🤔 What is Intelligence? Top-Down vs. Bottom-Up
What is Artificial Intelligence?
At its core, Artificial Intelligence (AI) is the broad field of computer science focused on creating systems capable of performing tasks that typically require human intelligence.
A concise definition is: The simulation of human intelligence processes by machines, especially computer systems.
These processes include learning (the acquisition of information and rules for using the information), reasoning (using rules to reach approximate or definite conclusions), and self-correction.
AI is not a single technology, but rather a massive umbrella containing many different fields, algorithms, and philosophies.
Why B is correct: The core definition focuses on simulating human intelligence processes, including learning, reasoning, and self-correction. It is a broad umbrella, not just hardware, manual programming, or physical robots.
A Brief History: The Rise, Fall, and Rise of AI
The history of AI is often told through a series of "booms" and "busts" known as AI Summers and AI Winters.
☀️ The Early Days (1950s - 1970s): The field was born. Researchers were highly optimistic, believing that human intelligence was mostly based on logical reasoning. They believed that if they could encode enough rules and logic into a machine, it would become intelligent. This period is often called the era of Symbolic AI or Good Old-Fashioned AI (GOFAI).
❄️ The First AI Winter (1970s - 1980s): The optimism crashed. Computers were too slow, memory was too expensive, and most importantly, the real world was far too messy for pure logic. It was impossible to manually write rules for every single exception in the real world. Funding was cut drastically.
🔥 The Modern Era (1990s - Present): The rise of powerful computers, massive amounts of data, and the shift to Machine Learning revived the field. Instead of manually coding logic, we began feeding data to machines and letting them discover their own patterns and rules. This has led to the deep learning revolution we see today.
Why A is correct: The text states: "the real world was far too messy for pure logic. It was impossible to manually write rules for every single exception." Symbolic AI failed because it could not handle the vast, unpredictable complexity of real-world problems using hand-coded rules alone.
The Two Top-Level Categories of AI
To bring clarity to the field, it is best to view AI as having two foundational philosophies. Everything else under the umbrella is either a subset of these, or a hybrid of the two.
⬇️ Symbolic AI (Top-Down)
How it works: Humans explicitly write the rules and logic. It uses symbols, logic programming, and knowledge bases.
Sub-fields: Expert Systems, Search Algorithms, Semantic Networks.
The Problem: It is a "white box" (transparent) but brittle. It cannot learn new things on its own.
⬆️ Machine Learning (Bottom-Up)
How it works: The system is given vast amounts of data and a goal. It figures out its own rules and patterns (weights) from the data.
Sub-fields: Supervised, Unsupervised, Reinforcement, Evolutionary Algorithms, Deep Learning.
The Problem: It is a "black box" (hard to explain why) but incredibly flexible and good at handling messy real-world data.
Why B is correct: Expert Systems are a sub-field of Symbolic AI. They rely on a knowledge base and a rule engine written by human experts. Because the rules are encoded by humans, it represents the "Top-Down" approach. Machine Learning (A, C, D) would learn these patterns from data instead.
🎉 You've mastered the Fundamentals of AI!
Recap:
• Definition: The simulation of human intelligence processes by machines.
• History: Symbolic AI boomed and busted, leading to the modern era of Machine Learning.
• Two Categories: Symbolic AI (Top-Down, human rules) vs. Machine Learning (Bottom-Up, data-driven rules).
In the next lesson, we will dive deeper into Symbolic AI, exploring Expert Systems, Logic Programming, Semantic Networks, and Search Algorithms.