Artificial Intelligence - Overview and History

What did Alan Turing contribute to the field of AI?

A means of assessing intelligence based off behavior (the Turing test)

What did Marvin Minsky contribute to the field of AI?

A psychology-oriented approach to AI research and the idea of the perceptron, which single-handedly started an AI winter.

What is the Turing test?

A test proposed by Alan Turing used to operationalize artificial intelligence by evaluating the extent to which a machine could

What did Warren McCulloch and Walter Pitts contribute to the field of AI?

Artificial neurons, which later inspired artificial neural networks and deep learning

What did George Boole contribute to the field of AI?

Boolean algebra and, indirectly through Claude Shannon, the basis for information theory and digital logic.

What did John von Neumann contribute to the field of AI?

Cellular automata and ideas about self-replication of machines

What are some examples of AI that use reasoning and knowledge representation to solve their problems?

DeepBlue RoboCup DARPA Grand Challenge (Self-Driving Car Challenge)

How does engineering contribute to AI?

Engineering produces real-world constraints that guide AI R&D.

What did Gottlob Frege contribute to the field of AI?

First-order logic and the idea that sentences can be expressed as logical proportions with a truth value

What factors most contributed to AI's growth in the 2010s?

Increased access to computational resources and improved methods in machine perception

What did Claude Shannon contribute to the field of AI?

Information theory and the idea of a chess-playing computer, which produced the ideas of look-ahead game-trees and static evaluation functions for board positions

What did John McCarthy contribute to the field of AI?

Its name ("artificial intelligence"), the programming language Lisp,

How do mathematics contribute to AI?

Mathematics help formalize behavior of intelligent systems and provide a basis for engineers to implement artificial intelligence.

What are the foundational issues of AI?

Philosophy, how reasoning should occur; engineering how resource constraints affect design; and mathematics, how intelligence functionally works

How do we handle uncertainty in decision-making?

Probabilistic reasoning,

What is artificial intelligence about?

Problem solving and decision-making

How do psychology and cognitive science contribute to AI?

Psychology and cognitive science help model human behavior, our starting point for building intelligence.

What is intelligence?

The capacity to learn and solve problems

What happened with the AI hype?

The first AI winter (1974-1980) came from Marvin Minsky laying waste to Frank Rosenblatt's perceptrons by revealing they failed at separating non-linearly separable data. This killed interest in the connectionist paradigm for over a decade. The second AI winter (1987-1993) came from significantly reduced funding cuts when researchers gradually discovered how difficult speech recognition and computer vision were when handling edge cases related to common sense reasoning. In the early 21st century, artificial neural networks and "deep learning" boomed in popularity until the field gradually became saturated with methods requiring increasing amounts of computational power to achieve state-of-the-art results.

What does it mean to "think humanly"?

Using human-like intelligence + thought and reasoning

What does it mean to "act humanly"?

Using human-like intelligence to replicate human behavior and actions. This approach seeks to understand and replicate all the errors and biases humans make, which is now the in the field of cognitive neuroscience.

What does it mean to "think rationally"?

Using ideal intelligence/rationality + thought and reasoning. This may not be feasible for intelligent behavior that cannot (yet) be modeled mathematically.

What does it mean to "act rationally?"

Using ideal intelligence/rationality to engage in thought and reasoning

When can exhaustive search be used?

Whenever a problem does not require extensive knowledge, like in: - Pathfinding - The map coloring puzzle

When is knowledge representation required?

Whenever a problem space requires knowledge with context and abstract knowledge, like in deductive reasoning