Lecture 01: Overview of Artificial Intelligence

The recommended textbook is:

Artificial Intelligence: A Modern Approach by Stuart Russell and Peter Norvig (Fourth Edition).

---

What is AI?

For thousands of years, humans have tried to understand how we think. How can a mere "handful of matter" perceive, understand, predict, and manipulate a world far larger and more complicated than itself?

Artificial Intelligence attempts not just to understand intelligent entities but to build them.

The approaches to AI can be grouped into a few core philosophies:

1. Acting Humanly: The Turing Test Approach

Proposed by Alan Turing in 1950, the Turing Test was designed to provide a satisfactory operational definition of intelligence. A computer passes the test if a human interrogator, after posing written questions, cannot tell whether the responses come from a person or a computer.

To pass the test, a computer would need capabilities like:

• Natural Language Processing (NLP) to communicate.
• Knowledge Representation to store what it knows.
• Automated Reasoning to answer questions and draw conclusions.
• Machine Learning to adapt to new circumstances.

The "Total Turing Test" extends this to interacting with objects in the real world, requiring Computer Vision and Robotics.

2. Thinking Humanly: The Cognitive Modeling Approach

This approach attempts to get inside the actual workings of human minds through:

• Introspection
• Psychological experiments
• Brain imaging

Once we have a precise theory of the mind, we can express it as a computer program. The field of cognitive science brings together AI computer models and psychological experimental techniques.

3. Thinking Rationally: The "Laws of Thought" Approach

Stemming from Aristotle's syllogisms, this approach tries to codify "right thinking." Logicians in the 19th century developed precise notation for statements about objects and relations. However, it is difficult to take informal knowledge (which is often less than 100% certain) and state it in formal terms, and there's a big difference between solving a problem "in principle" and doing it in practice computationally.

4. Acting Rationally: The Rational Agent Approach

A rational agent acts to achieve the best outcome, or when there is uncertainty, the best expected outcome. This approach is more general than the "laws of thought" because rationality includes more than just correct inference. It allows for mathematically defined standards of rationality and is the dominant approach in modern AI.

---

A Brief History of AI

The field of AI has gone through several cycles of immense optimism followed by periods of reduced funding ("AI Winters").

• 1943: Early AI work by Warren McCulloch and Walter Pitts proposing an artificial neuron model.
• 1950: Alan Turing publishes "Computing Machinery and Intelligence," introducing the Turing Test.
• 1956: The famous Dartmouth Workshop, organized by John McCarthy, where the term "Artificial Intelligence" was coined.
• 1969 - 1986 (Expert Systems): Programs like DENDRAL and MYCIN utilized domain-specific knowledge to solve problems, showing the efficiency of knowledge-intensive systems. However, difficulties in scaling and learning led to an "AI Winter."
• 1986 - Present (Neural Networks return): The reinvention of the back-propagation algorithm allowed connectionist models to excel at learning from examples.
• 1987 - Present (Probabilistic Reasoning): Integration of AI with statistics and operations research. Judea Pearl's work on Bayesian networks provided a robust framework for uncertainty.
• 2001 - Present (Big Data): Advances in computing and web technologies led to massive datasets, driving the performance of new learning algorithms.
• 2011 - Present (Deep Learning): Breakthroughs like AlexNet (2012), GANs (2014), ResNets (2015), YOLO (2015), AlphaGo (2016), BERT (2018), GPT-3 (2020), and DALL-E (2022) have marked milestones in complex strategy, NLP, and computer vision.

Today, AI prevails in robotic vehicles, speech recognition, machine translation, and autonomous planning.