Study sheet: Artificial Intelligence Fundamentals

📋 Course Outline

1. Artificial Intelligence and Its Scope
2. The AI Concept Hierarchy
3. Machine Learning from Data
4. Deep Learning and Neural Networks
5. Generative AI and New Content
6. Applications and Classification
7. Limitations and Hallucinations
8. Responsible Use and Verification

📖 1. Artificial Intelligence and Its Scope

🔑 Key Concepts & Definitions

• Artificial intelligence : a branch of computer science that aims to build systems capable of performing tasks that usually require human intelligence and simulating human thinking.

★ Must-know

📌 Most current AI systems have a narrow scope and are designed to perform one task or a limited group of tasks rather than everything a human can do.

Further detail

• Examples of AI applications include:
  • spam filtering
  • product recommendations
  • translation
  • speech recognition
  • image recognition

💡 Memory Hook

A simple calculator performs a narrow task, whereas AI imitates human intelligence across broader tasks.

📖 2. The AI Concept Hierarchy

🔑 Key Concepts & Definitions

• Machine learning : a method in which a system learns patterns from data to make predictions or judgments.
• Deep learning : an advanced approach within machine learning that uses neural networks to learn complex patterns and generally relies on relatively large amounts of data.
• Generative AI : uses deep learning to generate entirely new content and data, including text, images, and audio, rather than merely analyzing existing data.

📝 Essential Points

📌 Artificial intelligence is the broadest field, machine learning is a part of artificial intelligence, deep learning is a specialized approach within machine learning, and generative AI commonly relies on deep-learning models to create new content.

💡 Memory Hook

AI → ML → DL → GenAI: broad field to specialized content creation.

📖 3. Machine Learning from Data

★ Must-know

• In machine learning, the system studies many previous examples, discovers recurring patterns, and uses those patterns to predict the category or outcome of a new case.

Further detail

• A spam filter learns from previous email messages, identifies recurring patterns, and predicts whether a new message is spam.

• A recommendation system learns from previous purchasing data and uses the detected patterns to recommend a new product.

💡 Memory Hook

Data → pattern learning → prediction or judgment.

📖 4. Deep Learning and Neural Networks

🔑 Key Concepts & Definitions

• Artificial neural network : a computational model inspired in a simplified way by connections between neurons, using connected layers to process inputs and produce outputs.

★ Must-know

• 🔄 A visual neural-network process includes:
  1. processing the image through connected layers
  2. analyzing lines
  3. analyzing colors and shapes
  4. producing the final recognition or decision

Further detail

• Deep-learning systems can make mistakes when they have too few examples of a pattern or when the new pattern differs from the data on which they were trained.

💡 Memory Hook

A layered neural network transforms camera inputs into increasingly complex visual decisions.

📖 5. Generative AI and New Content

★ Must-know

📌 A traditional classification system selects a predefined option such as spam or not spam, whereas a generative system creates new content in response to a user prompt.

• 🔄 Generative AI interaction follows: a user prompt, application of a trained model, generation of text, images, audio, or other content

Further detail

• Generative AI can produce: texts, articles and stories, programming code, translations, images, logos, realistic pictures, and artistic paintings, music and sound transformations, videos and animations

• ChatGPT can generate a completely new text in response to a prompt asking for a short story about a robot.

💡 Memory Hook

Traditional systems choose among existing options, whereas generative systems create new text, images, audio, or video.

📖 6. Applications and Classification

📝 Essential Points

📌 Applications are classified according to their primary function: machine learning focuses on learning patterns for prediction, deep learning focuses on complex neural-network processing, and generative AI focuses on creating new content.

• AI-related tasks include:
  • speech recognition
  • image recognition
  • translation
  • recommendations

📖 7. Limitations and Hallucinations

🔑 Key Concepts & Definitions

• Hallucination : an answer produced by an AI system that appears convincing but is invented, completely wrong, or unsupported by reliable evidence.

★ Must-know

• An AI system is a powerful assistant but is not a final source of truth without human review.

Further detail

• AI systems may lack information about current events because their knowledge can stop at a particular date and may not reflect ongoing events.

• AI outputs may reflect bias or errors present in the data on which the system was trained.

💡 Memory Hook

Insufficient or biased training data → inaccurate, biased, or hallucinated output.

📖 8. Responsible Use and Verification

📝 Essential Points

📌 Generative AI should be used as an aid for understanding and thinking, with every line checked, rather than as a substitute for thinking or a complete replacement for writing code.

📌 The main benefit of generative AI is speed in producing large amounts of content, saving time on routine tasks, and supporting brainstorming, while the main risks are hallucinated information, difficulty distinguishing fact from imagination, and weakened critical thinking through total reliance.

• To verify AI output, check the original source, compare the answer with reliable sources such as official websites or school textbooks, test whether it agrees with prior knowledge and is logical, and accept personal responsibility for the information submitted.

💡 Memory Hook

Check sources → compare evidence → apply critical thinking → accept personal responsibility.

📊 Synthesis Tables

AI Concept Hierarchy

Concept	Role	Typical output or function
Artificial intelligence	Broad field	Simulates human intelligence and solves problems
Machine learning	Part of AI	Learns patterns from data for prediction or judgment
Deep learning	Specialized machine learning	Uses neural networks to learn complex patterns
Generative AI	Content-generating AI, commonly based on deep learning	Creates new text, images, audio, or other content