Introduction to AI

Seller presentation

Dirk deRoos
WW Technology Sales Enablement
Data & AI Manager
[email protected]

Gina Hawkins
WW Technology Sales Enablement
[email protected]
Seller guidance Slides in this presentation marked as
"IBM and Business Partner Internal Use
Only" are for IBM and Business Partner
References in this presentation to IBM
products, programs, or services do not
imply that they will be available in all

and legal use and should not be shared with clients

or anyone else outside of IBM or the
Business Partners’ company.
countries in which IBM operates. Product
release dates and/or capabilities referenced
in this presentation may change at any time

disclaimer
at IBM’s sole discretion based on market
© IBM Corporation 2023. opportunities or other factors and are not
All Rights Reserved. intended to be a commitment to future
product or feature availability in any way.
The information contained in this Nothing contained in these materials is
publication is provided for informational intended to, nor shall have the effect of,
purposes only. While efforts were made stating or implying that any activities
to verify the completeness and accuracy undertaken by you will result in any specific
of the information contained in this sales, revenue growth, or other results.
IBM and Business Partner publication, it is provided AS IS without
warranty of any kind, express or implied. All client examples described are presented
Internal Use Only In addition, this information is based on as illustrations of how those clients have
IBM’s current product plans and strategy, used IBM products and the results, they
which are subject to change by IBM without may have achieved. Actual environmental
notice. IBM shall not be responsible for costs and performance characteristics
any damages arising out of the use of, or may vary by client.
otherwise related to, this publication or
any other materials. Nothing contained in
this publication is intended to, nor shall
have the effect of, creating any warranties
or representations from IBM or its suppliers
or licensors, or altering the terms and
conditions of the applicable license
agreement governing the use of
IBM software.
What is AI?
AI refers to the ability
of computer systems
to attempt to mimic
the problem-solving
and decision-making Computer vision Data science
capabilities of the
human mind.

Natural language Robotics

processing (NLP)
AI milestones

1956 1980s 2011 2022

John McCarthy coins Neural networks IBM Watson beats OpenAI releases
the term 'artificial become widely used champions Ken ChatGPT, a generative
intelligence' at the first- in AI applications. Jennings and Brad AI chatbot based on a
ever AI conference at Rutter at Jeopardy! large language model.
Dartmouth College.

1950 1967 1997 2016

Alan Turing published the Frank Rosenblatt builds IBM's Deep Blue DeepMind's AlphaGo
Computing Machinery the Mark 1 Perceptron, supercomputer beats, program, powered
and Intelligence paper the first computer based then world chess by a deep neural
introducing the question: on a neural network champion, Garry network, beats Lee
“Can machines think?” that ‘learned’ through Kasparov in a chess
Sedol, the world
trial and error. match (and rematch).
champion Go player,
in a five-game match.
Artificial Intelligence (AI) AI can be defined as a technique that enables machines to mimic
cognitive functions associated with human minds – cognitive functions
Human intelligence exhibited by machines
include all aspects of learning, reasoning, perceiving, and problem solving.

Machine Learning (ML) ML-based systems are trained on historical data to uncover patterns.
Users provide inputs to the ML system, which then applies these inputs
Systems that learn from historical data
to the discovered patterns and generates corresponding outputs.

Deep Learning (DL) DL is a subset of ML, using multiple layers of neural networks, which are
interconnected nodes, which work together to process information. DL is
ML technique that mimics
well suited to complex applications, like image and speech recognition.
human brain function

Foundation Model
Generative AI systems

AI model built using a specific kind of neural network architecture,

called a transformer, which is designed to generate sequences
of related data elements (for example, like a sentence).

1950’s 1980’s 2010’s 2020’s

 Rules-based systems Machine learning systems

programmed with a series trained with a

of instructions & logical rules series of examples
for example: “a triangle has three sides” for example: “here are
pictures of many triangles”
What is a machine
learning (ML) model?
Example: Spam detection for email
Input Deployed ML model Output
new email spam detection valid email or spam

valid

spam

Labeled historical data Machine learning algorithm

valid and spam emails one of multiple possible
candidate algorithms

Machine learning model training

iterative process of experimenting and testing
Machine learning (ML) types
Classification models
Example: Spam detection for email

Input Deployed ML model Output

Classification models assign labels to model new email spam detection valid email or spam
inputs or assign them to specific categories.
Common use cases include: valid

• Fraud detection: predict whether a transaction

is fraudulent based on patterns in the data
spam
• Sentiment analysis: classify text as
positive, negative, or neutral
• Medical diagnosis: assign a disease
label to a patient’s case, based on
symptoms and medical history
• Image recognition: recognize objects
or identify people based on visible Labeled historical data Machine learning algorithm
features and characteristics valid and spam emails

Machine learning model training

iterative process of experimenting and testing
Machine learning (ML) types
Generative models
Example: Text generation

Generative models create new data in Input Deployed Output

response to input requests (prompts). prompt foundation model answer

Common use cases include:

• Text generation: generate new text
in response to a prompt, summarizing
text, or writing a lengthy essay
• Code generation: generate computer
code based on a textual description
of the proposed program
• Image generation: create images
based on the prompt

Unlabeled text data Machine learning algorithm

Machine learning model training

iterative process of experimenting and testing
Machine learning (ML) types
Regression models
Example: House price prediction

Input Deployed ML model Output

Regression models make predictions address & date house price predictor house price
based on the model input.
Common use cases include:
• Stock market analysis: securities
price prediction based on historical
data or news events
• Sales: forecasting based on
historical data or market trends
• Healthcare: predict patient outcomes
based on factors such as age, gender,
medical history, or treatment plans
• Customer behavior analysis: predict Labeled historical data Machine learning algorithm
future customer purchasing patterns real estate market data

based on demographic data, past purchase

history, and advertising campaigns Machine learning model training
iterative process of experimenting and testing
Machine learning (ML) types
Clustering models
Example: Market basket analysis

Clustering models identifies distinct groupings

of individual data points that share common
characteristics within a larger data set.
Common use cases include:
• Customer segmentation: group customers
based on similar preference, behaviors, Unlabeled Clustering Output
and demographics historical data algorithm items commonly
transaction data purchased together
• Genetic analysis: group genes with
similar functions or processes
• Social network analysis: identify communities
or groups within a social network
• Market basket analysis: identify items
that are commonly purchased together
Machine learning
operations (MLOps)
lifecycle Monitor model Find data

Use math, statistics, specialized

programming, and analytics to
extract meaningful insights
from data.
MLOps lifecycle
Designed to build, deploy, Deploy model Prepare data
manage, and monitor
ML models.

Test model Train model

Machine learning Supervised Semi-supervised
learning learning
(ML) methods
An operator provides By using a combination
the ML algorithm with of labeled and
a known dataset that unlabeled data, ML
includes desired algorithms can learn to
inputs and outputs. label unlabeled data.

Unsupervised Reinforcement
learning learning
The machine determines A technique that
the correlations and teaches an AI model
relationships by to find the best result
analyzing available data. by trial and error.
Large language
models (LLMs)

LLMs are machine learning

models that have been
trained on large quantities
of unlabeled text using
self-supervised learning
and can perform a variety
of natural language
processing tasks .

ChatGPT is an example of
a generative AI chatbot
developed using an LLM.
 What’s next with foundation models:

Conventional AI Zero-shot prompting Few-shot prompting Data-driven tuning

1000s – 100000s labeled No labeled data needed 1-10 labeled data points / 100s – 1000s labeled
data points / task task data points / task

Task Task Task Task Task Task

Task Task

Labeled tuning data

Model Model Handcrafted examples

Labeled Labeled
training training Pre-trained generative foundation model
data data
Traditional AI
applications
Speech Customer
recognition service
.

Computer Recommendation Automated stock

vision engines trading
Natural
language
processing

The technology that

gives computers the
ability to understand
text and spoken words
in much the same
way humans can.
Natural
language
processing Spam Machine
use cases detection translation

Chatbots and Social media Text

virtual assistants sentiment analysis summarization
Conversational
AI

Chatbots and virtual

assistants or agents,
can make it easy for
users to find the
information they need
by responding to their
questions and requests
through text or audio
input without the need
for human intervention,
providing 24-hour
availability to users.
Conversational
AI use cases
Customer support Accessibility HR processes

Healthcare IoT devices Computer software

Generative AI
applications
Text Code
generation generation

Music Generate Image & video

generation simulated data generation
Glossary of • Algorithm
AI terms •
•
Artificial general intelligence (AGI)
Artificial intelligence (AI)
• Artificial narrow intelligence (ANI)
• Artificial super intelligence (ASI)
• Bias
• Convolutional neural network (CNN)
• Deep learning (DL)
• Foundation model
• Generative AI
• Hallucination
• Large language model (LLM)
• Machine learning (ML)
• Natural language generation (NLG)
• Natural language processing (NLP)
• Natural language understanding (NLU)
• Neural networks
• Parameters
• Transformer model
These videos serve as a com plem ent to the learning in in this presentation

Resources

Review these videos to • AI vs Machine Learning

continue learning about • Natural Language Processing vs Natural
the concepts introduced
in this presentation. Language Understanding and Generation
• What are Generative AI models?
• What are Transformers (Machine Learning
Model)?
• Risks of Large Language Models (LLM)
• Why Large Language Models Hallucinate
• Artificial Intelligence – Are We There Yet?
Let’s get
started!