Four Views of AI

“The science and engineering of making intelligent machines, especially intelligent computer programs”

— John McCarthy

Intelligence: The computational part of the ability to achieve goals in the world.

There are varying kinds and degrees of intelligence in people, animals, and machines.

Four Categories of AI

1. Acting Humanly

Turing Test: Test of machine’s ability to exhibit intelligent behaviour equivalent to a human’s.

Proposed by Alan Turing in 1950

Aside: Human Behavior v.s. Intelligent Behavior, and more limitations

The Turing Test has many weaknesses.

Interrogators may be naive/unsuspecting (e.g., random online comments),
Human behavior isn’t necessarily intelligent (e.g., the first Loebner winner imitated human typing errors),
It only measures external behavior (e.g., are you actually thinking, or simulating thinking),
It may be irrelevant (e.g., aeronautical engineering texts don’t try to make machines that fly exactly like pigeons to fool other pigeons)

Aside: Turing Trap

Turing Trap: A focus on imitation over augmentation of humans means your research is just about replacing humans, which drives down worker wages and income.

This directly removes economic and political power from workers and gives it to people up the food chain.
In the modern day, there are strong incentives to do this to increase corporate profits (layoffs are already the first lever companies pull to please shareholders and keep stocks high)

2. Thinking Humanly

Cognitive Modeling: Validate by predicting and testing behavior of human subjects.¹1. This approach sits at the intersection of psychology, computer science, and linguistics.

Intertwined with scientific theories of how the brain works.
Roughly two approaches: Cognitive Science and Cognitive Neuroscience.

3. Thinking Rationally

Laws of Thoughts: What are correct arguments and thought processes?

It isn’t easy to state informal knowledge, formally.
Solving problems in principle and solving problems in practice are different.

4. Acting Rationally

Note — This is the route we will pursue in this course.

Rational Agents: Something that acts (an agent) doing the right thing (rationality)²2. Rationality here means maximizing expected performance given available information, not achieving perfect or omniscient outcomes..

More general than “thinking rationally”, and doesn’t try to imitate humans.

History of AI

Foundations

Topics	Description
Philosophy	Logic, methods of reasoning, language, rationality
Mathematics	Formal proofs, computation, decidability, probability
Economics	Utility, decision theory, game theory
Neuroscience	Physical substrate for mental activity
Psychology	Perception and motor control, experimental techniques
Computer Engineering	Building fast computers
Control Theory	Designing systems that maximize an objective

Major Events

Year	Description
1943	McCulloch & Pitts: Boolean circuit model of brain
1950	Turing’s “Computing Machinery and Intelligence”
1956	Dartmouth meeting: “Artificial Intelligence” adopted
1952—69	Look, Ma, no hands!
1950s	Early AI programs. Samuel’s checkers program, Newell & Simon’s Logic Theorist, Gelernter’s Geometry Engine
1965	Robinson’s complete algorithm for logical reasoning
1966—73	AI discovers computational complexity Neural network research almost disappears
1969—79	Early development of knowledge-based systems
1980—	AI becomes an industry
1986—	Neural networks return to popularity
1987—	AI becomes a science
1995—	The emergence of intelligent agents
2011—	Availability of very large data sets

Today

Branches of AI:

Logical AI, Search, Pattern recognition, Representation, Inference, Common sense knowledge and reasoning, Learning from experience, Planning, Heuristics, Genetic programming, Robotics, and more.

State-of-the-Art Uses:

Machine learning
Robotics: DARPA Grand/Urban Challenges, Roomba, PackBots
Speech recognition: banking agent, travel agent
Autonomous planning and scheduling: NASA’s autonomous planning programs
Game playing: IBM’s Deep Blue, Google’s AlphaGo
Spam fighting
Machine translation

This approach sits at the intersection of psychology, computer science, and linguistics.↩︎
Rationality here means maximizing expected performance given available information, not achieving perfect or omniscient outcomes.↩︎