1.2 Brief history of AI, Turing Test (BT104CO)

1. The Turing Test (1950)

Proposed by Alan Turing in his paper "Computing Machinery and Intelligence", the Turing Test was designed to provide an operational definition of intelligence. Instead of asking "Can machines think?", Turing suggested asking "Can machines behave like humans?".

A. The "Imitation Game"

A human interrogator communicates via text with two entities: a human and a computer. If the interrogator cannot reliably tell which is which after a series of questions, the computer is said to have passed the test.

Capabilities Required to Pass:
  • Natural Language Processing (NLP): To communicate successfully in a human language.
  • Knowledge Representation: To store what it knows or hears.
  • Automated Reasoning: To use stored information to answer questions and draw new conclusions.
  • Machine Learning: To adapt to new circumstances and detect patterns.

Note: The Total Turing Test also includes a video link to test perception and the ability to move objects (requiring Computer Vision and Robotics).

2. Brief History of AI: Key Eras

The history of AI is often viewed as a cycle of great expectations followed by periods of disappointment known as "AI Winters."

1943–1955

The Gestation Period

McCulloch and Pitts proposed the first model of artificial neurons (1943). Turing published the Turing Test paper (1950).

1956

The Birth of AI (Dartmouth Workshop)

John McCarthy coined the term "Artificial Intelligence". This officially established AI as a distinct field of research.

1952–1969

Early Enthusiasm & Great Expectations

Computers solving algebra problems and proving theorems. Newell and Simon created the General Problem Solver (GPS).

1966–1973

A Dose of Reality (First AI Winter)

Realization that "solving the world" was harder than puzzles. Problems due to combinatorial explosion and lack of power.

1969–1988

Expert Systems

Shift to "domain-specific" AI. Expert Systems (like MYCIN) used "If-Then" rules to mimic human experts.

1986–Present

Return of Neural Networks

Reinvention of Backpropagation allowed multi-layer neural networks to learn (Connectionism).

2011–Present

The Big Data & Deep Learning Era

Massive datasets, powerful GPUs, and Deep Learning breakthrough (ImageNet, LLMs like GPT).

3. Summary of Historical Transitions

Era Primary Focus Methodology
Foundational Can machines calculate? Mathematical Logic / Neurons
Symbolic AI Can machines reason? Search / Logic / Symbols
Knowledge-Based What do machines need to know? Rules / Expert Systems
Modern AI Can machines learn from data? Statistics / Neural Networks

4. Key Exam Concepts

Exam Tip

Emphasize that the Turing Test is a test of behavior, not consciousness. The Dartmouth Conference (1956) is considered the official "birthday" of AI.

Practice Quiz