ARTIFICIAL INTELLIGENCE

Artificial Intelligence in Our Daily Life:
To elaborate on some of its subdisciplines that directly affect our daily life, consider data mining, natural language processing, computer vision, robotics, and expert systems.

●       Natural Language Processing:
This is an AI application that understands, interprets, and generates responses in human language. Chatbots and virtual assistants, which analyze text and speech data to interact with users, are examples of natural language processing. Additionally, translation applications and sentiment analysis—such as determining whether social media comments are positive or negative—are other capabilities of natural language processing.

●       Computer Vision:
Computer vision is used to analyze, interpret, and understand digital images and video data. It includes applications such as object recognition, motion tracking, face recognition, and image classification. By working together with machine and deep learning, it offers technology that simplifies many aspects of human life, from healthcare to security.

●       Robotics:
Robotics involves the development of autonomous systems capable of performing tasks in the physical world and controlling machines. It enables robots to interact with their surroundings and users through various sensors. Examples include autonomous vehicles that drive by recognizing traffic signs and other objects in their environment, as well as industrial robots that work swiftly and precisely on production lines in factories.

●       Expert Systems:
Expert systems are capable of solving complex problems based on a knowledge-based set of rules that mimic human decision-making processes in specific domains. For instance, systems that provide diagnostic suggestions by accessing a medical knowledge base or financial advisory systems that analyze market data to offer investment recommendations are applications of expert systems.

MACHINE LEARNING

Machine learning is a subfield of artificial intelligence that enables computers to improve their performance by learning from data, and it has a wide range of applications. It is divided into three subcategories: supervised learning, unsupervised learning, and reinforcement learning.

Supervised Learning:
This involves training models using labeled data to make predictions and determine the correct function. It is a type of learning where both input and output values are provided.

Unsupervised Learning:
This approach aims to discover patterns and structures in unlabeled data. It is based on inferring the structure of the data and finding similarities or differences, with only input values provided and no corresponding output information.

Reinforcement Learning:
Reinforcement learning is a type of machine learning in which an agent learns through trial and error by using feedback from its actions, employing reward and punishment mechanisms to shape its decision-making process. It is used in strategic games such as chess and was notably applied in AlphaGo, the first computer program to defeat professional players in the game of Go.

DEEP LEARNING

Deep learning is a structure composed of neural networks that consists of an input layer, multiple hidden layers, and an output layer.

The hidden layers learn and refine the information received from the input layer to achieve highly accurate predictions of the desired output. The strength of the connections between these layers, known as weights, is learned from the training data.

Example Using a Classification Algorithm

Comparison of Machine Learning and Deep Learning

Deep neural networks, which are composed of many layers, have achieved high success by inferring complex features from data without the need for explicit feature specification. For example, to compare machine learning and deep learning, one could consider the task of determining the class of a flower from its image. In machine learning, features such as petal length and width need to be provided manually, whereas in deep learning, the system extracts these features on its own.

Step-by-Step Artificial Intelligence Training Process

Step 1: Data Collection

Step 2: Data Cleaning and Preprocessing
Data cleaning removes unnecessary information and handles errors, redundancies, and missing values. Meanwhile, preprocessing transforms the cleaned data to be compatible with the AI algorithm.

Step 3: Data Labeling
Data labeling is the process of annotating raw data with relevant labels that describe the data and make it machine-readable.

Step 4: Splitting the Dataset
While the training set is used to teach the model, the test set is used to evaluate its performance.

Step 5: Data Balancing and Bias Reduction

Step 6: Parameter Tuning

Step 7: Model Evaluation and Validation

How Should a Question Be Asked?

Large language models generate responses based on the commands they receive. For example, giving the instruction “You are a computer scientist” will yield a very different response compared to “You are an economist.” This difference is due to the underlying Transformer architecture. The relationships between words in the query create a contextual environment that shapes how the next word is processed and how the response is formed.

After the phrase “You are a lawyer,” the model tends to use technical jargon and prioritize legal terminology. In contrast, the phrase “You are an economist” causes the model to lean towards economic theories, trends, or statistical data. This dynamic shift is achieved by reweighting the probabilities of word choices based on the provided context.

●       Defining Roles:
Example: “You are a helpful assistant.”

●       Specifying Tasks:
Example: “Write a Python script for sorting algorithms.”

●       Specifying Output Style:
Example: “Explain this as if to a 5-year-old.”

These nuances help the model generate specific, accurate, and meaningful outputs. Additionally, providing all the relevant information in one go helps in obtaining the correct answer to the question asked.

Training link: https://youtu.be/o5fkdYAYvEY?si=LCLvaHybeatNCJ7A