Artificial Intelligence Introduction

A typical AI system receives information, processes data using algorithms or models, produces an output, and may use that output to perform an action.

Narrow AI

Narrow AI is designed to perform specific tasks. Most practical AI applications used today fall into this category.

Examples:

• Spam email detection
• Product recommendations
• Speech recognition
• Fraud detection
• Image classification
• Chatbots

General AI

General AI refers to a hypothetical system capable of performing a broad range of intellectual tasks at a human-like level.

Superintelligence

Superintelligence is a hypothetical concept describing AI whose capabilities would exceed human abilities across many areas.

AI is the broader field. Machine Learning is a major approach within AI, while Deep Learning is a specialized approach within Machine Learning.

Python is widely used for AI development because it provides a large ecosystem of libraries and tools.

• Simple and readable syntax
• Large AI and Machine Learning ecosystem
• Strong community support
• Excellent data-analysis libraries
• Easy experimentation and prototyping
• Integration with APIs and applications

Important Python Libraries

For basic AI and Machine Learning experiments, install common Python libraries using pip.

pip install numpy pandas matplotlib scikit-learn

For many beginner AI projects, these libraries provide enough functionality to work with datasets, train models, and evaluate predictions.

Suppose a real-estate company wants to predict the selling price of a house based on its size.

Historical data might look like this:

An AI or Machine Learning model can learn the relationship between house size and price.

We can use a Machine Learning algorithm to learn the relationship between house size and price.

from sklearn.linear_model import LinearRegression

X = df[["size"]]
y = df["price"]

model = LinearRegression()

model.fit(X, y)

prediction = model.predict([[1400]])

print("Predicted price:", prediction[0])

Predicted price: 76.5

The exact prediction depends on the dataset and trained model. The important concept is that the model learned a relationship from historical examples.

Email providers can use AI to identify messages that are likely to be spam.

The model can learn from previously labeled messages and identify patterns associated with unwanted emails.

E-commerce and streaming platforms can use AI-based recommendation systems to suggest content or products.

A telecom company may want to identify customers who are likely to stop using its service.

A trained model can use historical customer information to estimate churn risk for new customers.

Financial systems can use AI and Machine Learning to detect unusual transaction patterns.

Features may include transaction amount, location, time, transaction frequency, and account behavior.

An AI chatbot can process a customer's message and determine the appropriate response or workflow.

message = "Where is my order?"

if "order" in message.lower():
    response = "Please provide your order number."
else:
    response = "How can I help you?"

print(response)

Please provide your order number.

This simple example uses a rule. More advanced AI systems can understand intent, context, language variations, and conversation history.

Natural Language Processing (NLP) enables computers to work with human language.

Simple Python Example

review = "The product quality is excellent"

positive_words = ["excellent", "good", "great", "amazing"]

if any(word in review.lower() for word in positive_words):
    print("Positive review")
else:
    print("Review needs further analysis")

Computer Vision enables computers to analyze visual information.

• Object detection
• Face recognition
• Image classification
• Medical image analysis
• Quality inspection
• Document processing

Generative AI refers to AI systems that can generate new content based on learned patterns and user input.

1. Define the business or technical problem.
2. Collect relevant data.
3. Clean and prepare the data.
4. Select useful features.
5. Train an appropriate model.
6. Evaluate model performance.
7. Deploy the model when appropriate.
8. Monitor the model after deployment.

Machine Learning models should generally be evaluated on data that was not used during training.

from sklearn.model_selection import train_test_split

X_train, X_test, y_train, y_test = train_test_split(
    X,
    y,
    test_size=0.2,
    random_state=42
)

A model should be evaluated using metrics that match the problem.

An HR analytics system may estimate salary based on information such as experience, education, role, and location.

AI-based computer vision systems can assist professionals by analyzing medical images and identifying patterns that may require additional review.

An e-commerce company can use several AI technologies throughout its customer journey.

• Can process large amounts of data quickly.
• Can identify patterns that may be difficult to detect manually.
• Can automate repetitive tasks.
• Can provide predictions and recommendations.
• Can operate continuously when properly designed.
• Can support data-driven decision making.

• AI systems depend heavily on the quality of their data.
• Models can reproduce patterns or biases present in training data.
• Predictions can be incorrect.
• Complex models may be difficult to interpret.
• AI systems require monitoring and maintenance.
• Privacy and security must be considered when handling data.
• High-stakes applications require appropriate human oversight.

1. Learn Python fundamentals.
2. Learn NumPy and Pandas.
3. Learn data cleaning and visualization.
4. Learn Machine Learning fundamentals.
5. Build regression and classification projects.
6. Learn model evaluation.
7. Study neural networks and Deep Learning.
8. Explore NLP, Computer Vision, or Generative AI.
9. Build end-to-end real-world projects.

Business Problem

An online store wants to predict whether a customer is likely to purchase a product.

Available Data

• Customer age
• Number of website visits
• Time spent on website
• Number of products viewed
• Previous purchases

AI Workflow

Example Python Structure

import pandas as pd

from sklearn.model_selection import train_test_split
from sklearn.linear_model import LogisticRegression

data = {
    "age": [22, 35, 28, 42, 31, 25],
    "visits": [2, 8, 5, 10, 7, 3],
    "products_viewed": [3, 12, 7, 15, 10, 4],
    "purchased": [0, 1, 1, 1, 1, 0]
}

df = pd.DataFrame(data)

X = df[["age", "visits", "products_viewed"]]
y = df["purchased"]

X_train, X_test, y_train, y_test = train_test_split(
    X,
    y,
    test_size=0.2,
    random_state=42
)

model = LogisticRegression()

model.fit(X_train, y_train)

prediction = model.predict([[30, 6, 8]])

print("Predicted class:", prediction[0])

Building an AI system is not only a technical task. Developers should also consider how the system affects people and how data is collected and used.

• Starting with a complex model before understanding the problem.
• Using poor-quality or irrelevant data.
• Ignoring missing and incorrect values.
• Evaluating a model only on training data.
• Choosing a metric without considering the business problem.
• Assuming a high score always means the model is useful.
• Ignoring data leakage.
• Deploying a model without monitoring its performance.

• ☐ Define the problem clearly.
• ☐ Identify the required data.
• ☐ Clean and validate the data.
• ☐ Select useful features.
• ☐ Split data appropriately.
• ☐ Train a suitable model.
• ☐ Evaluate the model using relevant metrics.
• ☐ Check for overfitting and data leakage.
• ☐ Test the system with realistic examples.
• ☐ Monitor the system after deployment.

1. AI is a broad field; Machine Learning is one major approach within it.
2. Good data is essential for useful AI systems.
3. Always define the problem before selecting a model.
4. Separate training and evaluation data appropriately.
5. Use evaluation metrics that match the problem.
6. Do not assume model predictions are always correct.
7. Monitor models after deployment.
8. Protect sensitive and personal data.
9. Consider fairness, reliability, transparency, and security.
10. Use human oversight where the consequences of errors are significant.

Artificial Intelligence Introduction

Learn AI concepts with Python and real-time examples.