Basic Terms in AI Ethics (Part 2)

The 'Governance Guidelines for Implementation of AI Principles Ver. 1.1' is a set of guidelines formulated by the Ministry of Economy, Trade and Industry, which explains AI by dividing it into three broad categories: 'deep learning,' 'machine learning,' and 'artificial intelligence in a broad sense.' The terms used in these guidelines are essential knowledge for anyone involved with AI.

1. Deep Learning is a type of 'machine learning' that uses multi-layered 'neural networks' to extract complex patterns and the like. 'Deep learning' learns using large amounts of data and demonstrates high performance in many fields, such as 'image recognition,' 'natural language processing (NLP),' and 'speech recognition.'

2. Machine Learning is an approach that allows computers to learn automatically using data and perform tasks such as prediction and classification. Machine learning includes methods such as 'supervised learning,' 'unsupervised learning,' and 'reinforcement learning,' each of which corresponds to different tasks and data. Machine learning covers a wide range of technologies, including deep learning.

3. General AI (Artificial Intelligence in a broad sense) is an AI technology that broadly mimics human intelligence, and there are various approaches including machine learning and deep learning. General AI aims to possess a wide range of intellectual capabilities such as natural language processing, learning, 'knowledge representation,' 'inference,' 'problem solving,' 'knowledge acquisition,' 'visual recognition,' and 'speech recognition.' The goal is to realize AI that has intelligence as flexible and adaptable to situations as humans.

1. Natural Language Processing (NLP)
(1) Natural Language Processing (NLP) is a technology for computers to understand, analyze, and generate languages used by humans.

(2) GPT (Generative Pre-Trained Transformer) is a deep learning model for performing natural language processing (NLP). There are various types of NLP besides GPT, and many companies are developing and researching their own models and technologies, which are used in various applications and services.

Also, there are many open-source NLP libraries and frameworks designed to allow researchers and developers to easily work on NLP tasks. For detailed explanations of these, please see the specialized section we will create.

2. Neural Networks
(1) Neural Network is a machine learning model that mimics the nerve cells (neurons) of the human brain and is the foundational technology for machine learning and deep learning. A neural network consists of multiple layers, and each layer contains many neurons. As data passes through the layers, features are extracted, and ultimately, predictions or classifications are performed. Major types include FNN, CNN, RNN, LSTM, GRU, GAN, Autoencoder, VAE, Transformer, Capsule Network, SNN, etc., which are selected and combined according to the task and data.

3. Learning Methods
(1) Supervised Learning is a machine learning method where the model learns from data using known input data and corresponding correct labels.

(2) Unsupervised Learning is a machine learning method for finding structures or patterns from data for which no correct labels are provided.

(3) Reinforcement Learning is a machine learning method where an agent interacts with an environment and learns actions to maximize rewards.

4. Data Analysis
(1) Regression Analysis is a statistical method used to model relationships between variables and to estimate predictions or causal relationships.

(2) Classification is a task of assigning data to known categories and is one of the goals of machine learning.

(3) Clustering is a task of grouping data based on similarity and is a type of unsupervised learning.

5. Dataset
(1) Dataset is a collection of data used for machine learning and statistical analysis.

(2) Training Data is the portion of data used for a machine learning model to learn.

(3) Test Data is the portion of data used to evaluate the performance of a machine learning model.

6. Learning Optimization
(1) Loss Function is a function that quantifies the error between the model's prediction and the correct label, and is used for the purpose of minimization during learning.

(2) Optimization Algorithm is a method for updating model parameters to minimize the loss function.

(3) Overfitting is a phenomenon where a model fits too closely to the training data, resulting in decreased performance on unknown data.

(4) Generalization Performance is a performance metric indicating how accurately a machine learning model can predict or classify unknown data.

7. Features
(1) Feature Extraction is the process of extracting important information or patterns from data, which is used as input for machine learning models.

(2) Feature Selection is the process of selecting the most important features in a dataset, which contributes to improving model performance and reducing computational costs.

8. Model Evaluation and Selection
(1) Hyperparameter is an external setting value that influences the learning process of a machine learning model and is often adjusted manually.

(2) Cross-Validation is a method for evaluating the generalization performance of a model by dividing the dataset into multiple parts and using each as test data.