CUSTOMER SERVICE

Brian Lee Yung Rowe 71

DEMO: CUSTOMER SERVICE

Brian Lee Yung Rowe 72

WE’RE HIRING

We need

• data scientists

• software engineers

• devops

• QA/test automation

• script writers!

Apply at [email protected]

Brian Lee Yung Rowe 73

THANK YOU

Questions?

@cartesianfaith [email protected]

Brian Lee Yung Rowe 74