Inside AI

OVER 150 BILLION PURCHASES PER YEAR USE THIS AUTHOR’S AI

AKLI ADJAOUTE FOREWORD BY RAYMOND KENDALL

To my beloved wife, Nathalie, and our wonderful children, Ghislene and Eddy, as well as to my parents, brothers, and sisters. I want to express my profound gratitude for your constant encouragement and belief in my aspirations.

Brief Contents

• The rise of machine intelligence
• AI mastery: Essential techniques, Part 1
• AI mastery: Essential techniques, Part 2
• Smart agent technology
• Generative AI and large language models
• Human vs. machine
• AI doesn’t turn data into intelligence
• AI doesn’t threaten our jobs
• Technological singularity is absurd
• Learning from successful and failed applications of AI
• Next-generation AI
• Tracing the roots: From mechanical calculators to digital dreams
• Algorithms and programming languages
• Index

Contents

• Foreword
• Preface
• Acknowledgments
• About the book

Preface

Welcome to the world of artificial intelligence (AI), a domain where the boundaries between science fiction and reality often become indistinct. AI has captivated our collective imagination, particularly in 2022 and 2023, thanks to the release of ChatGPT. This groundbreaking product has played a pivotal role in democratizing AI usage by offering a user-friendly interface, empowering individuals without technical expertise to harness its benefits. ChatGPT boasts impressive capabilities, including answering questions, crafting narratives, composing music and poetry, and generating computer code.

For more than three decades, I’ve been passionate about artificial intelligence, dedicating my adult life to teaching and applying AI to address real-world challenges. In 1987, I established my first company, Conception en Intelligence Artificielle, in Paris before completing my PhD. We developed the MINDsuite platform, which seamlessly combines various AI techniques and has found successful applications in defense, insurance, finance, healthcare, and network performance. While leading this company, I also shared my expertise with students at the École Pour l’Informatique et les Techniques Avancées (EPITA), where I served as the head of the AI department and chaired the scientific committee.

In April 2000, I launched my second venture, Brighterion (acquired by Mastercard), in San Francisco. This company was founded to address the pervasive issues of payment fraud and cybersecurity, which pose significant challenges across various industries, leading to annual losses amounting to billions of dollars. Brighterion-powered software is now used by over 2,000 clients worldwide, with 74 of the largest U.S. banks relying on its technology to safeguard against fraud and risk. Annually, more than 150 billion transactions are processed through Brighterion software.

In this book, we embark on a transformative journey to educate readers about the fascinating world of AI. Whether you’re new to the field or a seasoned enthusiast, my aim is to equip you with a clear and comprehensive understanding of what AI truly is and what it can and cannot achieve. Throughout this exploration, we will discuss the expansive and multifaceted landscape of AI, marked by a diverse range of techniques and methodologies aimed at simulating human cognition.

Our journey will take us to the very heart of AI, where we’ll dissect these techniques and methodologies. From the early days of expert systems to the cutting-edge advancements in deep learning algorithms, you’ll gain a thorough comprehension of the full spectrum of AI techniques that drive AI applications. Along the way, we’ll also explore various aspects of human cognition, including imagination, intuition, curiosity, common sense, and creativity, to illustrate that current AI techniques still fall short of replicating these qualities.

Insights from both successful and unsuccessful AI projects will demonstrate that many human jobs remain beyond the capabilities of AI and refute the notion of technological singularity, which envisions a future where intelligent robots can replicate themselves, potentially leading to the end of human civilization. As we progress, we’ll also address ethical questions surrounding bias, fairness, privacy, and accountability. Drawing from my three decades of experience in developing and deploying mission-critical AI systems, I will outline the characteristics that, in my perspective, will define the next generation of AI platforms.

I firmly believe that it is crucial for every citizen to acquire knowledge about AI, given its pervasive effect on our modern world. Whether you are an aspiring AI developer, a business professional, an investor, a policymaker, or simply a concerned citizen, I welcome you to embark on this journey to discover the true essence of AI and its profound effect on our world. My hope is that, by the time you turn the final page of this book, you will not only possess the ability to discern AI reality from its illusions but also have the capacity to engage thoughtfully with the imminent AI-driven future that awaits us all.

Let the voyage begin.

Acknowledgments

Numerous individuals generously dedicated their time to reviewing this book, and I am sincerely grateful for the valuable comments and suggestions received. I extend my gratitude to Raymond Kendall, a dear friend who insisted I write this book. Lucien Bourely, a friend and former partner in my company, deserves special mention for his consistently valuable insights and encouragement.

My deepest appreciation goes to the great team of technical editors I was fortunate to have, namely Patrick Perez, Raymond Pettit, James T. Deiotte, Dick Sini, Shawn Nevalainen, François Stehlin, Florent Gastoud, Philippe Hallouin, and Philippe Perez. These exceptional experts invested their time and expertise in thoroughly reviewing my work. Through numerous discussions, their insights, suggestions, and meticulous inputs significantly contributed to refining the manuscript. Each editor brought a unique perspective and a wealth of knowledge to the table, enhancing the overall quality of the content. Their dedication to precision, helpful critiques, and collaborative approach played a pivotal role in shaping the content of the book. Throughout the different iterations leading to the final edition, their feedback acted as a guiding force, ensuring the narrative remained engaging, informative, and accessible to readers from diverse backgrounds.

The content of this book is profoundly influenced by my journey in applying AI to mission-critical applications. We encountered numerous challenges that required refining our AI algorithms, finding an efficient way to design models, and creating a storage technique suitable for storing intelligence while providing real-time responses in milliseconds to adhere to our stringent service level agreements. These agreements demanded scalability, resilience, adaptability, explicability, and compliance. I express my heartfelt gratitude once again to François Stehlin, Florent Gastoud, and Philippe Hallouin; an extraordinarily talented team that not only believed in my venture but also stood steadfast with me through every hurdle we encountered. Their intelligence and unwavering support were instrumental in turning challenges into triumphs, and for that, I am sincerely thankful.

Special thanks to Richard Vaughan, a CTO at Purple Monkey Collective, a research focused startup delivering machine learning and cloud guidance services. Richard is a highly experienced engineer who has worked across many different industry verticals and countries in a highly varied career, and worked as my technical editor on this book.

A special acknowledgment is reserved for Daniel Zingaro, whose compelling arguments and persuasive influence were crucial in deciding to incorporate a dedicated chapter on generative AI. This addition holds particular significance given the current prominence and extensive discussions surrounding generative AI in the broader field of artificial intelligence.

Finally, a big thank you to all the reviewers who provided feedback: To Alain Couniot, Alfons Muñoz, Andre Weiner, Andres Damian Sacco, Arnaldo Gabriel Ayala Meyer, Arturo Geigel, PhD, Arun Saha, Bill LeBorgne, Bonnie Malec, Clifford Thurber, Conor Redmond, Dinesh Ghanta, Georgerobert Freeman, James Black, Jamie Shaffer, Jeelani Shaik, Jereme Allen, Jesús Juárez, Kay Engelhardt, Lucian-Paul Torje, Marc-Anthony Taylor, Mario Solomou, Milorad Imbra, Mirna Huhoja-Dóczy, Piotr Pindel, Ranjit Sahai, Rolando Madonna, Salil Athalye, Satej Sahu, Shawn Bolan, Shivakumar Swaminathan, Simon Verhoeven, Stephanie Chloupek, Steve Grey-Wilson, Tandeep Minhas, and Tomislav Kotnik, your insight and suggestions helped make this book what it is.

About the book

In this book, the primary goal is to provide a comprehensive understanding of both the capabilities and limitations of artificial intelligence. We’ll explore a diverse range of AI techniques, spanning from expert systems to deep learning, and emphasize the distinctions between AI and human cognition. Insights drawn from real-world AI projects not only question the notion of machines taking over the majority of human jobs but also underscore the implausibility of the technological singularity concept. Ethical considerations, including issues like bias and privacy, will be addressed. Drawing on three decades of experience in applying AI to mission-critical applications, I outline the characteristics that define the next generation of AI platforms.

Who should read this book?

This book is a comprehensive guide for anyone interested in learning about artificial intelligence, an ever-evolving field that profoundly shapes our future, influencing how we learn, work, and live.

How this book is organized

Embark on an extensive exploration of the field of artificial intelligence within the 11 chapters of this insightful book. The journey begins with an introduction to fundamental principles, encompassing algorithms and programming languages, laying a solid foundation for understanding AI. Moving beyond, chapters 2 to 4 explore various AI techniques, covering expert systems, business rules, fuzzy logic, genetic algorithms, case-based reasoning, classical neural networks, deep learning, Bayesian networks, unsupervised learning, and smart agents. Chapters 5 and 6 shift focus to the advancements in generative AI and the comparison between human cognition and artificial intelligence. Subsequent chapters tackle diverse topics, including the limitations of AI, its impact on human jobs, and a critical examination of technological singularity. The book concludes with valuable insights from past AI projects, providing guidance for future endeavors and a visionary perspective on the next generation of AI platforms. Additionally, an insightful appendix complements the narrative by exploring the historical evolution of AI technology. Each chapter offers a unique lens into the multifaceted landscape of AI, making this book an essential read for both enthusiasts and those seeking a deeper understanding of this transformative field:

• Chapter 1—In the introductory chapter, we explore a range of real-world examples to showcase how AI is emerging as a pivotal force that propels positive transformations across diverse fields by enhancing efficiency and fostering innovation. Additionally, we also highlight the challenges that stem from the inherent inclination of AI algorithms and models towards errors.
• Chapter 2—In this chapter, we provide an overview of multiple AI techniques, accompanied by practical examples. We will explain expert systems, which rely on human expertise and inference procedures to solve problems, as well as case-based reasoning, a method that uses past experiences to tackle new challenges. Additionally, we will explore fuzzy logic as an elegant means of representing and capturing the approximate and imprecise nature of the real world. Finally, we’ll conclude this chapter with an examination of genetic algorithms, which offer a powerful, straightforward, and efficient approach to solving nonlinear optimization problems.
• Chapter 3—In this chapter, we will continue to explore various AI techniques. We’ll begin with data mining, a powerful AI technique used to extract valuable information, patterns, and associations from data. Following that, we’ll introduce artificial neural networks and deep learning, powerful algorithms for pattern recognition that have yielded impressive results in computer vision, natural language processing, and audio analysis. Next, we’ll briefly touch on Bayesian networks, a technique that encodes probabilistic relationships among variables of interest. To wrap up the chapter, we’ll explore unsupervised learning, a collection of algorithms designed to analyze unlabeled datasets and uncover similarities and differences within them.
• Chapter 4—In this chapter, we will introduce smart agents, a powerful artificial intelligence technique centered on the use of adaptive, autonomous, and goaloriented entities to address complex problems. We will specifically focus on a proprietary smart agent approach, providing an illustrative example to elucidate how each agent possesses the capability to assess inputs as either beneficial or detrimental with respect to its objectives. Furthermore, we will explore the adaptability of these agents, draw comparisons with more conventional approaches, and examine instances where this technique has been effectively employed to solve real-world challenges.

NOTE Chapters 2 to 4 contain a high-level explanation of some of the technical underpinnings of AI and can be skipped by those who want to dive into the discussion of the reality and illusion of current AI.

• Chapter 5—AI has witnessed numerous ups and downs, but the emergence of ChatGPT, OpenAI’s impressive chatbot, capable of composing poems, collegelevel essays, computer code, and even jokes, represents a pivotal moment. In this chapter, we will introduce generative AI, an impressive technology that offers a multitude of benefits across various domains and holds great potential for revolutionizing many industries. We will also examine its advantages, limitations, and the potential risks associated with the use of this technology.
• Chapter 6—In this chapter, we will explore various aspects of human cognition to illustrate what it means to be imaginative, intuitive, curious, and creative. We’ll show that current AI falls short in emulating these traits. We’ll compare human reasoning to AI to examine whether machines can replicate human-like thinking. Additionally, we’ll reflect on our limited understanding of the human mind. Furthermore, we will highlight that genuine comprehension is a prerequisite for vision, revealing the current limitations of AI algorithms in recognizing objects and their substantial gap in achieving human-like object and scene perception.
• Chapter 7—In this chapter, we will highlight that no matter how extensive the dataset or advanced the algorithms, AI programs ultimately fall short of attaining genuine intelligence. We will elaborate on the challenge AI encounters when attempting to extract true intelligence from data, as even with current AI techniques excelling in data processing, they continue to grapple with comprehending its deeper nuances.
• Chapter 8—We demonstrate that despite concern about AI taking our jobs, most human tasks are still out of the reach of AI.
• Chapter 9—The prevailing narrative often suggests that AI’s evolution will result in intelligent robots capable of replicating themselves, ultimately leading to the downfall of human civilization. While this scenario might be the stuff of compelling fiction, it doesn’t align with reality. In this chapter, we aim to debunk the notion of technological singularity as baseless and argue that our concerns should focus less on AI and more on the potential pitfalls of artificial stupidity.

• Chapter 10—Each AI project, whether it meets with success or faces hurdles, offers a wealth of valuable lessons. Drawing insights from these experiences empowers us to make informed decisions, steering our AI projects toward favorable outcomes while steering clear of common pitfalls. In this chapter, we will discuss insights gained from both the missteps and achievements of past AI

• projects. Furthermore, we will provide valuable guidance on assembling the right team, cultivating the necessary mindset, and crafting a promising strategy for your AI project.

• Chapter 11—In this chapter, I use my three decades of experience in the development and deployment of mission-critical AI systems where reliability, precision, and effect are not mere goals but absolute necessities. I will describe a set of characteristics that, in my perspective, will define the next generation of AI platforms.

liveBook discussion forum