00:00:25
Welcome to this NPTEL online certification course
on artificial intelligence and marketing. And now
00:00:31
we will talk about module 7. Now, as you can
see that we are we are talking about chapter 2,
00:00:37
that is how to go about developing marketing
strategies and plans using AI. In module 6,
00:00:42
we have talked about the customer value and
the role of artificial intelligence in the
00:00:46
value delivery process. Now this module 7 to
11, they are dedicated to understanding how
00:00:53
to transform marketing strategy with the help
of AI. So, this is chapter 2 and we are talking
00:01:00
about the transforming marketing strategy using
AI and which is part 1 and we are in module 7.
00:01:08
To have a overview of this module, we will start
with exploring the ideas framework and understand
00:01:14
five elements of technology landscape.
Then we will study the intelligence and
00:01:18
data landscape of ideas framework in detail.
Then we will understand different dimensions
00:01:24
of human and machine intelligence. After
that, we will study the trade-off between
00:01:29
applying human and machine intelligence to
solving marketing and business problems.
00:01:35
Thereafter, we will understand the shortcomings
of intelligent systems and the various kinds
00:01:41
of problems that come with them. And then we
will talk about exploring the building blocks
00:01:46
of the modern-day data foundation. So, now we
let us look at the ideas framework. So, these
00:01:54
the five elements of the technology landscape
are, one is intelligence, the second is data,
00:02:02
the third is expertise, the fourth is architecture
and the fifth is strategy and together they form
00:02:09
this ideas framework, IDEAS. Now, these elements
create opportunities for value creation and
00:02:21
distribution in innovative ways and change the
way business leaders can manage the procurement,
00:02:28
technology development, human resource
management and firm infrastructure.
00:02:32
So, these five elements will help the
leaders in in managing this these three,
00:02:39
fourth things. Technology implementations
for the sake of technology does not bring
00:02:44
the desired value for companies. So, it is
not it should not be just for the sake of the
00:02:50
technology. Technology based on deep learning
have little sense of causality space. So,
00:02:57
this causality means cause and
effect relationship. Space, time
00:03:11
or other fundamental concepts that human beings
effortlessly call on to move through the world.
00:03:17
Now, the first component of ideas framework is
intelligence. So, that is the first component of
00:03:23
ideas framework. Companies are now creating
applications and machines whose reasoning
00:03:30
ability is adaptable and savvy more like the
way human approaches a problem and the task. So,
00:03:38
it is more human-like. For example, new
generation of robots can generalize in real
00:03:51
world settings like warehouses manipulating
items without being told what to do.
00:03:58
Consider emotional AI which grew out of work with
autistic children to help them understand and
00:04:06
express their emotions. It is now evolving into
on board automobile AI that could be as effective
00:04:14
in saving motorist lives as seed welds by
leveraging the most powerful cognitive characters
00:04:21
of human that is awareness and adaptability. These
developments promise potentially more intelligent
00:04:30
solution to pressing commercial and social
challenges. Now, let us look at the intelligence
00:04:35
from the perspective of human versus machine.
So, we are talking about the human supremacy.
00:04:41
No machine powered by AI can yet match the ease
and efficiency with which even the youngest
00:04:47
human learn comprehend and contextualize.
Accidentally drop an object and a one year
00:04:55
old child who sees you reaching out
for it will retrieve it for you. So,
00:05:00
this is how one year old child will respond.
Now, throw it down on the purpose and the
00:05:06
child will ignore it. So, if it falls
then the child will pick it up for you.
00:05:11
If you throw it then the child
will ignore it. In other words,
00:05:18
even very small children understand that
other people have intentions. That is an
00:05:24
extraordinary cognitive ability that seems to
come almost pre-wired into the human brain. That
00:05:31
is not all. Beginning at a very young age,
they develop an intuitive sense of physics.
00:05:37
They begin expecting objects to move along smooth
paths. So, this is what a child expects. They
00:05:45
remain in existence, fall when unsupported and not
act at a distance. Now, where machine supremacy
00:05:55
comes in, yet AI can do many things that people,
despite being endowed with natural intelligence,
00:06:02
find impossible or difficult to do well. That is,
one is recognize patterns in vast amount of data.
00:06:11
So that defeats the greatest champion at chess, Go
and Jeopardy. Run complex manufacturing processes
00:06:19
efficiently and calls to customer service
centers. Run complex manufacturing processes
00:06:26
efficiently aid callers to customer service
centers. Analyze whether soil conditions,
00:06:32
satellite imagery to help farmers examine
crop yields. Scan millions of internet images.
00:06:40
In the fight against child exploitation, detect
financial fraud. Predict customer preferences,
00:06:47
personalize the advertising and much else. So,
this is where machines are superior to human. Now,
00:06:54
the middle ground between the fight of these
two, man and machine supremacy, automating
00:07:00
such task lies beyond not only the capabilities of
human, but also of traditional procedure logic and
00:07:07
programming. And most important, AI has enabled
humans and machines to complement each other.
00:07:15
So, it is not about man versus machine. It
is about how they can come together. So,
00:07:36
AI has enabled humans and machines to
complement each other, transforming
00:07:40
mechanistic processes into highly adaptive,
organic and human centered activities. The
00:07:46
next comes human like machine intelligence. A
2022 report by Accenture shows that more than
00:07:53
three-fourth of major companies currently
have deep learning initiatives underway.
00:07:59
Deep learning is a powerful subset of machine
learning. It works through neural networks
00:08:05
consisting of simple neuron like processing units
that collectively perform complex computations. AI
00:08:13
based on deep learning must be trained from the
bottom up on massive amounts of data and often
00:08:20
fine-tuned with additional data. So, deep learning
is trained from bottom up and massive amount of
00:08:26
data is required. But this data-hungry approach
is beginning to run into significant challenges.
00:08:34
Challenges of capacity, challenge
of affordability and the challenge
00:08:39
of sustainability. So, these are the three
challenges that this deep learning faces. So,
00:08:48
the three challenges that deep
learning faces are capacity,
00:08:54
affordability and sustainability. Meanwhile,
on the frontiers of research, the nature of
00:09:07
machine intelligence is taking radically
human turn, becoming less artificial and
00:09:14
more intelligent. Less like the autonomous vehicle
that has to be laboriously taught everything and
00:09:23
more like the human infant who comes equipped
with a remarkable efficient capacity to learn.
00:09:30
The quest for more human like AI after lying
dormant for decades have taken on a new life.
00:09:38
Impelled by the limits that current approaches
to intelligence are now running up against. For
00:09:42
senior leaders, navigating this dilemma begins
with an understanding of those limits. Now,
00:09:50
what is the trouble with intelligence? The
limit of the present deep learning-based
00:09:55
AI warrants a new approach to machine
intelligence which is more human like. So,
00:10:01
deep learning is more suitable
or suited to tackle for certain
00:10:06
narrowly defined problems and not as
a basis for more general intelligence.
00:10:12
So, this deep learning is not going to replace
general intelligence. Some key troubles with the
00:10:20
present-day AI are many AI systems are not all
that smart. AI image recognition was one of the
00:10:28
great AI success stories of recent years, but it
has been very easily confused by researchers. What
00:10:35
is at stake is not simply correctly classifying an
image but genuinely recognizing an object. As with
00:10:42
self-driving cars or delivering drones,
failure could have fatal consequences.
00:10:49
The next is, so that was one problem.
The next problem is complex systems
00:10:53
suffer from the black box problem.
AI systems are often used to help
00:11:01
make highly consequential decisions. Who
gets approved for a loan, who gets hired,
00:11:07
who wins payroll, who long, how long
a present sentence someone gets,
00:11:14
where and how a company ads are distributed on
social media and more. But many of these systems,
00:11:21
especially those that employ deep learning
are opaque, that is they are not transparent.
00:11:32
So, it is impossible to explain how these
deep learning algorithms working with
00:11:38
enormous amounts number of parameters and
many intricately interconnected layers of
00:11:43
abstraction reach their conclusion. And those
conclusions can sometimes be disastrous. So,
00:11:51
one is that we do not know how they have
reached that conclusion. And then that can
00:12:07
be sometimes disastrous. Resulting in racial
discrimination in loans and criminal justice,
00:12:15
respected brands whose ads on social media show up
next to neo-nausea content or conspiracy theories.
00:12:24
So, they lack fundamental
knowledge frameworks. So,
00:12:28
that is the third problem that they
lack fundamental knowledge frameworks.
00:12:33
Causation an essential component of common
sense. Much of the success of deep learning
00:12:38
has been driven by powerful ability to find
correlations. So, this is what we are looking
00:12:45
for such as that between a constellation
of symptoms and a particular disease.
00:12:50
But as we should all know by now correlation
is not causation. So, this correlation does
00:12:56
not mean cause and effect relationship. If machine
understood that one thing causes another then they
00:13:11
would not need to be retrained for each new task.
Instead, they should apply what they know in one
00:13:18
domain to the different domain. Now, let us talk
about the intelligence and human AI augmentation.
00:13:26
AI has enabled human and machine
to work together efficiently. And
00:13:30
such collaboration is creating an
array of new high value jobs. So,
00:13:35
now these humans and machines are coming together,
and they have created high value jobs. Now,
00:13:42
let us take a look at the case of Obeta a
German electronic wholesaler. At Obeta a
00:13:47
German electronic wholesaler whose warehouse is
run by Austrian warehouse logistic company NAP.
00:13:54
Human workers are teaching a new generation
of robo pickers how to handle different sized
00:14:00
and textured items. To train a robo NAP
workers put unfamiliar object in front of
00:14:06
it and see if it can successfully adapt
to them. When it fails it can update its
00:14:13
understanding of what it is seeing and try
different approaches. But when it succeeds
00:14:19
it gets a reward signal programmed
by humans to reinforce the learning.
00:14:25
Then a set of SKUs that is stock keeping
units differs totally from other sets.
00:14:40
The team reverts to supervised learning,
collecting and labeling a lot of new training
00:14:45
data as happens with deep learning systems.
NAP robo pickers are acquiring general purpose
00:14:52
abilities including 3D perception and an
understanding of how objects can be moved
00:14:57
and manipulated. In many cases the items have
not been pre-categorized, which is unusual
00:15:04
for industrial packaging systems. It means
that the robots are learning how to handle
00:15:10
them in real time. Now, this is a critical
skill to have when dealing with electronics
00:15:17
especially when you consider the different care
required to handle a light bulb and a stove.
00:15:25
What are the implications of using smart
robots? One is the increased reliability.
00:15:33
Previously NAP robo pickers reliably handled
only about 15% of the objects. The covariant
00:15:39
powered robots now reliably handle about 95%
of the objects. The second is the increased
00:15:48
speed. Robots are faster than humans, picking
about 600 objects an hour versus 450 for human.
00:15:57
The third is no layoff. Human workers instead
of losing their jobs have been retrained to
00:16:05
understand more about robotics and computers. Two,
we next discuss the future of intelligence. Now
00:16:12
let us look at the future of intelligence.
The authors of building machines that learn
00:16:18
and think like humans, a seminal piece on
the new direction in machine intelligence
00:16:24
state. As long as natural intelligence
remains the best example of intelligence,
00:16:29
we believe that the project of reverse
engineering, the human solution to
00:16:33
difficult computational problems will continue to
inform and advance the artificial intelligence.
00:16:42
The question of for senior leaders in which
more human like cognitive abilities detailed
00:16:49
next might be most relevant to capture value for
their businesses and delivering value to their
00:16:54
customers. So, the first is generalizing in real
world settings. While theoretical arguments rage
00:17:02
over deep learning versus some ideal versions of
artificial general intelligence as the means of
00:17:08
getting to more human like intelligence,
practitioners are not waiting. They are
00:17:14
drawing on all the disciplines of AI to open up
new possibilities for machine capabilities and
00:17:21
performance. For example, covariants are
looking to build out its brain to power
00:17:26
robots in manufacturing, agriculture, hospitality,
commercial kitchens and eventually people's homes.
00:17:33
The case of Alberta using smart robots
which we discussed is also an example of
00:17:39
this kind of applications. The next is the
survival of the fittest algorithm. For a
00:17:45
e-commerce seller like Zappos, irrelevant
search results are a perennial headache.
00:17:55
Because queries can now have multiple
different meanings to a website search
00:18:00
engine. Having accurate search results among
the enormous inventory can be very difficult.
00:18:06
So potential customers who enter such term
for a particular style of dress shoe who are
00:18:12
shown dresses instead will soon get fed
up and move on to the competitors. So,
00:18:17
the customer are asking for dress shoes and
the search engine is showing them dress so
00:18:24
they will soon be fed up and move on to
the competitors. To solve the problem,
00:18:29
Zappos is putting algorithm against each other in
a digital game of survival. A relevance text which
00:18:37
stimulates how users behave rewards the winning
algorithm by passing on its trait to the next
00:18:43
generation of algorithms. The best performing
algorithm goes live on the website until it
00:18:49
is superseded by a fitter one, continuously
improving the performance of the search engines.
00:18:56
Making better bets, so human routinely and often
effortlessly sort through probabilities and act
00:19:03
on the likeliest, even with relatively little
prior experience. Machines are now being taught
00:19:11
to mimic such reasoning through the application
of Gaussian processes that is probabilistic models
00:19:17
that can deal with extensive uncertainty, act on
space data and learn from experience. Example of
00:19:25
this is the project Loon by Alphabet that is the
Google's parent company. Then comes the closing in
00:19:32
on causation. AI is good at spotting correlations
and making valuable predictions based on them.
00:19:39
So that is what the problem with AI is.
For instance, GNS Healthcare, a Cambridge,
00:19:48
Massachusetts precision machine company uses
causal algorithms to help some of the world's
00:19:54
largest pharmaceutical companies understand
not only which patient responds to what drugs,
00:20:00
but also why do they respond. So
that is also important. So, what,
00:20:06
which patient will respond to what drug that
is one and why are they doing so. So that is
00:20:11
another important thing that it does so
what and why. Using Bayesian techniques,
00:20:20
this software platform translates data into
causal models, cause and effect models.
00:20:26
So this causal model is cause and effect
models. These techniques identify which
00:20:34
variables in a data set appears to have the
most influence on other variables. The next
00:20:41
comes taking on the time and space.
Researchers at the MIT IBM Watson AI
00:20:46
lab have recently developed a new technique for
training video recognition systems that is both
00:20:53
highly accurate and saves on computation
costs. The researchers were able to train
00:20:58
video recognition models three times faster
than existing state-of-the-art techniques.
00:21:04
So it has increased the efficiency of the
of the models. The increased speed could be
00:21:12
critical for the ability of autonomous vehicles
to recognize and react to visual information and
00:21:18
to predict what is likely to take place around
them. Retailers would benefit immensely from
00:21:25
such a video recognition technology. Next
comes calling on common sense. No AI system
00:21:32
currently deployed can reliably answer a
broad range of simple questions such as,
00:21:37
if I put my socks in a drawer, will
they still be there tomorrow? So that
00:21:42
is a simple question that the AI system
are not able to answer. How can you tell
00:21:48
if a milk carton is full? So, these are
some simple questions where the AI fails.
00:21:54
To help define what it means for machines to
have common sense, AI2 that is Allen Institute of
00:22:01
Artificial Intelligence is developing a portfolio
of tasks against which progress can be measured.
00:22:08
The Defence Advanced Research Project Agency,
DARPA, is investing $20 to $2 billion in AI
00:22:17
research. In its Common Machine Sense program,
researchers will create models that mimic core
00:22:24
domain of human cognition, including the domain
of objects that is intuitive physics, places,
00:22:33
spatial navigation and agents’ intentional
actors. So, this is what this machine common
00:22:40
sense program is trying to achieve. Researchers
at Microsoft and MacGill University have jointly
00:22:46
developed a system that has shown great promise
for untangling ambiguities in natural languages.
00:22:53
Untangling ambiguities in natural language,
a problem that requires diverse forms of
00:22:57
inference and knowledge. With such systems,
the goal is to produce both the right answer
00:23:03
and the rationale for the answer. It is not only
about the right answer but also the rationale for
00:23:08
that answer. Consider chatbots and voice digital
assistants, which often leaves users frustrated.
00:23:16
They would immensely benefit from these
advancements. Tracking of emotions. Affectiva,
00:23:23
a Boston startup founded in 2009 by
researchers from MIT framed Media
00:23:29
Lab and acquired in mid-2021 by Swedish
company SmartEye, is currently working
00:23:35
on an AI system designed to read emotions. So
now we are moving on to reading the emotions.
00:23:42
Affectiva algorithms read people faces to detect
their emotional and other cognitive states.
00:23:49
The technology is being used in
AI-assisted semi-autonomous cars. So,
00:23:54
we are not talking of only the autonomous car,
we are talking of semi-autonomous cars. The
00:23:59
company has also developed an emotion tracking
system that enables media and advertisers to
00:24:06
test responses to their programming and
video ads with the target audience. The
00:24:12
system is based on the analysis of more
than 7.6 million faces in 87 countries.
00:24:18
So this is the data set that is being used
for this kind of study. About one-fourth
00:24:25
of the Fortune Global 500 have used the
technology to test their ads around the
00:24:30
world and help them predict purchase
intent, sales lift or the likelihood
00:24:37
of content to go viral. The next component
of IDR's framework is data. Business data
00:24:44
is often locked in legacy on-site platforms
that are siloed, making it difficult if not
00:24:51
impossible for employees to get different
types of data to work together. Creating
00:24:56
a robust data foundation requires breaking
information out of traditional file silos.
00:25:01
So this is what is required. So that it
can be unified, one, optimally stored,
00:25:09
two and easily accessed three and readily
analyzed with new tools all in the cloud.
00:25:17
With a solid data foundation, more data from
more sources managed with the help of AI and
00:25:23
widely disseminated within an organization
can help in maximizing data potential. Now
00:25:30
let us look at the case of McDonald's. In
2018 McDonald was facing a major challenge.
00:25:39
Its competitors have used online delivery to
leapfrog its lock on the fast-food markets.
00:25:45
So the competitor were given online delivery
using online delivery for the competitors were
00:25:50
using online delivery. So, McDonald leaders
quickly devised an online delivery solution
00:25:55
through a global partnership with Uber Eats
that by 2019 was adding dollar 4 billion
00:26:03
to the annual sales. But top executives
knew that the company's long-term future
00:26:09
depends on making a rapid and complete
transformation to become data driven. So,
00:26:14
this kind of stopgap arrangement will not
last long and therefore something more
00:26:26
needs to be done and that is a complete
transformation to become data driven.
00:26:31
This required reconfiguring its restaurants
into enormous data processes complete with
00:26:36
machine learning and mobile technology to support
highly customized customer orders and delivery.
00:26:43
Data crunching also aids in calculating
how external factors from weather to big
00:26:49
sporting events would impact demand and
restaurants ability to serve customers.
00:26:54
So now you see that we are also looking
at the effect of external factors like
00:26:59
the change in weather or a big sporting
event and then predicting and forecasting
00:27:05
the restaurants’ ability to serve the
customer. And gathering and processing
00:27:10
data were important for developing new
products and initiatives that could be
00:27:13
immediately successful. Now implications of using
data driven approach is first is the financial
00:27:19
success. Within two years the transformation
effort successfully achieved financial results.
00:27:26
Few companies in the S&P 500 have outperformed
McDonald creating a modern data foundation. So,
00:27:33
mastering the use of big and small data
to generate value from AI requires that
00:27:38
organization lay a solid foundation and for that
three capabilities are key. First is modern data
00:27:45
engineering, second is AI assisted data governance
and the third is data democratization. So,
00:27:53
in modern data foundation data comes from a
variety of internal and external sources through a
00:27:59
number of organisms including batch and real time
processing and APIs. It get stitched together into
00:28:07
highly curated and reusable data sets that can
be consumed for a variety of analytic purposes.
00:28:13
A good foundation relies on reusable frameworks
for data ingestion and ETL that is extract,
00:28:20
transform and load that support diverse data
types. These frameworks also handle rules of
00:28:27
data quality and standardization so that
new data pipelines for analytic use cases
00:28:33
and data products can be developed quickly and at
scale on the cloud. AI assisted data governance,
00:28:40
cloud-based AI tools offer the
advanced capabilities and scale
00:28:44
to help automatically. Cleans, classify and
secure data gathered on the cloud as it is
00:28:52
ingested which supports better quality
data veracity and ethical handling.
00:28:58
Data democratization. A modern data foundation
gets more data into more hands. It makes data
00:29:05
accessible and easy to use in a timely manner
while enabling multiple ways to consume data
00:29:11
including self-service, AI, business intelligence
and data science. The latest cloud-based tools
00:29:18
democratize data and empower more people across
the enterprise to easily find and leverage data
00:29:24
that is relevant to their specific business needs
faster, create a modern data foundation. Together
00:29:31
these three capabilities help companies overcome
some of the most common barriers to value.
00:29:36
Data accessibility, data trustworthiness, data
readiness and data timeliness. So, these are
00:29:43
the four things. They enable companies to blend
data from big and small data sets together in
00:29:52
real time, build agile reporting and leverage AI
to create broadly accessible customers, markets
00:30:01
and operational insights that deliver meaningful
business outcomes. In order to conclude, we have
00:30:07
discussed the intelligence and data landscapes of
IDEA’s framework. There are many shortcomings of
00:30:12
deep learning-based AI. Machine intelligence is
now being augmented with human intelligence and
00:30:18
human have superior cognitive intelligence
and machines excels in pattern recognition.
00:30:24
We have also studied the implementation
of smart robot augmented with human
00:30:28
intelligence at Obeta helped them to increase
reliability and speed. Three key capabilities
00:30:35
are required to build a modern-day data
foundation. One is modern data engineering,
00:30:40
AI assisted data governance and the third
is data democratization. These are the five
00:30:45
books from which the material for
this module was taken. Thank you.
![The Complete Web Development Roadmap [2024]](https://ruhnrawpgxrzdrgaohnf.supabase.co/storage/v1/object/public/images/video/GxmfcnU3feo/thumbnail.jpg)