Unit Introduction
Welcome to the first unit of the AI Fundamentals course!
In this unit you will start immersing yourself in the world of artificial intelligence.
You will learn:
the different categories that compose AI and their applications
how AI works and the learning paradigms You can’t wait to get started, right?
Note that real humans worked on this course, it has not been generated by AI. Maybe it helped a bit with the phrasing of some sentences, but not much.
What is AI?
Artificial intelligence or AI is everywhere right now: from content creation to virtual
assistants. Let’s not forget personalized recommendations and automated
customer service. The list could go on for hours!
But what exactly is AI?
AI is a technology that enables computers to imitate human intelligence, allowing them to perform tasks like reasoning, learning, and problem-
solving, similar to how humans do.
Do not worry! Computers don’t have an actual brain or think like humans, they simply use math, statistics, and logic to process information and make decisions.
AI systems work by taking in a lot of information, learning from it, and then using what they learned to guess what might happen in the future.
Believe it or not, AI isn’t new! It’s been around for a long time and has evolved significantly, especially in recent years.
Let’s take a look at the history of AI.
History of AI
1950s
Turing test
Alan Turing laid the foundation for AI by proposing that machines could mimic human intelligence and introduced the Turing test to see if a machine’s behavior could be mistaken for a human’s.
Fun fact: This test was finally passed by a machine in 2013!
1956
Field of study
The term artificial intelligence was coined by John McCarthy, marking the official birth of AI as a field of study. 1980s
Machine Learning
The introduction of machine learning allows computers to learn from data and improve over time. 1997
Deep Blue
IBM’s Deep Blue, a chess playing computer, beats the world chess champion Garry Kasparov.
2010s
Deep Learning
This decade saw major advances in deep learning, where AI uses multiple layers to recognize complex patterns, for example enabling self-driving cars to understand their surroundings. 2017
Transformer model
The Transformer model is introduced, changing how computers understand and work with human language by capturing the relationships between distant words. 2020s
GPTs
OpenAI introduced the Generative Pre-trained Transformer (GPT) series, enhancing how computers understand and generate human language.
Categories of AI
AI has six main categories, each focusing on specific tasks and techniques. They can work both separately and together to create more advanced systems.
These categories are:
1 Machine Learning
2 Natural Language Processing
3 Expert Systems
4 Computer Vision
5 Speech Recognition
6 Robotics Imagine AI as a school called ‘AI will always love you’ with different classes, each
dedicated to a unique skill.
Let’s dive deeper into each category.
Machine Learning (ML)
Machine Learning (ML) is a part of AI that helps computers learn from data, find
patterns, and improve over time without needing humans to tell them what to do.
By looking at lots of existing information, ML algorithms can spot trends, make
choices, and improve their skills based on new examples.
It is like the math class where students learn to solve problems by practicing with examples.
NLP is focused on understanding and working with human language, while generative AI (used by ChatGPT for example) specifically creates new content, like text or images. Click + for more information.
Real world application:
Recommendations on what to watch on Netflix.
Natural Language Processing (NLP) Natural Language Processing (NLP) is a part of AI that helps computers understand,
interpret, and create human language. It lets machines process and analyze text or
speech like people do, making it easier for them to work with language.
NLP is the language class, where students learn how to read, write, and talk.
Click + for more information.
Real world application:
Autocorrect and autocomplete text.
Expert Systems An Expert System is a computer program that mimics how a human expert makes decisions in a specific area. It uses set rules to analyze information and draw
conclusions, applying knowledge from that field to help solve problems and
provide advice.
It is similar to a science class where students apply specific rules and knowledge from their field to solve problems.
Click + for more information.
Real world application:
Diagnosis of bacterial infections. 4: Computer Vision
Computer Vision
Computer Vision is a field of artificial intelligence that enables computers to
interpret and understand visual information from the world, such as images and
videos, allowing them to perform tasks like object recognition, image classification, and scene understanding. It is like the art class that helps students learn to interpret images and recognize patterns in visuals.
Click + for more information.
Real world application:
Facial recognition technology.
Speech Recognition
Speech Recognition is a technology that focuses on converting spoken language
into written text, allowing machines to understand and transcribe audio input,
such as voice commands and conversations. This technology facilitates human-
computer interaction by enabling users to communicate with machines through
their voices.
It is like the music class where students learn to interpret sounds.
Click + for more information. Real world application:
Virtual assistants.
Robotics Robotics is a part of AI that builds machines to interact with the real world, using
other AI tools like machine learning and computer vision to help them work and make decisions on their own.
It is like the PE class, where students learn through action and physical tasks.
Click + for more information.
Real world application:
Autonomous vehicles such as the Mars Rover.
Machine learning vs Deep learning
Machine learning Deep learning
Machine learning is the most widely used Deep learning is a part of machine learning category of AI because it forms the foundation that uses layered networks (similar to neurons for many AI applications. You will focus on this in the brain) to learn from large amounts of over the next few lessons.
raw data, which helps AI function like the human brain.
Deep learning is like an advanced math class where students dive deep into complex subjects, using layered understanding to master intricate topics and spot patterns.
Learning paradigms
Machine learning involves training an algorithm to learn patterns from data by providing examples and making predictions on new data. 🔑 Key Concept An algorithm is a set of instructions or rules. There are two different phases:
Training phase Prediction phase
During training, the algorithm Once trained, the
analyzes the data to identify mathematical model is used
patterns. From these to make predictions on new
patterns it develops a data. It applies the patterns
mathematical model that learned during training to represents the relationship interpret or predict outcomes
between this data. from new information.
🔑 Key Concept A mathematical model is a way to represent real-world things or problems using numbers and formulas to make predictions or understand how things work.
Types of learning paradigms
Machine learning includes three main types of learning, known as learning paradigms. Each paradigm offers a different method for how machines learn from data in terms of input provided and possible outputs. It is similar to three teachers who each have their own unique
teaching style.
They are:
1 supervised learning
2 unsupervised learning
3 reinforcement learning
Let’s have a look at each one in more detail. 1: Supervised learning
Supervised learning
In supervised learning paradigms the inputs are labeled data.
Think of them as examples with the correct answers. The model learns by analyzing these labeled examples to understand the relationships between inputs and outputs, allowing it to make accurate predictions on new, unseen data. For example, if you want your machine learning model to recognize different
kinds of fruit, you would provide labeled examples of apples, pears, and oranges. The model uses these examples to learn the differences and identify patterns
associated with each fruit, e.g. shape, color, texture, size. When a new image is given to the algorithm, it can predict which of the three fruits it is. It’s like Sarah the Teacher giving her students examples to help them understand and learn to classify different concepts.
Here are some examples in which supervised learning
is used.
Spam detection Image recognition Credit scoring Classifying emails as Identifying objects in Predicting whether a loan “spam” or “not spam” based images, such as recognizing applicant is expected to fail on labeled examples of cats and dogs, using to repay the loan based on previous emails. labeled images for training. historical data of previous applicants with known outcomes. Medical diagnosis Sentiment analysis Assisting doctors by Determining whether a predicting diseases based piece of text (like a review) on patient data and is positive, negative, or previous diagnoses. neutral using labeled examples of texts.
Unsupervised learning
In unsupervised learning paradigms the inputs are not labeled, the model doesn’t know any information about them. Instead, it finds patterns by grouping similar items on its own. This approach is useful when you have data but don’t know what you’re looking for, allowing the model to discover hidden connections or relationships within the information. For example, if you give the model various images of fruits, it will find relationships and sort them into groups based on the fruit characteristics. When
you show it a new image, it uses what it learned to place the fruit into one of the groups.
It’s like Ulysses the Teacher providing his students with a variety of materials and asking them to explore and find patterns on their own, encouraging them to group similar items without any guidance.
Here are some examples in which unsupervised
learning is used.
Customer segmentation Market basket analysis Anomaly detection Grouping customers based Identifying items frequently Detecting unusual patterns on purchasing behavior and purchased together to in data, such as fraud preferences without optimize product detection in financial predefined labels. placements and transactions. promotions.
Document clustering Image compression Organizing a collection of Reducing the size of images documents into topics or by grouping similar pixel themes without prior values to maintain quality categorization. while saving space.
3 : Re i n fo rcementlearning
Reinforcement learning
In reinforcement learning paradigms, the algorithm (called agent in this case) learns to make decisions by interacting with an environment and getting feedback as rewards or penalties. This method uses trial and error to get better over time as the agent tries
different options to find the best outcomes. This is especially helpful in complicated situations where the best choices aren’t clear.
Imagine a robot learning to navigate a maze. It receives a reward for reaching the exit and a penalty for hitting walls. Over time, the robot figures out the best path to take by trying different routes and learning from its mistakes.
It’s like Ralph the Teacher guiding his students to try different activities and learn from their successes and mistakes, helping them discover the best strategies for achieving their goals.
Here are some examples in which reinforcement learning is used. Game playing Robotics Autonomous vehicles Training AI agents to play Teaching robots to perform Training self-driving cars to video games, where the tasks, such as walking or navigate by rewarding them agent learns strategies by picking up objects, by for safe driving decisions playing against itself and rewarding them for and penalizing risky receiving rewards for successful actions and behaviors. winning. guiding them through trial and error.
Personalized Healthcare treatment recommendations plans Improving recommendation Developing treatment systems for movies or strategies for patients by products by adjusting using past treatment suggestions based on user outcomes to reward interactions and feedback. effective actions and improve future decisions.
Continuetothewrapupforthisunit
Wrap up
Artificial intelligence (AI) is a technology that gives computers the ability to imitate human intelligence, allowing them to perform tasks such as reasoning, learning, problem-solving solving and decision-making.
AI has several categories, each with specialized techniques for handling specific tasks. These include machine learning (teaching computers to learn from data), natural language processing (working with human language), computer vision (interpreting images and video), robotics (controlling machines), and expert systems (rule-based reasoning).
AI works in two steps: training and prediction. In the training step, the system learns from examples and creates rules that describe the relationships in the data. In the prediction step, it uses those rules to give answers or make predictions on new information.
Unit complete!
Well done!
You have completed the first unit! In the next unit you will learn more about Generative AI.
Mark this task complete to continue to the next unit.