Introduction: A Guide to Properly Understanding the Structure and Principles of Artificial Intelligence

I The Rise of Artificial Intelligence

Computers can read pictures
Definition of artificial intelligence
Misuse of AI terminology
History of artificial intelligence technology
Key examples of artificial intelligence

Ⅱ Data and Artificial Intelligence

Classification by data type
How to analyze and utilize data
Data that requires a lot of hands-on time
Big data and artificial intelligence
What is a data analysis model?
Key Use Cases of Data Analysis

III Machine Learning

Definition of Machine Learning
The emergence of linear functions
The meaning of hypothetical
Difficulty in finding weights
linear regression
Calculating loss costs
gradient descent
Try machine learning in practice
polynomial linear regression
binary classification
Multi-classification
Computers that read pictures
MNIST dataset
Recognize images

IV Deep Learning

neurons
perceptron
Verifying the Perceptron ①
Verifying the Perceptron ②
The emergence of multilayer perceptrons
The significance of multilayer perceptron
DNN
Shift in development paradigm
CNN ①
CNN ②
CNN ③
RNN
Utilization of RNN

V Unsupervised Learning

Three Types of Machine Learning
Supervised learning of character and word prediction models
The concept of unsupervised learning
K-means
GAN

VI reinforcement learning

Principles of reinforcement learning
Use Cases of Reinforcement Learning
Reinforcement learning in a greenhouse?

Ⅶ Large-Scale Language Model

The emergence of ChatGPT
LLM: A Machine Learning Issue
Natural language processing
Key features implemented through natural language processing
A leap forward in natural language processing technology
Basic concepts of word embeddings
How word embeddings represent words
Basic concepts of transfer learning
Transfer learning in natural language processing
Transfer learning for large-scale language models
Large-scale language models with transfer learning at various levels
Basic concepts of language models
Extension of the language model concept
Encoder-decoder model
context vector
Why Attention Is Needed
Attention mechanism
Transformers
Transformer's Attention
Features of the transformer mechanism
Large-Scale Language Models - BERT and GPT
Output of BERT and GPT
Secrets of High-Performance Language Models ①
Secrets of High-Performance Language Models ②
Limitations of large-scale language models
Significance of the emergence of large-scale language models
In the end, it's the same basic principle

Ⅷ High-performance machines and people

Vulnerabilities in AI Technology ①
Vulnerabilities in AI Technology ②
Vulnerabilities in AI Technology ③
The evil(?) nature of AI technology
Artificial intelligence is ultimately about data
Data cleaning and labeling
Obtaining data for language models
Does artificial intelligence have intelligence?
Strong AI and Weak AI
The possibility of the emergence of strong artificial intelligence
General purpose artificial intelligence
The potential of behavioral models
A world with humanoids
AI service development process
To become an AI expert
How should we respond going forward?

References
Source of images and photos
Search

Only those who understand artificial intelligence will survive!
An AI Guidebook to Enhance AI Literacy

This is truly the era of artificial intelligence.
Artificial intelligence is everything, and it will be everywhere.
As the surrounding environment changes, IT developers, planners, and practitioners can be assigned to AI-related tasks at any time.
If an AI beginner is assigned to a job, what should they do first?
If you have to work in the field without even having time to lay the foundation, you will usually end up in a difficult situation.
Moreover, the field of artificial intelligence has a wide variety of topics to study and a large amount of material to cover.
Here's a quick and easy way to understand AI technology from the ground up.
"Textbook on the Principles of Artificial Intelligence Structure" will be a helping hand to many who are struggling.

Unearthed through encounters on site
Talking about misconceptions and questions surrounding AI

This book is a compilation of what Senior Researcher Song Kyung-bin, who works in the AI ​​Convergence and Diffusion Team at the Korea Institute for Advancement of Intelligence Information Society, has learned while working on related projects in the field.
In short, this is an AI concept guide designed to help anyone learn the core principles and structure of artificial intelligence.

The author said he decided to write this book because the fragmentary information scattered on the Internet not only fails to fully understand artificial intelligence, but also fails to properly understand it.
In particular, he said he felt uncomfortable whenever he saw news or interviews treating AI as if it were an intelligent entity, despite the fact that AI currently lacks intelligence, defined as the "ability to engage in intellectual activity," due to technological limitations. He decided to personally address the misconceptions and questions surrounding AI.

You have to know correctly to deal with it correctly.
This book aims to provide practical help by providing accurate information about AI to those who need to jump into the field right away but feel they have no basic knowledge, as well as to the general public and students who want to learn about artificial intelligence.

Artificial intelligence begins with middle school math.
Core principles and structure of machine learning and deep learning technologies

This book is structured in steps so that anyone can easily begin studying artificial intelligence.
After explaining an easy concept, we explain a more difficult concept that builds on it.
According to the author, he tried to write a book that even elementary school students could understand, but due to his own lack of ability(?), he was satisfied with a level that middle school students could understand.
This passage reveals the author's confidence that even those with liberal arts backgrounds, let alone non-majors, can fully understand artificial intelligence technology.


Readers who read this book one chapter at a time will find themselves at the heart of understanding artificial intelligence, a field that others have given up on because it is too complex.
After a brief history of artificial intelligence, we begin with data types and analysis, and gain a thorough understanding of the technological trends and history of machine learning and deep learning, including how they emerged and developed to reach their current state.


When explaining the principles of machine learning, this book starts with linear functions.
The author's claim that even middle school students can understand artificial intelligence, as linear functions are learned in middle school, is not an exaggeration.
The book uses the simplest machine learning model (linear regression) to find the probability of a relationship as an example.

Linear regression is a model that expresses a phenomenon as a straight line on a graph.
In the example in the book, the value of the relationship increases linearly with the number of apples, which can be expressed as a formula: y=ax+b.
Here, x is the number of apples and y is the difference.
If we change this into the hypothetical form used in machine learning, it becomes y=wx+b.
A hypothesis is a formula that calculates a prediction result based on learning data.
As you can see from the shape of the equation, it is ultimately in the form of a linear function.

In machine learning, learning is the process of figuring out w and b in this formula.
Substitute the number of apples and price data into the hypothetical equation y=wx+b to optimize w and b.
That is, the goal is to find the values ​​of w and b that produce the correct answer.
A machine learning model that has gone through this process can predict how much five apples will cost.

Setting up hypotheses like this and finding the weights of those hypotheses is the foundation of artificial intelligence learning.
Machine learning technology, which must find answers solely from data without any specific rules or procedures, adjusts weights to move toward the correct answer.

The algorithm used at this time is gradient descent, which can be said to be the core of understanding machine learning.
This is because gradient descent is applicable to all machine learning tasks, from relatively simple models that find a mean value to incredibly complex large-scale language models.


To fully understand artificial intelligence, you need to know a variety of other concepts.
Especially for large-scale language models like ChatGPT, more complex concepts are used.
The book devotes the largest portion of its volume to a serious discussion of large-scale language models.
Although the content is complex, readers can learn technical concepts such as word embedding, attention in transformer models, and self-attention through this.
You can technically understand how language models understand words, sentences, paragraphs, and even the relationships between them.


If you understand correctly, you can see
The true nature of artificial intelligence and its future applications

The author says that it is necessary to properly understand the fundamental principles that are at the core of artificial intelligence technology.
Otherwise, we may fall into the misconception that ‘AI has its own intelligence’, leading to vague fears, anxiety, or excessive expectations or optimism.
Let's think back to when ChatGPT 3 first came out.
Many news media outlets have provided various explanations, such as saying that it has 175 billion parameters or that it is similar to neurons, but there have not been many cases where its technical nature and principles were clearly explained.
Vague and vague explanations are still often seen in media articles even after quite some time has passed.
Explanations that lack technical understanding like this can fuel many misunderstandings about artificial intelligence.


This may result in unnecessary costs and conflicts.
For example, if a practitioner working on an AI project misunderstands AI and aims for “self-learning AI,” it would be a serious problem.
There is currently no such technical concept in artificial intelligence, so we are completely on the wrong track.
On the other hand, if we resist technological change out of fear that artificial intelligence will completely replace humans and show interest in bizarre behaviors like the 21st-century Luddite movement, we will only lose our competitiveness.


This is when we need the right understanding.
Let's recall the basics of artificial intelligence.
Any machine learning technique starts with a hypothesis.
We must understand that no matter how advanced artificial intelligence is, it is ultimately nothing more than a formula with sophisticated weights.
Just look at the large-scale language models that are attracting people's attention these days.
Despite its ability to generate fluent sentences, it lacks common sense and is prone to bias depending on the training data.
Another issue that continues to be pointed out is that the language model cannot explain how its results came about.

With a sound technical understanding like this, practitioners can move AI projects in the right direction, and developers can accelerate their technical acquisition and understanding. Ordinary people without direct contact with the AI ​​field can let go of unnecessary worries and anxieties and calmly consider how to utilize AI in their own fields.

Artificial intelligence technology with limitless potential for development
Understand the basic principles and use them wisely.

Artificial intelligence is ultimately just an automated technology that finds weights.
However, we cannot simply ignore it.
Because it handles things amazingly well that it couldn't do before.
Its unique processing method, which uses data alone to find solutions, has incredible potential.
This is a reality that has been proven by the emergence of various models since the advent of ChatGPT.
It is showing outstanding performance not only in conversation but also in various fields such as finance, medicine, new drug development, translation, and autonomous driving, showing the possibility that it will one day replace many human tasks.

Artificial intelligence, or more precisely, machine learning technology, is still in its infancy and is likely to advance further in the future.
Of course, as mentioned earlier, there are many limitations and problems.
This is why the author continues to insist that in order to utilize AI wisely, one must properly understand the basic principles and structure of AI.
Let's properly understand the core principles of artificial intelligence with this book.
It will certainly help us wisely prepare for the era of artificial intelligence.