Quiz 2

🧠 History & Philosophy of AI

1818 words
9 min read
Python Week 1: the first filter for runtime behavior
Visual companion
Python
Type and operator map

Python Week 1: the first filter for runtime behavior

Open galleryRelated surface
View
Revision summary

What this note is really saying

Short form

# 🧠 History & Philosophy of AI ## 1. 🎯 Learning Objectives By the end of this topic, you will be able to: - Trace the historical evolution of AI from philosophical roots through modern deep learning - Explain the Dartmouth Conference (1956) and why it marks the birth of AI as a field - Define the combinatorial exp...

🧠 History & Philosophy of AI

1. 🎯 Learning Objectives

By the end of this topic, you will be able to:
  • Trace the historical evolution of AI from philosophical roots through modern deep learning
  • Explain the Dartmouth Conference (1956) and why it marks the birth of AI as a field
  • Define the combinatorial explosion problem and explain why it is the central challenge of AI
  • Distinguish between symbolic (logic-based) reasoning and machine learning (data-driven) approaches
  • Characterize intelligent agents along four dimensions: persistence, autonomy, proactivity, goal-directedness
  • Compare model-based reasoning vs. memory-based reasoning with concrete examples
  • Articulate Turing's vision and the Turing Test's role in AI philosophy

2. 📋 Prerequisites

PrerequisiteCourseWhy It Matters
Basic graph theoryBSCS2001Combinatorial explosion is fundamentally about graph growth
Propositional logicBSCS2002Symbolic AI relies on logical inference
Probability basicsBSMA1002ML approaches use probabilistic reasoning
Algorithmic complexityBSCS2002Understanding exponential vs. polynomial growth

3. 📖 Core Content

3.1 Intuition: What Is AI, Really?

Artificial Intelligence is the field of study concerned with making machines do things that would require intelligence if done by humans. In this course, intelligence is the ability to search through possibilities to find solutions. Imagine you are in a vast maze with billions of possible paths. An intelligent agent does not try every single path randomly — it uses strategies to find the exit efficiently. That is what AI search methods are: strategies for navigating impossibly large spaces of possibilities. The central problem that motivates everything in this course is combinatorial explosion: the phenomenon where the number of possibilities grows so fast (exponentially or factorially) that brute-force enumeration becomes impossible, even with the world's fastest computers.
Why This Matters: Combinatorial explosion is the single most important concept in AI. Every algorithm in this course — from DFS to A* to Genetic Algorithms — is designed to cope with it. Without combinatorial explosion, we would not need AI at all; we could just check all possibilities.

3.2 The Pre-History: Philosophical Roots (Aristotle to Turing)

The dream of mechanical reasoning dates back to ancient philosophy:
  • Aristotle (384–322 BCE): Developed syllogistic logic — the first formal system for reasoning. A syllogism is a pattern of reasoning where a conclusion follows from two premises (e.g., "All men are mortal; Socrates is a man; therefore Socrates is mortal"). This established the idea that reasoning can be reduced to formal rules.
  • Leibniz (1646–1716): Dreamed of a characteristica universalis — a universal language of symbols for all human thought — and a calculus ratiocinator — a mechanical device that could compute truth from these symbols. This is essentially the vision of AI 300 years before computers existed.
  • Babbage & Lovelace (1800s): Charles Babbage designed the Analytical Engine, a mechanical general-purpose computer. Ada Lovelace wrote the first algorithm intended for machine execution and speculated that machines might one day "compose elaborate and scientific pieces of music."
  • Turing (1936–1950): Alan Turing's 1936 paper "On Computable Numbers" introduced the Turing Machine — a mathematical model of computation. In 1950, his paper "Computing Machinery and Intelligence" asked "Can machines think?" and proposed the Turing Test: a machine is intelligent if a human interrogator cannot distinguish its responses from a human's. The Turing Test works as follows:
  1. A human interrogator communicates with two entities via text-only channel
  2. One entity is human, one is a machine
  3. The interrogator must determine which is which
  4. If the machine can fool the interrogator a significant fraction of the time, it is considered intelligent The Turing Test is controversial — is imitation truly intelligence? But it established the benchmark for AI research.

3.3 The Dartmouth Conference (1956): The Birth of AI

In the summer of 1956, John McCarthy (Dartmouth), Marvin Minsky (Harvard), Nathaniel Rochester (IBM), and Claude Shannon (Bell Labs) organized the Dartmouth Summer Research Project on Artificial Intelligence. The proposal read: "The study is to proceed on the basis of the conjecture that every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it." This was the first use of the term "Artificial Intelligence" as a field name. Key outcomes:
  • Demonstration of the Logic Theorist (Newell and Simon) — a program that could prove mathematical theorems
  • Introduction of heuristics — rules of thumb that guide search
  • Establishment of AI as an academic discipline

3.4 The Central Problem: Combinatorial Explosion

Let us understand combinatorial explosion with three concrete examples: Example 1: The Chessboard The game of chess has approximately 1012010^{120} possible games. By comparison, the observable universe contains about 108010^{80} atoms. If every atom were a supercomputer performing 101010^{10} calculations per second, it would take more than the age of the universe to enumerate all chess games. Example 2: The Traveling Salesman Problem (TSP) A salesman must visit N cities exactly once and return to the start. The number of possible tours is (N−1)!/2(N-1)!/2 (for symmetric TSP). For N=10: 181,440 tours. For N=20: ≈ 101610^{16} tours. For N=50: ≈ 106210^{62} tours — more than the number of atoms in the universe. Example 3: The 8-Puzzle The 8-puzzle has a state space of 9!/2=181,4409!/2 = 181,440 reachable states (half of 362,880 permutations are unreachable due to parity constraints). For the 15-puzzle: about 101310^{13} states. Brute force becomes infeasible very quickly. (Diagram)

3.5 Intelligent Agents

An intelligent agent is any entity that perceives its environment through sensors and acts upon that environment through effectors. Four Key Characteristics:
CharacteristicMeaningExample
PersistentOperates over time, not one-shotA chess program that plays an entire game
AutonomousMakes its own decisions within its designA self-driving car navigating traffic
ProactiveTakes initiative toward goalsA planning agent rearranging blocks
Goal-directedHas explicit objectives to achieveA TSP solver minimizing total distance
(Diagram)

3.6 Symbolic AI vs. Machine Learning

AspectSymbolic AI (GOFAI)Machine Learning
Core ideaIntelligence = manipulation of symbols using logicIntelligence = pattern recognition from data
Knowledge sourceHand-coded rulesLearned from examples
Reasoning styleDeductive (logic-driven)Inductive (data-driven)
StrengthExplainable, verifiable, works with little dataHandles perception, uncertainty, messy data
WeaknessFragile, does not scale to perception tasksRequires large datasets, black-box
ExampleExpert systems, theorem proversNeural networks, SVMs
Search roleState space search is centralSearch appears in optimization

3.7 Model-Based vs. Memory-Based Reasoning

Model-Based Reasoning: The agent builds an explicit model of the world (rules, constraints, physics). Reasoning is done by simulating the model forward or backward. Example: A chess program that evaluates positions using explicit material and positional rules. Memory-Based Reasoning (Case-Based): The agent stores past experiences (cases) and retrieves similar ones when facing new problems. No explicit model — just data. Example: A medical diagnosis system that finds similar patients from a database.
AspectModel-BasedMemory-Based
Knowledge representationRules, equations, logicStored cases/examples
InferenceDeduction from modelAnalogical matching
Data neededLowHigh
ExplainabilityHighModerate

4. 📐 Key Formulas / Concepts

ConceptDescriptionFormula
Combinatorial explosionGrowth rate that exceeds enumeration capacityO(bd)O(b^d) for branching factor bb , depth dd
State space sizeNumber of reachable configurationsVaries: N!N! permutations, bdb^d nodes
Branching factorAverage number of children per nodebb in search trees
Turing TestMachine indistinguishable from humanPass rate > 30% (arbitrary threshold)
AgentPerceive -> Reason -> Act cyclef:P∗→Af: P^* \rightarrow A

5. ⚠️ Common Pitfalls

Pitfall 1: Confusing AI with Machine Learning

The mistake: Treating "AI" as synonymous with "machine learning." Why it happens: Media coverage overwhelmingly discusses ML/DL. Correct approach: AI is the broader field; ML is one approach within it.

Pitfall 2: Underestimating Combinatorial Explosion

The mistake: Thinking "we will just enumerate all possibilities with a faster computer." Why it happens: Linear thinking — if a computer checks 1M states/second, a billion states seems reachable. Correct approach: Compute bdb^d for a realistic problem; even b=10,d=20b=10, d=20 gives 102010^{20} states.

Pitfall 3: Confusing the Turing Test with True Intelligence

The mistake: A program that passes the Turing Test is truly intelligent/sentient. Why it happens: The test sounds like a definitive benchmark. Correct approach: The Turing Test measures human-likeness of behavior, not consciousness.

6. 📝 Practice Questions

Q1: What was the significance of the Dartmouth Conference (1956)?
Answer: The Dartmouth Conference was the first formal gathering dedicated to AI as a field. It coined the term "Artificial Intelligence," brought together the founders (McCarthy, Minsky, Rochester, Shannon), and established AI as an academic discipline. Q2: Explain why combinatorial explosion is called the "central problem" of AI.
Answer: Combinatorial explosion refers to the phenomenon where the number of possible states grows exponentially or factorially with problem size. For any non-trivial problem (chess with 1012010^{120} states, TSP with (N−1)!/2(N-1)!/2 tours), exhaustive enumeration is impossible. Every AI search algorithm is designed to cope with combinatorial explosion. Q3: A TSP with 15 cities has how many possible tours? At 1 billion tours/second, is brute force feasible?
Answer: Number of tours = (15−1)!/2=43,589,145,600(15-1)!/2 = 43,589,145,600. At 1 billion/second, this takes about 43.6 seconds. For 20 cities: 19!/2≈6.0×101619!/2 \approx 6.0 \times 10^{16} tours -> 1.9 years. For 25 cities: 24!/2≈3.1×102324!/2 \approx 3.1 \times 10^{23} tours -> 9.8 million years. This demonstrates combinatorial explosion. Q4: Distinguish between symbolic AI and machine learning approaches.
Answer: Symbolic AI represents knowledge as explicit symbols manipulated through logic. It is explainable, requires little data, but is fragile. Machine learning learns patterns from data without explicit rules. It excels at perception and handling uncertainty but requires large datasets. Q5: List and explain the four characteristics of an intelligent agent.
Answer: (1) Persistent — operates over time; (2) Autonomous — makes own decisions; (3) Proactive — takes initiative toward goals; (4) Goal-directed — has explicit objectives. Q6: Would the 8-puzzle be solvable by brute force? What about the 15-puzzle?
Answer: The 8-puzzle with 181,440 states is easily brute-forceable. The 15-puzzle has about 101310^{13} states — manageable with good algorithms. The 24-puzzle has 102510^{25} states — infeasible for brute force. Q7: Explain model-based vs. memory-based reasoning with examples.
Answer: Model-based uses explicit rules/equations to simulate (e.g., physics engine). Memory-based stores and retrieves past cases (e.g., recommendation system). Q8: What is the Turing Test and what are its limitations?
Answer: The Turing Test proposes that a machine is intelligent if a human cannot distinguish its responses from a human's. Limitations: tests behavior not consciousness; simple programs like ELIZA can be deceptive; does not test vision, creativity, or physical interaction. Q9: How did Leibniz's vision anticipate modern AI?
Answer: Leibniz proposed a universal symbol language and a mechanical reasoning device, prefiguring symbolic AI where intelligence is computation over symbols. Q10: Why did the pendulum swing from symbolic AI to machine learning?
Answer: Symbolic AI could not handle perception, uncertainty, or messy real-world data. The rise of statistical methods, increased computation, and large datasets drove the shift to machine learning.

7. 🔗 Cross-References

Document outline

Keep your place and jump directly to a heading.

Table of Contents
System Normal // Awaiting Context

Intelligence Hub

Navigate the knowledge graph to generate context. The Hub adapts dynamically to surface backlinks, related notes, and metadata insights.

Open library →Launch Assessment Engine