AI FOR EVERYONE:

SECOND HALF
WHAT WE HAVE LEARNED SO FAR:

 Terminologies : Artificial Machine Deep Data

Intelligence Learning Learning Science

 What is ANI and AGI ?

 ML technique – Supervised Learning

 Workflow of ML and DS projects

 What makes an AI company

 What ML can do or cannot do

CASE STUDY:
SMART SPEAKER
OBJECTIVE :
 What does it feels like to work on a complex AI product.
 How you would write AI software to get a smart speaker to respond to
a verbal command such as "Hey device, tell me a joke" and execute on it ?
 STEP 1:
 USE A TRIGGER OR WAKE WORD SUCH
AS "HEY DEVICE".
 THE SPEAKER USES A MACHINE LEARNING
ALGORITHM TO INPUT THE AUDIO CLIP AND OUTPUT 0
OR 1 .
AUDIO 0/1
STEP 2:
SPEECH RECOGNITION

 NOW WHAT THE SOFTWARE HAS TO DO IS :

 MAP THE AUDIO CLIP " TELL ME A JOKE " TO A TEXT TRANSCRIPT , WHICH IN THIS
CASE ARE 4 WORDS .

AUDIO TELL ME A JOKE

STEP 3:
INTENT RECOGNITION

 THAT MEANS TO TAKE WHAT HAVE YOU SAID AND TO FIGURE OUT WHAT YOU ACTUALLY WANTED TO DO.

o " TELL ME A JOKE " / MUSIC ?

" DO YOU KNOW SOMETHING FUNNY " / Intent mapping ALARM ?
" WANT TO LISTEN A JOKE " . NEWS ?
JOKE ?
CALL ?
MANY MORE ….
STEP 4 – THE LAST :
EXECUTE JOKE / COMMAND

A CLASSIC PIECE OF CODE WRITTEN BY A SOFTWARE

ENGINEER THAT EXECUTE A JOKE .
AT A GLANCE:

OTHER FUNCTIONS : KEY STEPS :

 PLAY MUSIC  TRIGGER WORD DETECTION
 MAKE A CALL  SPEECH RECOGNITION
 CURRENT TIME  INTENT RECOGNITION
 UNIT CONVERSION  SPECIALIZED PROGRAM TO
EXECUTE THE COMMAND
 GK QUESTIONS
 NAVIGATION
CASE STUDY :
SELF DRIVING CAR
OBJECTIVE :

 A simplified description of self-driving car.

 Help you to better understand how you can

piece together multiple AI components
in order to build these amazing things.
BASIC KEY STEPS FOR DECISION :
WORKFLOW: Input Image via
Sensors / camera/radar

Motion planning
Steer/accelerate/brake
SOME OTHER COMPONENTS :
Trajectory
Car Detection
Prediction
Input Image via Motion planning
Sensors / Steer/accelerate
camera/radar Pedestrian Trajectory /brake
Detection Prediction

GPS/Maps Lane Detection

Obstacle
Detection

Traffic Light
Detection
ROLES OF
AI TEAM :
 • Design Specialized
Software software
Engineer • Set timer
• Telling a weather etc.

• Gather DATA
• Generating A to
ML B Mapping
Engineer • Build Algorithm
needed for the
product ML
Scientist

• Extend state-of-art in
ML
ML • Some publish papers.
Researcher • Focus on
doing Research

State-of-the-art (sometimes cutting edge) refers to the highest level of general development, as of a device, technique, or scientific field achieved at a
particular time
 • Examine DATA and
Data provide insights
Scientist • Make Presentations
to executive.

• Organize DATA
Data • To make sure Data
is safe and easily
Engineer
accessible because
Data will be in bulk.

• Helps what to build

AI • Tells what is feasible.
Product • Tells what is
Manager valuable
AI TRANFORMATION PLAYBOOK

This Photo by Unknown author is licensed under CC BY.

 Review :
Gaining Execute pilot project to gain momentum.
Momentum

AI Team Build an in-house AI team.

Training Provide broad AI Training

Strategy Develop an AI strategy

Communications Develop internal and external communication

EXECUTE PILOT PROJECT
TO GAIN MOMENTUM :

 More important for the initial project is to

succeed rather than be the most valuable.
 Show traction within 6 to 12 months.
 First pilot project can be either in-house or out
sourced to get expertise and build momentum
faster.
BUILD AN IN-HOUSE AI TEAM :

 Companies should consider hiring an internal team of AI talents.

 This team is responsible for gaining :
 AI capabilities
 Supporting different divisions with AI solutions
 Developing recruiting and retention standards
 Building a platform that can be used company-wide .
PROVIDE BROAD
AI TRAINING :
 Employees should be
encouraged to receive an
AI education via
 A combination of in-
person consulting and
digital courses on AI
and machine learning .
DEVELOP AN AI STRATEGY :
 Use the virtuous cycle of AI .
 Better product leads to more customers , more customer leads to
more data and more data leads to better product
 This create an industry-specific advantage that you are uniquely
positioned to defend against new entrants on the market.

Better
Product

More More
Data user
EXAMPLES :

 Deploying model to different users results in more

data , more training and more accuracy .

 1) Smart Speaker – Alexa , Cortana , Siri

 2) Facebook – Auto Tagging
 3) Online Ad Recommendations
DEVELOP AN INTERNAL AND EXTERNAL COMMUNICATIONS:

Promote insightful
Companies should
internal explanations
maintain good
Invest in appropriate Attract and retain of AI to relieve
relationships with
marketing talents. employee job
investors and
replacement
governments.
concerns.
AI PITFALL
TO AVOID :
 DO DONT

 Be realistic about the  Expect AI can solve everything .

limitations of AI .
 Hire ML Engineers and count solely
 Include cross functional talent to on them to come up with use
find feasible and valuable cases .
projects .
 Expect your project to work for the
 Go through the iterative process . first time.
 Work with the AI team to establish  Expect Traditional planning
timelines , milestone , KPI etc. process to apply without changes.
 Think you need a super star AI
Engineer .

KPI - key performance indicator is a type of performance measurement. KPIs evaluate the success of an
organization or of a particular activity in which it engages.
How to Choose Your
First AI Project ?
 Get friends TO LEARN ABOUT AI

START BRAINSTORMING PROJECT

NO PROJECT IS TOO SMALL AND IT'S BETTER

TO START SMALL AND SUCCEED, THAN TO
START TOO BIG AND NOT SUCCEED

HIRE A FEW ML OR DS PEOPLE TO HELP YOU

(COMPANY/IN-HOUSE)

WHEN YOU GO BIGGER HIRE AI LEADER (VP

OF AI /CAIO)
SURVEY OF MAJOR AI APPLICATIONS AREA :
COMPUTER VISION :

 IMAGE CLASSIFICATION / OBJECT RECOGNITION -

 OBJECT DETECTION - Panda

 IMAGE SEGMENTATION -

 TRACKING -
NATURAL LANGAGE PROCESSING :
 Text Classification - Sports
New
Article News

Politics

 Information Retrieval – web search

 Name Entity Recognition - Pakistan has a friendly relationship with China

 Machine Translation - ‫ اسالمی جمہوریہ پا ِكستان‬------ (input)

Islamic Republic of Pakistan ------ (output)
 NLP Continue :

 Parts of speech tagging

  Speech Recognition – Speech to text

SPEECH : 


Speaker ID – Predict the speaker
Speech Synthesis – Text to Speech
 Wake Word Detection - " Hey Device "
ROBOTICS :

 Perception – Figuring
out what's in the world
around you
 Motion Planning –
Finding a path for the
robot to follow
 Control – Sending
command to the
motors to follow the
path
SURVEY OF MAJOR AI
TECHNIQUES :
UNSUPERVISED
LEARNING :

 Unsupervised
learning is where the
input data is
unlabeled and the
system tries
to learn structure
from that data
automatically,
without any human
guidance There are 3 clusters or groups in this image
 Unsupervised
Learning Technique
– CLUSTERING :
TRANSFER LEARNING :

 Transfer learning is a research problem in machine learning that

focuses on storing knowledge gained while solving one problem
and applying it to a different but related problem .

CAR DETECTION Truck Detection

REINFORCEMENT
LEARNING :
 In reinforcement learning, an agent receives
information about its environment and learns
to choose actions that will maximize some
reward .
 Example :
A neural network that looks at a videogame
screen and outputs game actions in order to
maximize its score can be trained via
reinforcement learning.
GENERATIVE
ADVERSARIAL
NETWORK (GAN) :

 GAN focuses to generate the

data from scratch.
 It means GAN is able to
produce or to generate new
content .
KNOWLEDGE GRAPH :

 Information that is drawn from a

Knowledge Graph means a database
that lists people and key information
about these people.
 Such as their birthday, when they pass
away, their bio, and other properties of
these individuals.
AI AND SOCIETY

 AI and Hype
 Limitations Of AI
 Bias
 Adversarial Attacks
 AI developing Economies and Jobs
A REALISTIC VIEW OF AI

 Goldilocks Rule For AI

"Neither Be Too Optimistic Nor Be Too Pessimistic
About What AI Can And Cannot Do"

 Too Optimistic : Sentient / super-intelligent AI killer robots coming

soon
 Too Pessimistic : AI cannot do anything so an AI winter is coming
 Just Right : AI can't do everything, but will transform industries
Too Optimistic

we should not be too

optimistic about AI technologies and
having an unrealistic view of AI technologies
may make people think that
sentience or super intelligence,
artificial, general intelligence is coming soon,
and we should invest a lot of resources
into defending against AI evil killer robots.
 AI cannot do everything.
There are some things AI cannot do and so, another AI
winter is
coming.
Too Pessimistic  AI Winter
The term AI winter refers to a couple of episodes in
history when AI had been over-hyped and when
people figured out that AI couldn't do everything that
they thought it would.
It resulted in a loss of faith and
a decrease in investment in AI.
 One difference between AI now and the earlier winters of
a few decades
ago, is that AI today is creating tremendous economic
value.

Just Right We also see a surprisingly clear path for it to continue to

create even more
value in multiple industries.
So, the combination of these two things ensures that AI will
continue
to grow for the foreseeable future.
Even though, it is also true that AI cannot do everything.
Rather than being too optimistic or too pessimistic the
story of Goldilocks
learn that something in-between is just right.
LIMITATIONS OF AI

Explain Biased AI
Performance
ability is through
Limitations
hard biased data

Adversarial
attacks on
AI
Performance
Limitations "The toy arrived two days late,
so I wasn't able to give it to my niece for her
birthday.
As we have studied in the Can I return it?"
previous week about the
performance limitations of AI as
if you want an AI model to
differentiate between the types
of emails, it is possible but if you
are willing from your AI model to
write a proper email in reply of
the email you have received it
is kind of impossible for now
may be soon it will be possible.
 One of the limitations of AI is that
explainability is hard and many high-

Explainability Is performing AI systems are black boxes.

Meaning that it works very well
Hard but the AI doesn't know how
to explain why it does what it does.
 For Example
Explainability How does a human look at this

Is Hard and say, that's a coffee mug?

You know there are some things
you can point to like, there's a
room for liquid and it has a
To be fair, humans are handle.
also not very good But we humans are not very good at explaining,
at explaining how we how we can look at this and decide what it is.
make decisions But because AI is a relatively new thing,
ourselves. the lack of explainability is
sometimes a barrier to its acceptance.
 Biased AI through biased data

As a society, we do not want to discriminate against individuals based

on their gender, based on their ethnicity and we want people to be treated fairly. But
when AI systems are fed data that doesn't reflect these values, then an AI can become
bias or can learn to discriminate against certain People.AI systems are already making
important decisions today, and will continue to do so in the future as well. So, bias
matters.
Biased AI through biased data

For Example :
If an AI system is trained primarily on data of lighter skin individuals, then it will
be more accurate for that category of individuals to the extent that these
systems are used in, for example, criminal investigations, this can create a very
biased and unfair effect or dark-skinned individuals
Why Bias Matters
How To Reduce Bias
• Zero out the bias in words
Techincal • Use less biased and/or more inclusive
Solutions data

Transparency • constantly check what types of bias,

And/Or Auditing if any, these AI systems are exhibiting

Processes

• the organizations are becoming

Diverse Workforce more heterogeneous mix of people
in terms of gender,age,ethnicity etc
Technical Solutions

Zero out the bias in words Use less biased and/or more
inclusive data
Try to use less bias and or more
Researchers have found that when
an AI system learns a lot of inclusive data.
different numbers with which to store For example, if you are building a
words, there are few numbers that face-recognition system, and make
correspond to the bias. sure to include data from multiple
If you zero out those numbers, ethnicities, and all genders, then
just set them to zero,
your system will be less biased and
then the bias diminishes
more inclusive.
Significantly.
Transparency And/Or Auditing
Processes

Many AI teams are subjecting their systems to better transparency and or

auditing processes, so that we can constantly check what types of bias, if
any, these AI systems are exhibiting, so that we can at least recognize the
problem if it exists, and then take steps to address it.
For example, many face recognition teams are systematically checking
how accurate their system is on different subsets of the population to
check whether, it is more or less accurate on dark-skinned versus light-
skinned individuals. Having transparent systems as well as systematic auditing
processes increases the odds that will at least quickly spot a problem, in case there
is one, so that we can fix it
Adversarial Attacks On AI
ØPicture Perturbation

ØPhysical Attacks
 Picture Perturbation means to make minor
changes in the pixels of the image, the changes
that are almost imperceptible to the human eye.
But an AI system sees the world very differently

Picture than you and I do.

Perturbation
 Physical Attacks

Physical Attacks means to add some extra content in a

picture which can fool the AI system
For Example :
A different group of researchers showed that if you
affects stickers like these onto a stop sign, you can
fool an AI system into not seeing the stop sign at all. It
things there's something else there other than a stop
sign.
Physical Attacks

A group at Carnegie Mellon University

was able to design a funky pair of glasses like this.
So, that when does man wears this pair of glasses,
he can fool an AI system into
thinking that he is actress Milla Jovovich.
Adversarial Defenses

Ø Defenses do exist, but incur some cost

Ø Similar
to spam vs anti-spam, we may be
in an arms race for some applications
Adverse Uses Of AI

 Deep Fakes
 Synthesize video of people doing things they never did
 Undermining of democracy and privacy
 Oppressive Surveillance
 Generating fake comments
 Spam vs Anti-Spam and Fraud vs Anti-Fraud
AI and Developing
Economies
Leap Frog
Leapfrogging is the notion that areas
which have poorly-
developed technology or economic
bases can move themselves forward
rapidly through the adoption of
modern systems without going
through intermediary steps.
Leap Frog

Examples
 Mobile Phones
 Developing Countries switched directly on mobile phones instead of
landlines which were used by developed countries.
 Online Education
 Developing Countries which do not have much universities and institutes
for higher education have directly switched for online education.
How Developing Economies Can
Build AI

 Focus on AI to strengthen a country's vertical

industries
Developing countries today should not try to build their own
Web search engine. There are already great web search engines
and that was last decades competition.
Instead, if a country has a very strong vertical industry then that
country should build AI to automate that industry, it will even further
Strengthen that industry.
How Developing Economies Can Build AI

 Public Private Partnerships To Accelerate

Development
Public-private partnerships, meaning governments and
corporations working together, can really help accelerate a
vertical industry's AI developments.
In highly regulated sectors, ranging from healthcare
to transportation like self-driving cars to finance, there are
certain outcomes that we want and certain outcomes that
we don't want.
How Developing Economies Can
Build AI
 Invest In Education
AI is still so immature. There's still plenty of room for every nation
to learn more about AI, maybe even build up its own AI
workforce and participate in a significant way in this AI
powered world that we're building.
AI Impact On Jobs World Wide
Conclusion :

What you have learned ?

 What is AI ?
 Building AI Projects
 Building AI in your company
 AI and Society

Best Of Luck For Your Test !