IACT-TAP New School Thinking - Prototype Workshop

> this slide was intentionally left blank_

Disclaimer are usually fine prints…
BUT AS A REBEL; I’MMA START WITH IT.
Disclaimer: Everything is just based on my research & my own personal experience/ understanding :-)
WHAT I AM ABOUT TO PRESENT
IS NOTMY AREA OF EXPERTISE.

HELLO.
Questions? Tweet @jonnyha with #hashtag:
#Qs4Jonny

SOME THINGS I’VE DISCOVERED ABOUT
“CHATBOTS”
v.20170619
Rewards!

IN THIS DECK:
WE WILL COVER THE FOLLOWING TODAY…
1. OMGWTFBBQ!!1
Artificial Intelligence? What’s AI?
2.BASICS OF A CHATBOT
A quick one, some history & stuff
3.MY LEARNINGS
6 points
4.HANDS-ON WORKSHOP?
Yay!
Questions? Tweet @jonnyha - #Qs4Jonny

OMGWTFBBQ!!1
ARTIFICIAL INTELLIGENCE?
WHAT’S AI?

DATA SEARCH

SELF DRIVING CARS

FLIGHT & HOTEL BOOKINGS

E-COMMERCE/ CONTENT PERSONALIZATION
This is Jake :-)

AI { ML > DL/ANN || NLP/NLU }
• Machines,
showing intelligent behaviours; doing things “indistinguishable” from
human behaviour
• Machines,
recognising patterns like human beings

• Machine Learning is a tool
• Provides systems the ability to
automatically learn & improve
from experience without being
explicitly programmed
• Focuses on the development of
computer programs that can
access data and use it learn for
themselves
• Make decisions in situations
they’ve never seen

Learn themselves:
(Un)supervised learning
Make decisions in situations
they have never seen before

NICE. EVERYTHING’S SORTED!
A MACHINE THAT LEARNS…

• Artificial Neural Network
• ANN find patterns by combining
layers of artificial neurons (like
our brain)
• Information that flows through
the network affects the structure
of the ANN because a neural
network changes (learns?!)
based on that input and output
• Marketing term: “Deep Learning”?

Source: playground.tensorflow.org

Source: Google Proprietary & Confidential

GOOGLE’S #QUICKDRAW
WHAT IS THIS?

• Process of communication between computer agent & human
• Understanding of communication - with context!
Source: http://nlp.stanford.edu/~wcmac/papers/20140716-UNLU.pdf

SO MANY THINGS; SO LITTLE TIME.
• Q: I want to know about everything! Why are we focused on NLP?
• A: Great! Visit lmgtfy.com & because we are going to talk about chatbots today.
Not web-crawling bots, not robotics, botnets, automated testing bots, trading bots,
moderator bots, etc.
“bot, is a software application that runs automated tasks (scripts) over the
Internet. Typically, bots perform tasks that are both simple and
structurally repetitive, at a much higher rate than would be possible for a
human alone.”
- via Wikipedia

IN THIS DECK:
1. OMGWTFBBQ!!1
3.MY LEARNINGS
13 points
Yay!

Open-ended chat
Input is from free text, bot detects intent of
question and answers are selected from scripted
responses
• Pros: allows more freedom of input
• Cons: harder to craft and control and
requires more moderation
SO… IS CHATBOT NOT A BOT?
WHAT’S A CHATBOT?
• A chatbot is a computer program
designed based on “conversations”
i.e. It simulates conversation with
human users
• A chatbot exists on 3 main
platforms: Web, app, social
• Usually A.I. (NLP!) is used, so the
chatbot can “understand” the
context and complicated tasks
compared to a non-AI bot
e.g. “Book a flight to Phuket” &
someone might say that “I need a
flight to Phuket”
Source: https://en.wikipedia.org/wiki/Internet_bot
Closed chat
Generally uses buttons. Answers drawn from
predetermined choices through a decision tree
• Pros: simple experience and good
introduction for a brand into using chatbots
as a channel
• Cons: can feel restrictive and repetitive if not
carefully considered

THE NEW THING <VS> ANOTHER THING
“
MESSAGING IS THE
NEW BROWSER,
AND BOTS ARE THE
NEW WEBSITES
”
– Mike Roberts
Head of Messaging @Kik

THE NEW THING <VS> ANOTHER THING
“
MESSAGING IS THE
NEW BROWSER,
AND BOTS ARE THE
NEW WEBSITES
”
– Mike Roberts
Head of Messaging @Kik
“
WEBSITES, ARE NOT
THE NEW STORES.
APPS ARE NOT THE
NEW WEBSITES.
BOTS, ARE THE NOT
NEW APPS.
”
– Noone
But if you ask me:
“New, but not replacements”
Users will have preferences; but it will
become an essential product mix.

WHY CHATBOT? WHY NOW?
A FEW SIMPLE REASONS… 
• One app, to have all the services
(No need to download apps)
– Simple & seamless
– Improving experience & be where
the users are
– Easier to “install”(?)
• Interactions to conversations
(GUI to CUI)
– Conversational Branding
– Conversational Engagement
– Conversational Commerce
• Automating simple, everyday
mundane tasks
– Easy to create, and cost effective
– 24/7, no sleep required!
– Guaranteed responses
– Cheap to scale, unlike humans, so
volume can be handled :-)

CHATBOTS AREN’T NEW…
THEY’VE BEEN AROUND FOR AGES.
• Eliza (1964 ~ 1966)
– Looks for pronouns & verbs
– “You” becomes “I” and vice versa
– Computer Therapist
• Racter (~1984)
– Generates English language prose at random
– Attempted to parse text inputs, identifying
significant nouns and verbs
– ‘Generated’ a book “The Policeman’s Beard.”
• A.L.I.C.E. (~1994)
– Artificial Linguistic Internet Computer Entity
– AIML
(Artificial Intelligence Markup Language)
– Loebner Prize winner
(But not Turing Test)
• Mitsuku (2013~)
– AIML technology
– Won the 2013 and 2016 Loebner prize
• Et cetera
Source: https://en.wikipedia.org/wiki/Loebner_Prize; https://en.wikipedia.org/wiki/Artificial_Linguistic_Internet_Computer_Entity; https://en.wikipedia.org/wiki/Racter

SO WHY NOW?
GRAPHS THAT YOU MIGHT SEE EVERYWHERE
Source: Companies. Business Insider; AppAnie, https://www.appannie.com/en/insights/understanding-global-messaging-app-user/; Insight guide – The message heard around the world; https://venturebeat.com/2017/02/01/are-chatbots-more-conversational-or-controversial/
WHY NOW?
• Messaging users on the rise
• Users prefer messages than calls
• Users want quick responses
• We now have “Big Data”!
• Processing power now more powerful than ever
“Gartner predicts that intelligent automation will
manage 85 percent of businesses’ customer
relationships by 2020.”

LEARNINGS
>6 MONTHS <60 MINUTES =6 POINTS
v2017May

A CHATBOT PROJECT IS
MORE THAN MEET THE EYES
CHATBOT
Agent
Response
Tool
WoW &
Operational
Changes
Resource
Management
What does
the agents
see?
Agent
Management
Case
Management
Knowledge
Management
Knowledge
Taxonomy &
Optimisation
Connecting
back to the
bot?
CRM
Lead
Management
Feedback
Loop
Up/Cross sell
Intelligence
Other data
sources
Feedback
Management
Chat
channels
Web Chat
App Chat
Standalone
ChatApps
Social Chat
Chatbot
Design
Onboarding Use Cases
Integrations Architecture
Hosting/
Storage?
Security?

A CHATBOT PROJECT IS
MORE THAN MEET THE EYES
CHATBOT
Agent
Response
Tool
WoW &
Operational
Changes
Resource
Management
What does
the agents
see?
Agent
Management
Case
Management
Knowledge
Management
Knowledge
Taxonomy &
Optimisation
Connecting
back to the
bot?
CRM
Lead
Management
Feedback
Loop
Up/Cross sell
Intelligence
Other data
sources
Feedback
Management
Chat
channels
Web Chat
App Chat
Standalone
ChatApps
Social Chat
Chatbot
Design
Onboarding Use Cases
Integrations Architecture
Hosting/
Storage?
Security?
UNDERSTAND THINGS
CONCEPTUALLY.
HAVE CLARITY ON:
SHORT/ MID/ LONG TERM.
(Chat channel, knowledge management, use cases, etc.)

A.I. REALITY <VS> A.I. UNICORN
AI UnicornFind the right AI partner

Lots of data
Complex
mathematics in
multidimensional
space
AI Unicorn

AI Unicorn
(THINGS AREN’T AS ‘MAGICAL’ AS IT SEEMS )
Collect
data
Organise
data
Create
models
Cloud to
compute
Deploy
models
Store
data
Analyse
data
Refine
models
≠

AI UnicornFind the right partner
≠

YOU CANNOT HAVE “MACHINE LEARNING”
WITHOUT DATA
A simpleBot is running on (many) “IF/ELSE” rules;
A smartBot on is learning. To learn, it requires data to train.

YOU CANNOT HAVE “MACHINE LEARNING”
WITHOUT DATA
A simpleBot is running on (many) “IF/ELSE” rules;
A smartBot on is learning. To learn, it requires data to train.
YOU CAN’T LEARN
A.I./ ML/ NLP
IMMEDIATELY.
(But there are tools out there, to do things fast!)

REMEMBER “QUICKDRAW”?
USERS PROVIDED
THE TRAINING DATA FOR “AUTODRAW”!
Source: @jonnyha’s assumption!

THE AI UNICORN?
UNDERSTANDING
LANGUAGE(S) IS HARD
AN OXFORD/SERIAL
COMMA?!
Woei Hern invited
the strippers, Jonny and Esther.
<VS>
Woei Hern invited
the strippers, Jonny and Esther.

THE AI UNICORN?
UNDERSTANDING
LANGUAGE(S) IS HARD

THE AI UNICORN?
BUT THE REAL CHALLENGE IS
GIVING THE RIGHT ANSWER!
Yes, we need to train for specific
intent & entity with NLP… but the
bulk of the work lies in preparing
the right answers:
Taxonomy for knowledge bank
(a structured way we access
knowledge)

THE AI UNICORN?
BUT THE REAL CHALLENGE IS
GIVING THE RIGHT ANSWER!
Yes, we need to train for specific
intent & entity with NLP… but the
bulk of the work lies in preparing
the right answers:
Taxonomy for knowledge bank
(a structured way we access
knowledge)
THINK ABOUT
YOUR ANSWERS;
NOT UNDERSTANDING
QUESTIONS.
Q: ARE YOU A CONCIERGE?
Q: ARE YOU FULFILLING A
REQUEST?

THINGS ARE
FAST MOVING!
https://developer.apple.com/sirikit/; http://www.thestar.com.my/news/world/2017/07/05/sex-robots-are-here/
• 30,000 bots (in our initial presso);
now >100,000+ bots & growing
• Facebook F8 happened
• FB Group bots launched
• More apps on SiriKit Integration (iOS11)
• Discovery for FB bots started
• FB Payments for bots enabled
• Chatfuel enabled analytics; Chatfuel enabled payments
• Google bought API.ai & API.ai bought Chatbase analytics
• API.ai is now Google’s “Cloud Conversation Engine”
(mid-July)
• More brands (and competitions (shh)) are getting into it
• Katy Perry has launched a bot
• SexTech with AI SexDroids are here
• Baidu acquires natural language startup Kitt.ai, maker of
chatbot engine ChatFlow
• Et cetera…

THINGS ARE
FAST MOVING!
https://developer.apple.com/sirikit/; http://www.thestar.com.my/news/world/2017/07/05/sex-robots-are-here/
• 30,000 bots (in our initial presso);
now >100,000+ bots & growing
• Facebook F8 happened
• FB Group bots launched
• More apps on SiriKit Integration (iOS11)
• Discovery for FB bots started
• FB Payments for bots enabled
• Chatfuel enabled analytics; Chatfuel enabled payments
• Google bought API.ai & API.ai bought Chatbase analytics
• API.ai is now Google’s “Cloud Conversation Engine”
(mid-July)
• More brands (and competitions (shh)) are getting into it
• Katy Perry has launched a bot
• SexTech with AI SexDroids are here
• Baidu acquires natural language startup Kitt.ai, maker of
chatbot engine ChatFlow
• Et cetera…
START YESTERDAY;
NOT TOMORROW.
(While it’s still “new’ish”)

THE BEST BOTS TODAY ARE
FUNCTIONAL & HAS NARROW SCOPE
Today, there are very few good bots out there.
But there will be, soon.
Chatbots have not even come close to touching their potential.

THE BEST BOTS TODAY ARE
FUNCTIONAL & HAS NARROW SCOPE
Today, there are very few good bots out there.
But there will be, soon.
Chatbots have not even come close to touching their potential.
HAVE FOCUS;
FOCUS SMALL.
BE FOCUSED.
THINK BIG; START SMALL.

HAVING ONE HUMAN PERSONALITY;
BUT STILL CLEARLY NOT HUMAN
Best practice #1: Chatbot that has a human personality
Best practice #2: Don’t lie to users, let them know you’re a bot
Solution: A bot, but not a bot… A very fine line…
https://developers.google.com/actions/design/
WHAT DO YOU THINK?
• 1x Personality for Males?
• 1x Personality for Females?
• 1x Personality for Students?
• 1x Personality for every segment?!
• 1x Personality for social users?
• 1x Schizo bot?

Worst onboarding, ever: https://www.messenger.com/t/theassistantbot;

Worst onboarding, ever: https://www.messenger.com/t/theassistantbot;
Personification “Chatbotification” of the brand.
You are now a person “Chatbot”.
Just like how you meet any person “Chatbot”
for the first time; introduce yourself.
In a more industry specific term:
ONBOARDING
(i.e. Manage users’ expectations; and
have clear CTAs)

CONCEPTUALLY: IT’S SIMPLE
EXECUTIONALLY: IT’S CLEARER
EXAMPLE: SLBDB-41
“LEARNING THROUGH DOING”
Link: https://www.linkedin.com/in/sulinlau/

EXAMPLE: SLBDB-41
Link: https://www.linkedin.com/in/sulinlau/
FOLLOW HER ON LINKEDIN;
BUT SAY “HI” AT YOUR OWN RISK.

EXAMPLE: SLBDB-41
Visit: https://www.facebook.com/trythinkthings/ & https://twitter.com/SLBDB_41

EXAMPLE: SLBDB-41
DISCLAIMER:
SLBDB-41 TOOK AROUND ~16 HOURS TO BUILD.
BECAUSE IT’S A “SIMPLE BOT”.
(No contextual understanding, no flow, no fulfillment; no integrations)

QUESTIONS?
LET’S CHECK ON #Qs4JONNY BEFORE THE LAST POINT
Q: Can you share this deck?
A: If you can find it. It will be made available
public; on Slideshare. Soon. Maybe.

EXAMPLE: SLBDB-41
• Started off with 100+ intents/topics – Smalltalks
• (Not so friendly ) User test, +20~ intents/topics
– Mainly vulgarity, sexual intent, insulting another, parents
• First 2 days of launch: +20~ intents/topics
– smalltalk.agent.movies
– smalltalk.agent.childhood
– smalltalk.dialogue.utterance
– smalltalk.dialogue.animals
– intent.telltime
– smalltalk.about.geography
• Today, over 160+ intents/topics, and still unlikely sufficient

EXAMPLE: SLBDB-41
• Started off with 100+ intents/topics – Smalltalks
• (Not so friendly ) User test, +20~ intents/topics
– Mainly vulgarity, sexual intent, insulting another, parents
• First 2 days of launch: +20~ intents/topics
– smalltalk.agent.movies
– smalltalk.agent.childhood
– smalltalk.dialogue.utterance
– smalltalk.dialogue.animals
– intent.telltime
– smalltalk.about.geography
• Today, over 160+ intents/topics, and still unlikely sufficient
MOST WORK TO BUILD A BOT IS
AFTER LAUNCH.
(Monitor analytics; update knowledge; train new intents; retrain conversations; and
repeat forever?!)

SUMMARY OF LEARNINGS:
6x POINTS
1. THINK BIG
– Understand things conceptually, but know what you want
2. KILL THE UNICORN!!1
– Artificial Intelligence is not magic; but NLP tools are here!

<<6 SECONDS COMMERCIAL BREAK>>
1. THINK BIG
kill it.

SUMMARY OF LEARNINGS:
6x POINTS
1. THINK BIG
3. SPEED
– Evolution of the chatbot space: Start, now!
4. FOCUS
– There are no “best” today; Start small & be focused
5. BRAND
– Give your chatbot a personality, onboard the users.
6. PROTOTYPING
– 16 hours learning with “SLBDB-41”: Do, do, do. Stop talking & just do.

IN THIS DECK:
1. OMGWTFBBQ!!1
3.MY LEARNINGS
6 points
Yay!

IACT-TAP New School Thinking - Prototype Workshop

“THINK BIG; START SMALL; START NOW”
Source: https://www.forbes.com/sites/chunkamui/2016/01/03/6-words/#3d5d6ba51a3b
Know what you are
working towards;
Understanding through doing,
face (small) blockages;
Learning through doing, & better
understand the bigger picture;

“THINK BIG; START SMALL; START NOW”
A THOUGHT BEFORE THE WORKSHOP…

READING & RESOURCES
Join a community?
• ChatbotsMagazine.com
• ChatbotsJournal.com

READING & RESOURCES
Join a community?
• GDGKualaLumpur

CREDIT: SOME POINTERS CAME FROM THIS LEGEND.
CONNECT & SAY HI.
Link: https://www.linkedin.com/in/tjlsharp/

“BUT I’M NOT A CODER!”
NEITHER AM I.
Bot Building Platforms:
– API.ai (Multiple channel integrations available)
– Chatfuel (Very easy for controlled flow conversations – but Facebook Only)
– MOTION.ai (A good start to create conversation flow)
– QnAmaker.ai (For FAQ bots; if you have a good FAQ page!)
– IBM & Watson
• Bluemix Trial Sign Up
• Bluemix Watson Catalog Info
• Watson Virtual Agent information
– RECAST.ai
– INIT.ai
– MEYA.ai
– GUPSHUP.io
Bot Analytics:
– dialoganalytics.com
– chatbase.com
– botanalytics.co
– dashbot.io
– bot-metrics.com
Note: Not yet tried :-S – Fine print: Tools shown are based on self-research; I do not work for, nor endorsing any of the tools above.

WOW. THAT’S A LOT OF STUFF.
SO, AGAIN -- WHAT’S A CHATBOT?
Open-ended chat
Input is from free text, bot detects
intent of question and answers are
selected from scripted responses.
• Pros: allows more freedom of
input
• Cons: harder to craft and control
and requires more moderation
Closed chat
Generally uses buttons. Answers
drawn from predetermined choices
through a decision tree
• Pros: simple experience and good
introduction for a brand into
using chatbots as a channel
• Cons: can feel restrictive and
repetitive if not carefully
considered

I’M SO NOT PREPARED FOR THIS “HANDS-ON” PART.
(BLAMING WOEI HERN)
BUT VENN DIAGRAMS MAKE ME LOOK SMART.
* for API.ai + Chatfuel; see: https://chatbotslife.com/how-to-integrate-api-ai-with-chatfuel-48baafd4613
OPEN ENDED CHAT
1. Launch API.ai for an NLP-bot;
2. Go to “Console”
3. Sign in with your Gmail
4. Click on “Pre-built Agents”
5. Understand how it works
6. Build your own intents
CLOSED-CHAT
1. Go to Chatfuel for a
decision tree bot
2. Start messing around with the
onboarding; flow; etc.
3. Understand how it works
4. Try the “AI” (NLP) engine
Hybrid?
Many platforms can do this;
Try Motion.ai * (Web + fb Integration)

CLICK ON OUR PROFILES & CONNECT
• Today’s trainers: “Old school is new school”
• Mentions & credits: “People of interest”

READING & RESOURCES
• https://www.slideshare.net/gjuljo/chatbot-development-workshop-with-the-
microsoft-bot-framework?qid=cbf6967a-7503-4f62-ae63-
bac6de04549b&v=&b=&from_search=20
• https://medium.com/@RecastAI/10-chatbot-tips-to-build-awesome-bots-on-
recast-ai-aebbc5ed4ac7
• https://www.themarketingtechnologist.co/are-we-willing-to-embrace-chatbots/
• https://getvoip.com/blog/2017/04/21/the-current-state-of-chatbots-in-2017/
• https://www.smashingmagazine.com/2016/12/conversational-design-essentials-
tips-for-building-a-chatbot/
• http://www.uxmatters.com/mt/archives/2017/02/designing-conversational-
chatbot-user-interfaces.php
• https://designmodo.com/chatbots-design/
• https://www.ibm.com/developerworks/library/cc-cognitive-chatbot-
conversational-design/
• https://www.washingtonpost.com/news/the-switch/wp/2016/04/07/why-poets-
are-flocking-to-silicon-valley/
• https://www.smashingmagazine.com/2017/04/applications-machine-learning-
designers/
• https://www.chatbotalk.com/chatbots-missing-their-brand-identity.html
• https://developers.google.com/actions/downloads/get-started.pdf
• https://developers.google.com/actions/design/
• http://app.smrtp.io/h0bk/QlO6fdd4cE
• https://blog.prototypr.io/a-guide-to-developing-bot-personalities-c6eba213d77b
• https://chatbotsmagazine.com/how-to-improve-your-chatbot-in-3-simple-steps-
36f9d26d7f2f
• https://venturebeat.com/2017/07/03/what-it-takes-to-build-a-compelling-
chatbot/
• https://chatbotsmagazine.com/4-subtle-ways-to-make-your-chatbot-more-
human-d482c7e57b3a
• https://chatbotsmagazine.com/19-best-practices-for-building-chatbots-
3c46274501b2
• https://chatbotsmagazine.com/10-simple-tips-on-bot-strategy-and-design-
a4b48116ee76
• http://garage.mybluemix.net/posts/chatbot-best-practices/
• https://medium.com/chat-bots-developers/introduction-to-text-clustering-
50d3718ddb01
• http://www.gartner.com/imagesrv/media-products/pdf/rage_frameworks/rage-
frameworks-1-34JHQ0K.pdf
• https://www.forbes.com/sites/blakemorgan/2017/03/21/how-chatbots-will-
transform-customer-experience-an-infographic/#4e29b1b67fb4
• https://medium.com/@obaid/conversational-user-interfaces-case-studies-
4c040a6272c
• https://medium.com/@Paul_Boutin/why-most-chatbots-fail-1c085b74d6ad
• https://cloudplatform.googleblog.com/
• https://cloud.google.com/blog/big-data/
• https://messengerify.me
• http://www.google.com

COPYRIGHT NOTICE
HELLO, IT’S ME. I’MMA SUE YOU!
PLEASE ENSURE PERMISSION IS GRANTED BY ME BEFORE
DISTRIBUTING THIS DECK.
IF NOT –
DO NOT DISTRIBUTE THIS DECK;
IT HAS BEEN MADE FOR INTERNAL USE (MY OWN USE),
AND IS FOR REFERENCE ONLY; THE DECK MAY NOT CONTAIN
COMPREHENSIVE INFORMATION;
AND IS AN INDIVIDUAL’S POINT-OF-VIEW
(WHOM MAY OR MAY NOT BE A SUBJECT MATTER EXPERT);
AND CERTAIN THINGS MAY ORIGINATE FROM OTHER
SOURCES THAT I MAY HAVE MISSED IN REFERENCES

IACT-TAP New School Thinking - Prototype Workshop

More Related Content

Similar to IACT-TAP New School Thinking - Prototype Workshop (20)

Recently uploaded (20)

IACT-TAP New School Thinking - Prototype Workshop