No-Code Artificial Intelligence: The new way to build AI powered applications (English Edition)

By Ambuj Agrawal

“No-Code Artificial Intelligence” is a book that enables you to develop AI applications without any programming knowledge. Authored by the founder of AICromo (https://aicromo.com/), this book takes you through an array of examples that shows how to build AI solutions using No-code AI tools.

The book starts by sharing insights on the evolution of No-code AI and the different types of No-code AI tools and platforms available in the market. The book then helps you start building applications of Machine Learning in Finance, Healthcare, Sales, and Cybersecurity. It will also teach you to create AI applications to perform sales forecasting, find fraudulent claims, and detect diseases in plants. Furthermore, the book will show how to build Machine Learning models for a variety of use cases in image recognition, video object recognition, and data prediction.

After reading this book, you will be able to build AI applications with ease.

• Language: English
• Publisher: BPB Online LLP
• Release date: Mar 7, 2023
• ISBN: 9789355513465

Book preview

CHAPTER 1

What is AI?

Introduction

In this chapter we will go through the basics of Artificial Intelligence (AI) and look at different types of AI systems. We will also discuss the basics of Machine Learning and understand different types of Machine Learning techniques such as Supervised Learning, Unsupervised Learning and Reinforcement Learning (RL). Furthermore, we would look at some of the latest AI research and its impact on our day to day lives.

Structure

In this chapter, we will cover the following topics:

Introduction to AI

Types of AI Systems

Introduction to Machine Learning

Supervised Learning

Unsupervised Learning

Reinforcement Learning

Bleeding Edge AI Applications

Objectives

After studying this chapter, you will be able to decide and implement relevant AI techniques on your projects. You will also learn about Machine Learning techniques that can be used for a variety of use cases and look at some of the latest advancements in the AI space.

Introduction to Artificial Intelligence

AI refers to the capability of machines to take actions and make decisions without being explicitly programmed. It can also refer to the capability of machines that display features associated with the human brain such as problem solving and learning.

In simple terms, AI is intelligence exhibited by machines. The ideal characteristic of AI is its ability to rationalize and take actions that have the best chance of achieving a specific goal. AI is a very broad field with many subareas and it involves the study of automated recognition, understanding of signals, reasoning, planning, decision-making, learning and adaptation.

Many of the devices you use in your day to day life use some form of AI within them. From your refrigerator, smartphone, television, car, and so on use some form of AI to make decisions for you. For example, your refrigerator uses the temperature sensors to determine the cooling rate. Your smartphone uses AI to perform fingerprint recognition, battery saving and optimizing pictures. Your car uses AI to improve fuel efficiency, safety features and temperature control.

In the next section, we will look at 4 different types of AI systems such as Reactive Machines, Limited Memory, Theory of Mind and Self-aware AI. We will also look at an alternate system for classification of AI systems which divides it into 3 different types such as Artificial Narrow Intelligence (ANI), Artificial General Intelligence (AGI), and Artificial Superintelligence (ASI).

Types of AI systems

In this section we will discuss different types of AI systems such as Reactive Machines, Limited Memory, Theory of Mind, Self-aware AI, ANI, AGI and ASI.

There are 4 main types of AI systems such as Reactive Machines, Limited Memory, Theory of Mind and Self-aware AI, which all are explained as following:

Reactive Machines: Reactive Machines is the oldest form of AI system with extremely limited capacity. These AI systems react to current scenarios but do not have an ability to learn from past actions. Thus, they can only be used to solve problems with a limited set or combination of inputs. Reactive Machines are used in spam filters, Netflix’s recommendation engine and chess-playing computer. IBM’s Deep Blue is a popular example of a reactive AI machine that beat chess Grandmaster Garry Kasparov in 1997.

Limited Memory: Limited Memory machines have the capability of learning from historical data to make decisions in addition to the capabilities of reactive machines. Majority of the applications of AI that we use today use Limited Memory machines. For example, for image recognition an AI system is trained using thousands of pictures and their labels. This AIs system stores past data in the form of training images to learn to recognize images with high accuracy. Nearly all present day AI applications such as chatbots, virtual assistants, language translation systems and self-driving vehicles use limited memory AI systems.

Theory of Mind: Theory of Mind system is an AI system which is in the research phase and it is not fully developed for practical applications. Theory of Mind is an active area of AI research system to allow machines to understand mental states of self and others and to communicate things that are not concrete like needs, ideas or concepts. It involves the ability of machines to teach, build shared goals, intentional communication, and so on. With Theory of Mind, the AI machines will have capability to combine advanced pattern recognition and model-based reasoning, develop intelligence with common sense and learn the best way for optimal learning.

Self-aware: Self-aware system is an AI system which only exists hypothetically at the present moment. Self-aware AI is an AI akin to the human brain with developed self-awareness. It is the AI system that we see in sci-fi movies like Terminator, Space Odyssey, Matrix, and the like. Creating this AI system is a bit far away in the future, probably centuries away from materializing and is the ultimate objective of most of the AI research. This type of AI will have ions, needs, beliefs and desires of its own and will be able to understand and evoke emotions in those it interacts with.

There is an alternate system of classification for AI systems which is generally used by technology companies and they divide the AI system into 3 types such as Artificial Narrow Intelligence (ANI), Artificial General Intelligence (AGI), and Artificial Superintelligence (ASI).

Artificial Narrow Intelligence (ANI): ANI is the type of AI system that represents all the existing AI, including the latest and most capable AI systems created to date. Artificial Narrow Intelligence refers to AI systems that are developed for particular tasks and can perform those specific tasks autonomously. These AI systems can do nothing more than the specific tasks they are programmed to do and hence they have a very limited or narrow range of competencies. Latest AI technologies in natural language generation, speech recognition, decision management, deep learning, and so on, allfalls under ANI.

Artificial General Intelligence (AGI): AGI involves AI systems that can learn, perceive, understand and function completely like a human being. They can build multiple competencies for different ranges of tasks and have the ability to form generalizations across different domains. With AGI the AI system would be just as capable as humans and they will replicate our thinking and innovation capabilities.

Artificial Superintelligence (ASI): ASI is the pinnacle of AI research with it becoming the most capable form of intelligence on earth. ASI would be exceedingly better than the intelligence of human beings because of faster data processing, greater memory and decision-making capabilities. AGI and ASI could provide us with great benefits by improving human society but at the same time they can threaten our existence.

In this section we have looked at different types of AI and a variety of use cases associated with these types of AI systems. In the next section we will look at a subset of artificial intelligence called Machine Learning (ML) that refers to AI programs that can automatically learn from and adapt to new data without being assisted by humans.

Introduction to Machine Learning

A subset of Artificial Intelligence is ML that provides computers with the ability to learn without being explicitly programmed. It allows the computer programs to automatically learn from and adapt to new data without being assisted by humans. Majority of the real world AI applications we see today use some form of machine learning algorithms to achieve desired outcomes.

A typical computer application is programmed and takes data as an input. It then performs some computation on the Data according to the Program and produces the desired output as shown in Figure 1.1.

Figure 1.1: Typical Computer Application

A machine learning system takes the Data and desired output as inputs to generate the required program as the Output as shown in Figure 1.2. The machine learning generated Program (the output) can then be used in a typical application to predict or perform some computation on the data to produce the desired output. Please refer to the following figure:

Figure 1.2: Machine Learning based computer application

ML is a vast field with a variety of techniques and a large variety of use cases. It can further be subdivided into three main categories which are:

Supervised learning: Supervised learning involves labeled data and can be used for classification (grouping similar instances) and regression (learning what normally happens to draw inferences from datasets ). For example, learning to classify emails as spam and not spam based on training data of spam and not spam email messages.

Unsupervised learning: Unsupervised learning involves unlabeled data and can be used to discover patterns in the dataset. For example, learning the patterns in the English language from the Wikipedia pages.

Reinforcement Learning (RL): RL involves learning from experimentation based on rewards and feedback loop and is generally used to train agents in simulated environments. For example, teaching a bot to play a computer game and maximize the score and chances of winning will use a RL algorithm.

In this section we have looked at the basics of ML and the different types of ML systems. In the next section we will look at Supervised Learning (type of ML) in more detail with a few examples of the use cases of Supervised Learning techniques to learn from labeled data.

Supervised learning

Supervised learning is a machine learning technique that takes in labeled data as an input and learns the pattern in the data to either perform classification by grouping similar instances or perform regression which involves predicting outcomes accurately based on a given input. The input data is fed in the supervised learning models and the model adjusts its weight until it is able to correctly predict the desired outcomes using the cross validation process.

The training dataset for supervised learning models uses inputs with correct outputs to learn over time. It measures accuracy through a loss function which is adjusted until the error has been sufficiently minimized.

Supervised learning can be either used to perform either classification or regression:

Classification: Classification is used to classify data into specific categories. This algorithm recognizes specific types in the dataset and attempts to draw conclusions on how these types should be defined. Common classification algorithms are linear classifiers, Support Vector Machines (SVM), decision trees, k-nearest neighbors and random forest.

Regression: Regression is used to predict relationships between variables. It is commonly used to make future projections such as housing prices, future growth and sales revenues. Common regression algorithms are Linear regression, logistic regression and polynomial regression.

Supervised machine learning involves various algorithms and computation techniques. Following are the explanations of some of the commonly used Supervised learning methods:

Naive Bayes: Naive Bayes is a classification algorithm that uses the principle of class conditional independence from the Bayes Theorem. This means that the presence of one feature is independent from the presence of another feature and each feature has an equal effect on the result. There are three main types of Naive Bayes classifiers algorithms such as Multinomial Naive Bayes, Bernoulli Naive Bayes, and Gaussian Naive Bayes. Naive Bayes are primarily used for spam identification, text classification and recommendation systems.

Linear regression: Linear regression is a regression algorithm that is used to predict relationships between variables. It is used to identify the relationship between a dependent variable and one or more independent variables and is used to make predictions about future outcomes. Simple linear regression involves only one independent variable and one dependent variable. Multiple linear regression involves multiple independent variables. Linear regression seeks to plot a line of best fit (straight line when plotted on a graph) which is calculated through the method of least squares.

The Figure 1.3 shows an example of linear regression where we want to predict the relationships between price and mileage from the input dataset.:

Figure 1.3: Linear regression algorithm for finding relationship between variables

Logistic regression: Logistic regression is a regression algorithm that is used when the dependent variable has binary outputs, such as true and false or yes and no. Logistic regression is similar to linear regression but is mainly used to solve binary classification problems such as spam identification.

Support vector machine (SVM): Support vector machine is a supervised learning algorithm that can be used for both data classification and regression tasks. It is generally used for classification problems where the algorithm involves constructing a decision boundary (or hyperplane) where the distance between two classes of data points is at its maximum. This decision boundary separates the classes of data points (for example, spam vs. non-spam) on either side of the plane.

The Figure 1.4 shows an example of a hyperplane constructed by the support vector machine algorithm to classify the given dataset:

Figure 1.4: Hyperplane constructed by the support vector machine to classify data

K-nearest neighbor (KNN): K-nearest neighbor which is also known as the KNN algorithm is a classification algorithm that classifies data points based on their proximity and association to other data points. KNN algorithm calculates the distance between data points (usually through Euclidean distance) and uses this distance to assign a category based on either the most frequent category or the average in that segment. KNN is typically used for recommendation engines and image

Reviews for No-Code Artificial Intelligence

[samiraagrawal7 · Rating: 5 out of 5 stars]

Amazing book for getting started with understanding No Code AI.