CHAPTER 1
Artificial Intelligence

Artificial intelligence (AI) is changing the world as we speak and will continue to do so over the next 10 years and into the future. It will affect our kids and their families, and it will certainly affect us.

"Artificial intelligence would be the ultimate version of Google. The ultimate search engine that would understand everything on the Web. It would understand exactly what you wanted, and it would give you the right thing. We're nowhere near doing that now. However, we can get incrementally closer to that, and that is basically what we work on."

-Larry Page

"The pace of progress in artificial intelligence (I'm not referring to narrow AI) is incredibly fast. Unless you have direct exposure to groups like DeepMind, you have no idea how fast it is growing at a pace close to exponential."

-Elon Musk

What AI Is-And What It Is Not

Artificial intelligence will help assist workers rather than replace them altogether. For years we have heard that humans will one day be replaced by computers; while that is true for some tasks, artificial intelligence will aid many of us with our mundane day-to-day tasks and allow us to focus on higher-value activities.

Tools like OpenAI's ChatGPT will help make access to information quicker and easier by allowing us to ask it to explain things in a way that we can understand without necessarily having to do a tremendous amount of research. For example, a programmer trying to write a line of code could ask ChatGPT to write it by describing what they are trying to accomplish. Still, the output will require the programmer to validate and optimize the code for their exact situation and improve the performance of the code to ensure it is flawless.

Another good example is the use of robot vacuums; many of us have embraced and use them daily. While these vacuums do a great job in general, there are hard-to-reach places in our homes that they can only reach with a human intervening to either move the obstruction or perform the vacuuming themself.

Forms of AI

Artificial intelligence takes many forms in the world. Each form has its place, and that's what makes artificial intelligence truly intelligent. In the upcoming sections, the following types of AI will be reviewed: machine learning, machine teaching, reinforcement learning, computer vision, natural language processing, deep learning, and robotics.

Machine Learning

Machine learning is a type of artificial intelligence that attempts to develop statistical models and algorithms that help computers make decisions or predictions by learning from data to perform a task. The following image is an example of a machine learning model leveraging the Microsoft Azure platform.

The two main types of machine learning algorithms are supervised learning algorithms and unsupervised learning algorithms:

• Supervised Learning: A type of machine learning in which the algorithm's dataset is labeled and trained and the correct output is provided for each input. Supervised learning attempts to teach a model that predicts net-new data production. Here are some examples of supervised learning algorithms:
  • Linear Regression: Used for regression problems that aim to predict a continuous value. The algorithm finds the line of best fit that lowers the difference between the predicted and actual values.

    Linear regression is a way of finding a relationship between two things, like how the amount of rain affects the amount of water in a bucket. We can use this relationship to make predictions, like how much water will be in the bucket if it rains a certain amount.

    Imagine we have a ball that can go faster or slower depending on how much we push it. Using a scale to measure the force, we can measure how fast the ball goes utilizing a stopwatch.

    The harder we push the ball, the faster it goes. We can use linear regression to create a simple equation that helps us predict how fast the ball will go based on how hard we push it. The equation might look something like this:

    Source: Microsoft / https://learn.microsoft.com/en-us/azure/architecture/example-scenario/ai/many-models-machine-learning-azure-machine-learning/?last accessed March 28, 2023

    This means that if we push the ball with a force of 8, we can predict that the ball will go at a speed of 19 (2 × 8 + 3). If we push the ball with a force of 12, we can predict that the ball will go at a speed of 27 (2 × 12 + 3).

  • Logistic Regression: Used for binary classification issues. The goal is to predict one of two possible outcomes. The algorithm finds the best boundary that divides the data into separate classes.

    Logistic regression is a way of predicting whether something will happen. It is like guessing if it is going to rain tomorrow. We can use logistic regression to make predictions based on patterns in the data we collect.

    For example, say we have a bag of candy; some are blue, and some are red. Could we predict whether we will pick a blue or red candy from the bag?

    We can use logistic regression to create a simple equation that helps us make this prediction based on the features of the candy, such as its size or weight.

    Using logistic regression, we can predict whether something will happen based on the patterns we observe in the data. This is a valuable tool for understanding the world and making informed decisions.

  • Decision Trees: Used for classification and regression problems. The algorithm builds a treelike structure by dividing the data into smaller and smaller groups based on the values of the features.

    Decision trees are a way of making decisions by following a series of steps, like a flowchart. So based on the information we have, we can use decision trees to choose for us.

    Imagine we want to decide what to do based on the weather outside. We can create a decision tree with two branches, one for nice days and one for stormy days. Here is an example of a decision tree.

    This decision tree tells us that if it is nice outside, we should play tennis, but if it is raining, we should play video games instead.

    We can make the decision tree more complex by adding more branches and choices, such as it is too cold to be outside if the temperature is below 45 degrees. This helps us make better decisions by considering all the factors that might be important.

    Using decision trees, we can make choices based on our information and follow a logical process to make the best decision. This is a valuable tool for making decisions in everyday life, like what to eat for breakfast or what game to play with friends.

  • Neural Networks: Used for many problems, such as natural language processing, image classification, and speech recognition. Neural networks consist of levels of interconnected neurons that process the input data and generate the output.

    Neural networks are like magic boxes that can learn to do things alone. They comprise many small parts that work together to solve problems, similar to how our brain works.

    Suppose we wanted to teach a computer to recognize different birds. We can use a neural network to do this. We can show a neural network pictures of birds like eagles, hawks, and sparrows. The neural network will examine the images and determine what makes each animal different.

    As the neural network learns, it will start recognizing picture patterns. For example, it might know that eagles have large wingspans and large talons. It will use these patterns to make predictions about new pictures it sees.

    Once the neural network has learned enough, we can give it a new picture and ask it to tell us what bird is in the image. It will use what it learned to predict the bird in the picture, whether it's an eagle or a hawk.

    We can use neural networks to teach computers to learn and make decisions independently. This is a powerful tool that can be used to solve many problems, like recognizing animals, predicting the weather, or playing games.

• Unsupervised Learning: A machine learning algorithm trained on an unlabeled dataset without the correct output. Unsupervised learning attempts to locate patterns and structures in the data it has not yet learned. Here are some examples of unsupervised learning algorithms:
  • K-Means Clustering: Used for clustering problems to separate the data into a designated number of groups that are called clusters. The algorithm finds the centers of each cluster and can assign each data point to the closest center.

    K-means clustering is a way of organizing things into groups based on their similarities. We can use K-means clustering to group together things that are similar.

    For example, say we have a bunch of nails, some long and some short. We can use K-means clustering to group the nails based on their appearance.

    Let's start by grouping the nails based on their color. We could put all the black nails in one group, all the white nails in another, and so on. Then, we could group the nails within each color group based on their shape, like all the nails that are long in one group, all the nails that are short in another group, and so on.

    After we have sorted the nails into groups, we could see that all the black and long nails were together in one group and all the white and long nails were together in another...