AI- Artificial Intelligence

The term AI was first coined by John McCarthy in 1956 in Dartmouth conference. As per Wiki-

In 1955 John McCarthy, then a young Assistant Professor of Mathematics at Dartmouth College, decided to organize a group to clarify and develop ideas about thinking machines. He picked the name ‘Artificial Intelligence’ for the new field. He chose the name partly for its neutrality; avoiding a focus on narrow automata theory, and avoiding cybernetics which was heavily focused on analog feedback, as well as him potentially having to accept the assertive Norbert Wiener as guru or having to argue with him

Father of AI

Artificial Intelligence is a fairly involved area, a lot of things come from psychology, philosophy, mathematics, economics, linguistics, biology, and computer engineering. It includes many fields such as a database, information retrieval, statistics, and machine learning.

Types of Learning Systems

Supervised: “right answers” are provided for sufficient training examples. The computer tells “right answers” for new input. Performance measure. (Classification and regression)

Unsupervised: “right answers” are NOT provided and the computer tries to make sense of the data. How good the spread of items is. (clustering and association rule)

Semi-supervised: “right answers” are provided for few training examples only

Active: computer can ask questions. Needs less training. Opposite is passive learning

Lazy: learner do not consolidate the findings. Reinforced: hit and trial method to minimize cost. (game playing)

Transfer: Learning a task B to do A. (cycle riding for bike riding) Deep: processing like the human brain

Success Stories

Deep Blue of IBM In 1997 beat current chess champion Gary Kasparov 1-Kasparov, 2-DeepBlue, 3-draw
AlphaGo of Alphabet Inc. In March 2016, it beat human professional Go player Lee Sedol in a five-game match by 4/1

Waymo: A safer driver that is always alert and never distracted First driverless ride on public roads in 2015 giving a ride to a sole blind In public: 2020
German Traffic Sign Recognition Benchmark (GTSRB) 99.46% against 98.84% of human

Google mapped every single location in France in two hour Images acquired from Google street view
Example of an image search That can taking care of color and pose of the object in the image


In pursuit of computers doing things which at the moment, people do
better, AI attempts to build intelligent entities called Agents 1
Agent perceives the environment through sensors and acts upon the environment through actuators. Our approach is to build a rational agent. How well an agent can behave depends on the nature of environment. Some environments are more difficult. Agents choice of action can depend on percept sequence.

Intelligent Agent
We need a Rational Agent
Rational Agent is one that does the right things with time.
Sequence of actions of agent leads to sequence of states of the
Note: Rationality is not perfection. Rationality maximizes expected
performance whereas perfection maximizes actual performance.

A performance measure could be used to evaluate how the
sequence of state of environment is desirable (NOT of agent)
Design performance measure according to what one actually
wants in the environment, rather then how agent should behave
What is desired is not easy to define (simple life or ups and down)
(every one in moderate poverty or some rich some more poor).

Omniscient vs Autonomous Agents
Omniscient agent knows actual outcome of its actions (this is
An autonomous agent learns to compensate for partial or
incorrect prior knowledge

Agent’s structure can be viewed as −

  • Agent = Architecture + Agent Program
  • Architecture = the machinery that an agent executes on.
  • Agent Program = an implementation of an agent function.

Agent Types

Agents are of basically four types, like Simple Reflex Agent, Model-Based Reflex Agent, Goal-Based and Utility based.

Goal Based Agent

AgentTask environment is described by Performance, Environment,
Actuators, Sensors (PEAS)


The environment could be Fully Observable vs Partially Observable, Single-agent vs Multiagent, Deterministic vs Stochastic, Episodic vs Sequential, Static vs Dynamic, Discrete vs Continuous, Known vs Unknown


A General Learning Agent has Critic to determine how the agent is doing, Learning agent to make rules to improve/adapt, and Problem Generator to suggest experiments under different condition.

Comments 2

Leave a Reply