3. AI is getting
everywhere
inside the firm!
While current adoption
levels in Marketing are
comparatively lower
(25% widescale adoption
to AI being critical)
compared to most other
functions, it is projected
to rise up to 64% and
touch adoption levels
similar to most other
functions.
Source: https://www.insiderintelligence.com/content/5-charts-ai-role-marketing
4. Where all is AI entering
the Marketing?
• At a high-level, AI could be used for
marketing automation, personalization
and forecasting.
• Marketers are using AI to help with
their more top-of-funnel efforts, but
that may soon change.
• AI could take on a bigger role in
measurement and analytics as marketers
attempt to optimize their campaigns.
Source: https://www.insiderintelligence.com/content/5-charts-ai-role-marketing
5. What is driving this?
• Gather data-rich insights in a fraction of the time: AI also plays a pivotal role in
data analytics and performance measurement. AI enables marketers to track
campaign performance from mass-market messaging right down to individual social
media posts. With the power of AI, marketers can leverage thousands of data points
to optimise their measurement framework according to the targets and metrics that
matter to the business.
• Understand your target audience better: The data collected by AI can help
provide us with a better understanding of our customers by collecting and analysing
their social, behavioural and sales data so you can target your message to the right
audience and better anticipate their needs.
• Streamline operations: Data drives modern marketing practices and AI plays a
central role in achieving better business efficiencies and delivering better outcomes
across marketing operations. AI can help streamline marketing operations as it will
minimise the inefficiencies in your business’ operations allowing for more time to be
spent on strategic action.
Source: https://www.salesforce.com/in/resources/guides/role-of-ai-in-marketing/
6. What’s
impeding it?
• More than half (54%) of executives
worldwide feel their companies lack
the expertise and talent to
implement AI, according to a
Boston Consulting Group and MIT
Sloan Management Review survey.
Other barriers include a lack of
training and knowledge among staff
members, buy-in from senior
leaders, funding or resources, and
awareness about AI.
• These barriers are unlikely to change
in the year ahead as budgets are
tapped, margins are tighter than
ever, and layoffs continue.
8. ChatGPT Demo
• ChatGPT interacts in a conversational way.
The dialogue format makes it possible for
ChatGPT to answer followup questions, admit
its mistakes, challenge incorrect premises, and
reject inappropriate requests. ChatGPT is a
sibling model to InstructGPT, which is trained
to follow an instruction in a prompt and
provide a detailed response.
• Since its public beta in late 2022, it has
become among the fastest adopted hi-tech
products!
Source: https://openai.com/blog/chatgpt/
9. How ChatGPT is pre-trained?
• We trained this model using Reinforcement Learning from Human Feedback
(RLHF), using the same methods as InstructGPT, but with slight differences in the
data collection setup. We trained an initial model using supervised fine-tuning:
human AI trainers provided conversations in which they played both sides—the user
and an AI assistant. We gave the trainers access to model-written suggestions to help
them compose their responses. We mixed this new dialogue dataset with the
InstructGPT dataset, which we transformed into a dialogue format.
• To create a reward model for reinforcement learning, we needed to collect
comparison data, which consisted of two or more model responses ranked by quality.
To collect this data, we took conversations that AI trainers had with the chatbot. We
randomly selected a model-written message, sampled several alternative completions,
and had AI trainers rank them. Using these reward models, we can fine-tune the
model using Proximal Policy Optimization. We performed several iterations of
this process.
Source: https://openai.com/blog/chatgpt/
11. Limitations of ChatGPT
• ChatGPT sometimes writes plausible-sounding but incorrect or nonsensical answers. Fixing this
issue is challenging, as: (1) during RL training, there’s currently no source of truth; (2) training the
model to be more cautious causes it to decline questions that it can answer correctly; and (3)
supervised training misleads the model because the ideal answer depends on what the model knows,
rather than what the human demonstrator knows.
• ChatGPT is sensitive to tweaks to the input phrasing or attempting the same prompt multiple
times. For example, given one phrasing of a question, the model can claim to not know the answer,
but given a slight rephrase, can answer correctly.
• The model is often excessively verbose and overuses certain phrases, such as restating that it’s a
language model trained by OpenAI. These issues arise from biases in the training data (trainers prefer
longer answers that look more comprehensive) and well-known over-optimization issues.12
• Ideally, the model would ask clarifying questions when the user provided an ambiguous query. Instead,
our current models usually guess what the user intended.
• While we’ve made efforts to make the model refuse inappropriate requests, it will sometimes respond
to harmful instructions or exhibit biased behavior. We’re using the Moderation API to warn or block
certain types of unsafe content, but we expect it to have some false negatives and positives for
now. We’re eager to collect user feedback to aid our ongoing work to improve this system.
Source: https://openai.com/blog/chatgpt/
12. Dall-E 2 Demo
• DALL·E 2 is a new AI system that can create realistic images and art from a
description in natural language. It can combine concepts, attributes, and styles.
• DALL·E 2 has learned the relationship between images and the text used to describe
them. It uses a process called “diffusion,” which starts with a pattern of random dots
and gradually alters that pattern towards an image when it recognizes specific aspects
of that image.
• DALL·E 2 can expand images beyond what’s in the original canvas, creating
expansive new compositions.
• DALL·E 2 can make realistic edits to existing images from a natural language
caption. It can add and remove elements while taking shadows, reflections, and
textures into account.
• DALL·E 2 can take an image and create different variations of it inspired by the
original.
Source: https://openai.com/dall-e-2/
19. But, there’s
also a lot of
Hype!!!
https://www.gartner.com/en/articles/the-4-trends-that-prevail-on-the-gartner-hype-cycle-for-ai-2021
20. …with alarmingly high failure rates!
• Gartner: 85% of AI projects will fail and deliver erroneous outcomes through 2022.
• MIT SMR: 70% of companies report minimal or no impact from AI. Among the
90% of companies that have made some investment in AI, fewer than 2 out of 5
report business gains from AI in the past three years.
• VentureBeat: 87% of data science projects never make it into production. Tom
Siebel says 99% of internal AI projects fail!
• Andrew Ng (HBR): It is not unusual for teams to celebrate a successful proof of
concept, only to realize that they still have another 12-24 months of work before the
system can be deployed and maintained.
https://research.aimultiple.com/ai-fail/
21. …And Low ROI & Long Payback periods!
The ROI for AI projects varies greatly,
based on how much experience an
organization has.
Leaders showed an average of a 4.3%
ROI for their projects, compared to only
0.2% for beginning companies.
Payback periods also varied, with leaders
reporting a typical payback period of
1.2 years and beginners at 1.6 years.
https://www2.deloitte.com/us/en/insights/industry/technology/artificial-intelligence-roi.html
22. So, What is AI?
Hint: There is nothing really “Intelligent” about it!
23. What is “Artificial Intelligence”, Or AI?
Prof John McCarthy, Father of Artificial Intelligence:
• Intelligence is the computational part of the ability to achieve goals in the
world. Varying kinds and degrees of intelligence occur in people, many
animals and some machines.
• Artificial Intelligence is the science and engineering of making intelligent
machines, especially intelligent computer programs. It is related to the
similar task of using computers to understand human intelligence, but AI
does not have to confine itself to methods that are biologically observable.
http://jmc.stanford.edu/artificial-intelligence/what-is-ai/index.html
24. Good Old Fashioned AI (GOFAI)
• Symbolic artificial intelligence is the term for the collection of all methods
in artificial intelligence research that are based on high-level symbolic (human-
readable) representations of problems, logic and search. Symbolic AI was the
dominant paradigm of AI research from the mid-1950s until the late 1980s.[1][2]
• John Haugeland gave the name GOFAI ("Good Old-Fashioned Artificial
Intelligence") to symbolic AI in his 1985 book Artificial Intelligence: The Very Idea,
which explored the philosophical implications of artificial intelligence research.
In robotics the analogous term is GOFR ("Good Old-Fashioned Robotics").[3]
• Researchers in the 1960s and the 1970s were convinced that symbolic approaches
would eventually succeed in creating a machine with artificial general
intelligence and considered this the goal of their field.
• However, the symbolic approach would eventually be abandoned, largely because of
the technical limits of this approach. It was succeeded by highly mathematical
Statistical AI which is largely directed at specific problems with specific goals, rather
than general intelligence. Research into general intelligence is now studied in the
exploratory sub-field of artificial general intelligence.
Symbolic Artificial Intelligence, https://en.wikipedia.org/wiki/Symbolic_artificial_intelligence
26. Key milestones
• 1943: Walter Pitts & Warren McCullogh develop a computer model based on Neural Networks of human brain using a combination of algorithms and maths they called “threshold logic” to mimic the
thought process.
• 1950: Alan Turing proposes the imitation game, aka “Turing Test”
• 1952: Hodekin-Huxley paper of brain as neurons forming an electrical network
• 1956: John McCarthy coins the term “Artificial Intelligence” and organizes Dartmouth Summer Research Project, the first conference on AI.
• 1960s: Research labs established at MIT, Stanford, SRI, etc. to mimic human intelligence by problem-solving or playing games like checkers or chess.
• 1960: Henry Kelley develops the basics of continuous Back Propagation (backprop) model
• 1962: Stuart Dreyfus develops chain rule to simplify backprop.
• 1965: Alexey Grigoryevich Ivakhnenko & Valentin Grigorʹevich Lapa develop Deep Learning algorithms using polynomial activation functions and statistical analysis at each layer.
• 1970s: MYCIN was able to diagnose certain kinds of bacterial infections based on symptoms input.
• 1970s: A “prospector” expert system uncovers a hidden mineral deposit of porphyr molybdenum (a form of copper deposit) at Mount Tolman in the state of Washington.
• 1973-80s: First “AI Winter”
• 1981: John Searle proposes “Chinese Room”
• 1980s: Development of Expert Systems bring some successes (e.g. DEC’s XCON)
• 1985-90s: Second “AI Winter”
• 1979: Kunihiko Fukushima develops “Neocognitron” the first Convolutional Neural Network (CNN) with multiple pooling and convolutional networks that allows computer to “learn” to recognize visual
patterns using manually-adjustable “weights” of certain connections.
• 1990s: Focus shifts to “Intelligent Agents”
• 1997: IBM Deep Blue beats World Chess Champion Garry Kasparov (Artificial Intelligence)
• 2011: IBM Watson beats human players on US game show Jeopardy (Machine Learning)
• 2012: Deep Learning
• 2014: Ian Goodfellow creates Generative Adversarial Networks (GANs)
• 2016: Google’s AlphaGo beats boardgame Go master Lee Sedol (Deep Learning)
28. AI, ML, NN, DL…
• Machine learning, deep learning, and neural networks are all sub-fields
of artificial intelligence. However, deep learning is actually a sub-field
of machine learning, and neural networks is a sub-field of deep
learning.
• The way in which deep learning and machine learning differ is in how
each algorithm learns. Deep learning automates much of the feature
extraction piece of the process, eliminating some of the manual human
intervention required and enabling the use of larger data sets.
• Classical, or "non-deep", machine learning is more dependent on
human intervention to learn. Human experts determine the set of
features to understand the differences between data inputs, usually
requiring more structured data to learn.
Machine Learning, https://www.ibm.com/cloud/learn/machine-learning
29. Machine Learning
• Term coined by Arthur Samuel in 1959
• Machine learning is a branch of artificial intelligence (AI) and computer
science which focuses on the use of data and algorithms to imitate the
way that humans learn, gradually improving its accuracy.
• Through the use of statistical methods, algorithms are trained to make
classifications or predictions, uncovering key insights within data mining
projects. These insights subsequently drive decision making within
applications and businesses, ideally impacting key growth metrics.
https://www.ibm.com/cloud/learn/machine-learning
30. Potential of AI
•AI is the new electricity – Andrew Ng
•AI is more profound than fire or electricity – Sundar
Pichai
•People should stop training radiologists – Geoff Hinton,
2016, https://youtu.be/2HMPRXstSvQ
31. Where to use AI?
If a typical person can do a mental task with less
than one second of thought, we can probably
automate it using AI either now or in the near
future. – Andrew Ng
https://hbr.org/2016/11/what-artificial-intelligence-can-and-cant-do-right-now
32. So, why now?
After decades of start/stops, finally AI seems to be at the cusp of its (third) resurgence.
33. AI is the New Software!
• Machine learning was a huge leap from programmed instructions and if-
then statements that merely simulated the very human process of thinking
and making decisions.
• With machine learning, the machine no longer needs to be explicitly
programmed to complete a task; it can pour through massive data sets
and create its own understanding. It can learn from the data and create its
own model, one that represents the different rules to explain relationships
among data and use those rules to draw conclusions and make decisions
and predictions.
• A machine learning algorithm is a mathematical function that enables the
machine to identify relationships among inputs and outputs. The
programmer’s role has shifted from one of writing explicit
instructions to creating and choosing the right algorithms.
Rose, Doug. Artificial Intelligence for Business: What You Need to Know about Machine Learning and Neural Networks.
34. How is AI different from Software?
Traditional Software AI Software
Reasoning Deductive Inductive
Inputs Data + Program Data + Output
Logic Manually pre-programmed to perform a
specific task on a given dataset
Programmed to automatically keep learning rules
from a given dataset
Output Output Models, Rules
Learning Learns one-time from the programmer “Learns” constantly being the data
Resource Code Data
Solutions Deterministic Probabilistic
Output Consistently remains the same Can improve with usage (or degrade over time)
Business
model
One-time development efforts, followed by
multiple sales, and small maintenance effort
(optional)
Each project is one-off, and needs full lifecycle
management mandatorily
35. Limitations
of AI
When people talk about AI, machine learning, automation,
big data, cognitive computing, or deep learning, they’re
talking about the ability of machines to learn to fulfill
objectives based on data and reasoning. This is
tremendously important, and is already changing business
in practically every industry. In spite of all the bold
claims, there remain several core problems at the
heart of Artificial Intelligence where little progress
has been made (including learning by analogy, and
natural language understanding). Machine learning
isn’t magic, and the truth is we have neither the data
nor the understanding necessary to build machines
that make routine decisions as well as human beings.
https://hbr.org/2016/11/how-to-make-your-company-
machine-learning-ready
36. Failure Modes of AI
• Brittleness
• Embedded Bias
• Catastrophic Forgetting
• Explainability
• Quantifying Uncertainty
• Common Sense
• Math
Source: https://spectrum.ieee.org/ai-failures
37. Challenges
• Unavailability of Skills & Talent
• Technology is quite good, but still maturing
• Data is often siloed or not available, or poor quality, etc.
• Business case, and alignment of strategy and business model to an AI-
firm
• Societal concerns like ethics, privacy, transparency, bias, surveillance, etc.
• Governance issues such as data management, regulations, legal
frameworks, etc.
• …..and many many more…!!!
39. OK, what is AI in Marketing?
• AI is a broad term that can be used to describe any technology that can learn and make
automated decisions based on past learnings. All learning that an AI will develop comes from
the data it receives. The more data the AI gathers, the faster it can adapt to fit the needs of
your audience. Unlike human employees, AI solutions operate 24/7 and become increasingly
capable of more complex functions over time. It’s these kinds of capabilities that make
artificial intelligence so powerful. (Salesforce)
• AI marketing uses artificial intelligence technologies to make automated decisions based
on data collection, data analysis, and additional observations of audience or
economic trends that may impact marketing efforts. AI is often used in digital
marketing efforts where speed is essential. AI marketing tools use data and customer profiles
to learn how to best communicate with customers, then serve them tailored messages at the
right time without intervention from marketing team members, ensuring maximum
efficiency. For many of today’s digital marketers, AI is used to augment marketing teams or
to perform more tactical tasks that require less human nuance. (Marketing Evolution)
40. Where all can AI help me?
• Reach: attract visitors with a range
of inbound techniques
• AI-generated content
• Smart content curation
• Voice search
• Programmatic media buying
• Act: draw visitors in and make them
aware of your product
• Propensity modeling
• Predictive analytics
• Lead scoring
• Ad targeting
• Convert: nudge interested consumers
into becoming customers
• Dynamic pricing
• Web and app personalization
• Chatbots
• Re-targeting
• Engage: keep your customers
returning
• Predictive customer service
• Marketing automation
• 1:1 dynamic emails
• …?
Source: https://www.smartinsights.com/managing-digital-marketing/marketing-innovation/15-applications-artificial-intelligence-marketing/
43. Four kinds of Marketing AI
• Categorizing applications according to
their intelligence levels and structure can
help companies plan the rollout of their
marketing AI.
• Simple stand-alone apps are a good place
to begin because they’re easier to set up,
but their benefits are limited.
• Once companies acquire AI skills and
amass data, they can add apps that are
more advanced and are part of other
platforms, working their way up to
integrated machine learning, which has the
potential to create the most value.
Source: https://hbr.org/2021/07/how-to-design-an-ai-marketing-strategy
44. “Marketing
Myopia”
• Half the money I spend on advertising
is wasted; the trouble is, I don't know
which half. – John Wannamaker,
1938 - 1922
• A nearsighted focus on selling
products and services, rather than
seeing the “big picture” of what
consumers really want – Theodore
Levitt, 1962
• Marketers have lost the forest for the
trees, focusing too much on creating
products for narrow demographic
segments rather than satisfying needs.
– Clayton Christensen, 2006
45. YMMV! (“Your mileage may vary”)
“While marketing AI holds enormous promise, we urge CMOs to be realistic about
its current capabilities. Despite the hype, AI can still accomplish only narrow tasks, not
run an entire marketing function or process. Nevertheless, it’s already offering
substantial benefits to marketers—and in fact is essential in some marketing
activities—and its capabilities are rapidly growing. We believe that AI will ultimately
transform marketing, but it’s a journey that will take decades. The marketing
function and the organizations that support it, IT in particular, will need to pay long-
term attention to building AI capabilities and addressing any potential risks. We urge
marketers to start developing a strategy today to take advantage of AI’s current
functionality and its likely future.”
– Davenport, T. H., Guha, A., and Grewal, D. (2021). How to Design an AI marketing
strategy. Harvard Business Review. Jul-Aug 2021.