AI in Human Terms

Table of Contents

Intro: How Did We Get Here So Fast?

Scraps of Evolution

Part 1: Acronyms, Math, and Science, Oh My

What is AI?

From Being Told, to Learning

Model Machines

Machine Learning

Training Models: Supervised and Unsupervised Learning

Supervised Learning

Unsupervised Learning

Reinforcement Learning

Neural Networks

Deep Learning

Feedforward Neural Networks (FNN), You Can’t Go Back

Recurrent Neural Networks (RNN), I Remember Something

Convolutional Neural Networks (CNN), I See Something

Wrapping Up Deep Learning

Part 2: A Brief History of AI, The 1950s

Intelligent Machines

The Mind

The Maze

Dartmouth

Teaching Computers To Read

The 1960s: Ready for Liftoff

DARPA’s Big Save

LISP’s List

ELIZA, My Fair Lady?

Good Neighbors

Building Blocks

The 1970’s: Expert Solutions

Shakey The Robot

DENDRAL, MYCIN, and Prospector’s Prospects

Frames

Speech

Sometimes You Need To Go Back

The Mansfield Amendment & The Lighthill Report

The 1980’s: Winter

The New Experts

Backpropagation Revisted

Robots

The 1990’s: Bigger Data, Small Machines

Big Data Part 1: Deep Learning & Network Architecture

Rise of Le Machines

Support Vector Machines

Big Data Part II: Language and Information

The New Millenium: AI Goes Mainstream

The New Machines

Even Bigger Data (Forest and Trees)

The Semantic Web

Reinforcement Learning

From Forest to Mines

Big Data Includes Big Images

The Tens / The Twenties / A New Spring

Deep Learning Revolution

Rise of the Large Language Models

The 2020s: A Pandemic Leads To AI Spring

Large Language Models Capture Everyone’s Imagination

Attention Is All You Need

Large Language Models

How do LLMs work?

So what is a prompt?

Dreams, we all have them

Fine Tuning

Retrieval Augmented Generation (RAG)

Agents (specialized programs)

Bias and LLMs

Other Applications of Generative Models

Generative Images

Generative Audio

Final Thoughts & Recommended Resources

Intro: How Did We Get Here So Fast?

Imagine waking up one day to find that artificial intelligence, once a distant dream, is now intricately woven into the fabric of your daily life.

In recent years, as I’ve spoken with a diverse group of individuals from parents, businesspeople, retail workers, to medical staff and many managers across industries - it’s become clear that artificial intelligence, or AI, has taken many by surprise. The common refrain during these conversations is one of disbelief: This came out of nowhere. What happened? Yet, as sudden as the advent of Generative AI (GenAI) might seem, its roots are deep and complex, often overlooked in our daily hustle.

Figure 1 - Timeline

Let’s start with a brief timeline, a favorite starting point in my presentations. The journey to today’s AI began decades ago, not with flashy startups or tech juggernauts, but with the bulky mainframes of the mid-20th century. These evolved into minicomputers and, eventually, into the personal computers we know today—like the Apple II launched in 1977, a landmark in computing history.

Following personal computers, the next significant leap was the internet, introduced as the World Wide Web in 1993. Originating from Advanced Research Projects Agency ARPANET in the late ‘60s, it connected researchers globally and, after years of development and the convergence of key technologies like browsers and web servers, gained massive traction by the late ‘90s.

Up next, mobile technology, too, transformed our lives. Recall the early car phones of the late ‘80s—a precursor to the smartphones that would redefine our communication landscape with the introduction of the iPhone in 2007. These devices merged touch screens, internet connectivity, and apps into an indispensable tool.

Fast forward to November 2022: the launch of ChatGPT by OpenAI. Suddenly, AI was not just a utility but a household name, with roughly 100 million people interacting weekly with ChatGPT and 1.6 billion visits to the OpenAI site over a month. To grasp the rapid emergence of GenAI, we must look back to foundational technologies like Google Translate and other natural language solutions that set the stage years earlier, as well as non-generative AI approaches.

These technological shifts didn’t just appear; they evolved, unnoticed, becoming as essential as the internet, laptops, and mobile phones are today. AI has been a quiet companion in our technological journey, tracing back to pioneers like Alan Turing in 1946 and computing in general with Ada Lovelace, whose 1843 Notes contained what is considered the first algorithm.

Let’s delve into the impact of AI understanding the concepts and history followed by a focus on Large Language Models (think ChatGPT). The goal is to demystify these technologies using relatable terms, making clear how integral AI has become to our modern existence.

Scraps of Evolution

…the kind of control you’re attempting is not possible. If there is one thing the history of evolution has taught us, it’s that life will not be contained. Life breaks free, it expands to new territories. It crashes through barriers. Painfully, maybe even…dangerously, but and…well, there it is.

– Malcolm to John Hammond, Jurassic Park¹

If there’s a lesson to be learned from Jurassic Park, it’s that our creations often take on lives of their own. Transform ancient mosquitoes into dinosaurs, and those very dinosaurs might just fight back. View life—or, for our purposes, technology and specifically AI—from any angle, and it inevitably swerves from our most meticulous plans. Reflect on this iconic line from Jurassic Park, substituting ‘life’ with ‘AI’:

There’s a poignant reason I chose Jurassic Park to start a dialogue about AI. Imagine if the events of Michael Crichton’s film unfolded today—dinosaurs roaming through New York City, theme parks brimming with visitors, and an investor frenzy around the latest Elon Musk venture, ‘TyrannosaurusX’. When we substitute ‘life’ with ‘AI’ in the dialogue, the analogy sharpens: in recent years, AI has been ‘breaking free’ on its own terms.

Technology sneaks up on us. Suddenly, OpenAI’s ChatGPT bursts onto the scene, and within months, what seemed like overnight, its user base exploded to over 150 million by May 2023. ChatGPT swiftly became a term as common as household names. The entire world buzzed about artificial intelligence—a concept so potent it captured global imaginations.

Yet, AI isn’t a sudden phenomenon. But if it wasn’t for putting a friendlier face on AI, democratizing it’s use, we may have been waiting longer.

For nearly seven decades, artificial intelligence has been brewing, growing significantly more sophisticated over the past twenty years. AI didn’t just appear; we nurtured it, fed it with our data, our daily technology use—connecting with friends, deciding what to watch, ordering food, navigating cities.

We enabled its evolution.

Now, we face an urgent question: How will AI ‘find a way’? Humanity inches ever closer to crafting an intelligence that could surpass its creators. AI’s potential to enhance or threaten our way of life hangs in balance. As artificial intelligence seeps into consumer hands worldwide, we hit a pivotal moment. It could surpass us, and stringent regulations may be necessary and unable to curb the unchecked ambitions of AI firms. Understanding AI’s roots—and its trajectory—is critical.

We need to drill into the amber, extracting the DNA that shaped today’s AI. Understanding how to teach a computer to complete ‘I want peanut butter and…’ with ‘jelly’ or use a Shakespearean twist to create ‘a spread of peanut paste’ requires no math or computer science degree, just curiosity.

For almost seven decades, we’ve been unraveling the artificial brain—pondering whether a computer can play chess, solve puzzles, or write essays. Today, AI’s capabilities challenge the breadth of human knowledge, raising profound questions:

Should we, or shouldn’t we?

As we stand at this critical juncture, our choices will shape the future. AI demands our engagement. Deciding whether AI will redefine our work and life, knowledge—particularly human knowledge—is our greatest ally. We need to deconstruct the complexity of AI’s concepts and history to navigate our future.

We are not mere bystanders in the saga of AI. We are its architects and benefactors. The power that AI affords us is unprecedented—we can communicate across languages we don’t speak, predict calamities, and save lives, yet we also face risks like eroding privacy and escalating surveillance.

The choice is ours. Being informed is essential. As AI’s DNA continues to weave into our daily existence, albeit often unnoticed, becoming part of our lives. How we choose to interact with it, what we expect from others, corporations and governments, and how we harness its potentials or mitigate its risks, will define our future.

This book is structured in three parts: a breakdown and explanation of key AI concepts simplifying the jargon (in human terms), a concise history of AI, and a brief spotlight on large language models powering generative AI. While you can explore each part as you’d like I invite you to join me on this AI journey— through its theoretical underpinnings, it’s origins, to understanding Generative AI. Who knows? We might even encounter some proverbial dinosaurs along the way.

Part 1: Acronyms, Math, and Science, Oh My

Pure mathematics is, in its way, the poetry of logical ideas.

– Albert Einstein

While the realms of math (from which AI primarily springs) and science (specifically computer science and machine learning) can seem as daunting as a Jurassic T-rex, there’s no need for panic or a frantic escape in a Jeep. Our journey into AI won’t require deep scientific knowledge or familiarity with endless three-letter acronyms that might send you checking your rearview mirror in terror. Instead, we’re here to demystify AI, breaking it down into accessible, practical terms—no dinosaurs included.

This part attempts to strip away the complexities of AI, but introduce some specific terms. You don’t need to be a science, math, or philosophy major; just bring your sense of adventure and perhaps a bit of common sense. Our goal is to explore the multifaceted ways AI influences everyday life without getting bogged down in the theoretic details that researchers and scholars might appreciate. We’re not here to build our own AI models; rather, we’re here to understand AI in human terms. Behind every daunting concept in science or math lies a basic idea that most people can grasp. Here, we distill AI to its core elements to make the concepts more relatable and less intimidating.

Technology moves at a breakneck pace, with yesterday’s breakthroughs quickly becoming today’s old news. The use of AI, while a staple at tech giants like Google, Amazon, Meta, and Microsoft for decades, has only recently entered the public consciousness through the advancements in Large Language Models. This is fortunate because it gives the rest of us time to catch up. As we begin to understand these models and approaches, we’re better equipped to engage with critical questions: Should we? How will we? and What impacts will this have? In the next set of chapters certain words will be highlighted in bold as they represent some of the core terms used in AI. Let’s dive in, ready to untangle the complexities of AI and explore its profound implications on our world.

What is AI?

In human terms, think of AI as an ‘intelligent machine.’ While various definitions exist, at its core, AI is about machines performing tasks typically done by humans, predominantly making predictions. Will you buy a new dress? What’s tomorrow’s weather? Will a stock price rise or fall? Is that a cat or a dog in the picture? Draft an email for a job application or even