Introduction to AI learning

Week 1

Week 2
But with today's wealth of big data and the need for more complex evidence-based decisions, such a rigid
approach often breaks or fails to keep up with available information. Cognitive Computing enables people
to create a profoundly new kind of value, finding answers and insights locked away in volumes of data.
Conventional computing solutions, based on the mathematical principles that emanate from the 1940's,
are programmed based on rules and logic intended to derive mathematically precise answers, often
following a rigid decision tree approach.

Cognitive computing - we go through four key steps.

1. First, we observe visible phenomena and bodies of evidence.
2. Second, we draw on what we know to interpret what we are seeing to generate hypotheses
about what it means.
3. Third, we evaluate which hypotheses are right or wrong.
4. Finally, we decide, choosing the option that seems best and acting accordingly. Just as humans
become experts by going through the process of observation, evaluation, and decision-making,
cognitive systems use similar processes to reason about the information they read, and they can
do this at massive speed and scale.
Unlike conventional computing solutions, which can only handle neatly organized structured data
such as what is stored in a database, cognitive computing solutions can understand unstructured
data, which is 80 percent of data today. All of the information that is produced primarily by humans
for other humans to consume. This includes everything from literature, articles, research reports to
blogs, posts, and tweets. While structured data is governed by well-defined fields that contain well-
specified information, cognitive systems rely on natural language, which is governed by rules of
grammar, context, and culture. It is implicit, ambiguous, complex, and a challenge to process. While
all human language is difficult to parse, certain idioms can be particularly challenging. In English for
instance, we can feel blue because it's raining cats and dogs, while we're filling in a form, someone
asked us to fill out. Cognitive systems read and interpret text like a person. They do this by breaking
down a sentence grammatically, relationally, and structurally, discerning meaning from the semantics
of the written material. Cognitive systems understand context. This is very different from simple
speech recognition, which is how a computer translates human speech into a set of words. Cognitive
systems try to understand the real intent of the users language, and use that understanding to draw
inferences through a broad array of linguistic models and algorithms. Cognitive systems learn, adapt,
and keep getting smarter. They do this by learning from their interactions with us, and from their own
successes and failures, just like humans do.
 Read and interpret unstructured data, understanding not just the meaning of words but also the
intent and context in which they are used.
 Reason about problems in a way that humans reason and make decisions.
 Learn over time from their interactions with humans and keep getting smarter.

Terminology and Related Concepts of AI

AI enables a world where machines cannot only:
- understand human language,
- but also predict our needs,
- recognize our faces,
- and provide safety against threats.

AI in autonomous vehicles:
Machine learning
Deep learning
Natural language processing
Computer vision

Artificial intelligence is a branch of computer science that focuses on creating systems capable of
performing tasks that typically require human intelligence. AI systems will typically demonstrate
behaviors associated with human intelligence, such as planning, learning, reasoning, problem-solving,
knowledge representation, perception, motion, manipulation, social intelligence, and creativity.

Narrow AI or weak AI perform specific tasks. General AI, also known as strong AI, possesses human-
like cognitive skills and is capable of learning and adapting across various tasks, whereas super AI or
conscious AI aims to surpass human intelligence and is currently theoretical and not yet realized.

Machine learning is a subset of AI that uses computer algorithms to analyze data and make intelligent
decisions based on what it has learned without being explicitly programmed. Machine learning
algorithms are trained with large datasets and learn from examples. They do not follow rule-based
algorithms. Machine learning enables machines to solve problems on their own and make accurate
predictions using the provided data.
Deep learning – is an important concept and a specialized subset of machine learning. It uses multi-
layered neural networks, known as deep neural networks to analyze complex data and simulate
human decision-making. Deep learning algorithms can label and categorize information and identify
complex patterns in data. It enables AI systems to learn continuously and improve the quality and
accuracy of results by evaluating the correctness of decisions.

Neural networks - a computation model inspired by the human brain's neural structure. Neural
networks consist of interconnected nodes or neurons and have three layers. While the input layer
receives and processes raw data, hidden layers perform complex computations and transform data.
The output layer converts processed data into output format, producing the result.

Supervised learning
Regression – looking at relationships between features X and result Y (dependent and
independent variables) Y is continous
Classification – focuses on discrete values it identifies. Is the process of predicting the class of
given data points.
Forms:
Decision trees
Support vectors
Logistic regression
Random forests
Machine learning data sets:
 Training – showing some algorithms(data) to be True and false (giving examples) / Training
subsets are the data used to train the algorithm.
Validation – is used to validate our results and fine-tune the algorithm’s parameters.
Test sets – is data the model has never seen before and used to evaluate how good our model is.
(Accuracy, precision and recall)
Neural network – structures that imitate human brain
Unsupervised learning (Clustering)
Reinforcement learning

Deep learning – Specialized subset of machine learning. Enables natural language understanding
capabilities of AI systems by analyzing unstructured data. Understands context and intent being
conveyed
When creating deep learning algorithms, developers and engineers configure the number of layers
and the type of functions that connect the outputs of each layer to the inputs of the next.
Imagine captioning
Voice recognition and Transcription
Facial recognition
Medical imagining
Neural networks are the reason deep learning algorithms become more efficient as the datasets
increase in volume.
An artificial neural network is a collection of smaller units called neurons, which are computing
units modeled on the way the human brain processes information. Artificial neural networks borrow
some ideas from the biological neural network of the brain, in order to approximate some of its
processing results. These units or neurons take incoming data like the biological neural networks and
learn to make decisions over time. Neural networks learn through a process called backpropagation.
Backpropagation uses a set of training data that match known inputs to desired outputs. First, the
inputs are plugged into the network and outputs are determined. Then, an error function determines
how far the given output is from the desired output. Finally, adjustments are made in order to reduce
errors. A collection of neurons is called a layer, and a layer takes in an input and provides an output.
Any neural network will have one input layer and one output layer. It will also have one or more
hidden layers which simulate the types of activity that goes on in the human brain. Hidden layers take
in a set of weighted inputs and produce an output through an activation function. A neural network
having more than one hidden layer is referred to as a deep neural network. Perceptrons are the
simplest and oldest types of neural networks. They are single-layered neural networks consisting of
input nodes connected directly to an output node. Input layers forward the input values to the next
layer, by means of multiplying by a weight and summing the results. Hidden layers receive input from
other nodes and forward their output to other nodes. Hidden and output nodes have a property called
 bias, which is a special type of weight that applies to a node after the other inputs are considered.
Finally, an activation function determines how a node responds to its inputs. The function is run
against the sum of the inputs and bias, and then the result is forwarded as an output. Activation
functions can take different forms, and choosing them is a critical component to the success of a
neural network. Convolutional neural networks or CNNs are multilayer neural networks that take
inspiration from the animal visual cortex. CNNs are useful in applications such as image processing,
video recognition, and natural language processing. A convolution is a mathematical operation, where
a function is applied to another function and the result is a mixture of the two functions. Convolutions
are good at detecting simple structures in an image, and putting those simple features together to
construct more complex features. In a convolutional network, this process occurs over a series of
layers, each of which conducts a convolution on the output of the previous layer. CNNs are adept at
building complex features from less complex ones. Recurrent neural networks or RNNs, are recurrent
because they perform the same task for every element of a sequence, with prior outputs feeding
subsequent stage inputs. In a general neural network, an input is processed through a number of layers
and an output is produced with an assumption that the two successive inputs are independent of each
other, but that may not hold true in certain scenarios. For example, when we need to consider the
context in which a word has been spoken, in such scenarios, dependence on previous observations has
to be considered to produce the output. RNNs can make use of information in long sequences, each
layer of the network representing the observation at a certain time.

Quiz 2:

Question 1
Which of these statements is true?
1 / 1 point
Cognitive systems can derive mathematically precise answers following a rigid decision tree approach
Cognitive systems can learn from their successes and failures
Cognitive systems can only translate small volumes of audio data into their literal text translations at massive
speeds
Cognitive systems can only process neatly organized structured data
Correct
Cognitive systems learn, adapt, and keep getting smarter by learning from their interactions with us and from
their own successes and failures, just like humans do.

2.
Question 2
Which of these statements is true?
1 / 1 point
AI is the subset of Data Science that uses Deep Learning algorithms on structured big data
Artificial Intelligence and Machine Learning refer to the same thing since both the terms are often used
interchangeably
Data Science is a subset of AI that uses machine learning algorithms to extract meaning and draw inferences
from data
Deep Learning is a specialized subset of Machine Learning that uses layered neural networks to simulate
human decision-making
Correct
Deep Learning enables machines to continuously learn on the job and improve the quality and accuracy of
results by determining whether decisions were correct.
3.
Question 3
Which of the following is NOT an attribute of Machine Learning?
1 / 1 point
Machine Learning defines behavioral rules by comparing large data sets to find common patterns
Takes data and rules as input and uses these inputs to develop an algorithm that will give us an answer
Takes data and answers as input and uses these inputs to create a set of rules that determine what the Machine
Learning model will be
Machine Learning models can be continuously trained
Correct
Machine Learning algorithms are trained with large sets of datasets to determine the relationships between
inputs and desired results to build the machine learning models.

4.
Question 4
Which of the following is NOT an attribute of Unsupervised Learning?
1 / 1 point
It is useful for clustering data, where data is grouped according to how similar it is to its neighbors and
dissimilar to everything else
The algorithm ingests unlabeled data, draws inferences, and finds patterns from unstructured data
It is useful for finding hidden patterns and or groupings in data and can be used to differentiate normal
behavior with outliers such as fraudulent activity
Takes data and rules as input and uses these inputs to develop an algorithm that will give us an answer
Correct
This statement is not an attribute of either Machine Learning or Unsupervised Learning. Machine Learning
techniques such as unsupervised learning are not fed rules. Rather they determine the rules from data.

5.
Question 5
Which of the following is an attribute of Supervised Learning?
1 / 1 point
Tries its best to maximize its rewards by trying different combinations of allowed actions within the provided
constraints
Relies on providing the machine learning algorithm unlabeled data and letting the machine infer qualities
Relies on providing the machine learning algorithm human-labeled data - the more samples you provide, the
more precise the algorithm becomes in classifying new data
Relies on providing the machine learning algorithm with a set of rules and constraints and letting it learn how
to achieve its goals
Correct
Supervised learning relies on giving the algorithm human-labeled data for training. The greater the number of
samples that the algorithm is trained on, the greater is its precision in classifying new data.

6.
Question 6
Which of the following statements about datasets used in Machine Learning is NOT true?
1 / 1 point
Training data is used to fine-tune algorithm’s parameters and evaluate how good the model is
Testing data is data the model has never seen before and is used to evaluate how good the model is
Validation data subset is used to validate results and fine-tune the algorithm's parameters
Training subset is the data used to train the algorithm
Correct
Training data is used to train the algorithm. It is the Validation data that is used to fine-tune algorithm’s
parameters and evaluate how good the model is.
7.
Question 7
When creating deep learning algorithms, developers configure the number of layers and the type of functions
that connect the outputs of each layer to the inputs of the next.
1 / 1 point
True
False
Correct
Deep Learning algorithms rely on several layers of processing units, or neurons, where each layer passes on its
output to the next layer, which processes it and passes it onto the next. The number of layers and the types of
functions that connect the outputs of each layer to the inputs of the next are configured by developers.

8.
Question 8
Which of the following fields of application for AI can be used at the airport to flag weapons within luggage
passing through the X-ray scanner?
1 / 1 point
Chatbots
Computer Vision
Natural Language
Speech
Correct
Computer Vision enables machines to interpret digital images and video sequences and perform tasks like
object identification.

9.
Question 9
Which of these activities is NOT required in order for a neural network to synthesize human voice?
1 / 1 point
Continue to correct the sample and run it through the classifier, repetitively, till an accurate voice sample is
created
Generate audio data and run it through the network to see if it validates it as belonging to the subject
Deconstruct sentences to decipher the context of use
Ingest numerous samples of a person’s voice until it can tell whether a new voice sample belongs to the same
person
Correct
Deconstructing sentences to decipher the context of use is a feature of Natural Language Processing, not
Speech Synthesis.

10.
Question 10
Which one of these ways is NOT how AI learns?
1 / 1 point
Supervised Learning
Reinforcement Learning
Unsupervised Learning
Proactive Learning
Correct
AI learns in three different ways - Supervised, Unsupervised, and Reinforcement Learning.
One fascinating application of AI is in healthcare, specifically IBM Watson's use for oncological
diagnostics and treatment planning. Watson uses advanced natural language processing (NLP) and
machine learning (ML) algorithms to analyze vast amounts of medical literature, patient records, and
clinical trials.
Here's how it works:
Data Ingestion: Watson ingests unstructured data from diverse sources such as medical journals, clinical
trial data, and patient health records. This includes both structured data (like lab results) and unstructured
data (like doctor's notes).
Natural Language Processing (NLP): Using NLP, Watson interprets and understands the context and
nuances of medical literature. It can discern the relationships between symptoms, diagnoses, treatments,
and outcomes.
Machine Learning (ML): Watson employs ML algorithms to continuously learn from new data. It
identifies patterns and correlations that may not be apparent to human clinicians, thereby improving its
diagnostic accuracy over time.
Cognitive Computing: Watson's cognitive computing capabilities allow it to mimic human thought
processes. It can generate hypotheses, evaluate evidence, and rank potential treatment options based on
confidence levels.