Introduction to Artificial Intelligence

Introduction to Artificial Intelligence

What is AI?

A - Artificial

I - Intelligence

There are so many definitions around  the world, but basically, AI is concerned with the design of intelligence in an artificial device (an agent)

That means, having artificial or manmade devices and intelligence built into them.

What is Intelligence?

We can take intelligence as something that characterizes human.

Artificial intelligence is having behaviour like a human.

There are 2 ways of how the intelligence is concerned by artificial intelligence.

1. Have a machine or a system that behaves intelligently like a human.

But humans are not always completely intelligent, even there are some intelligent people among us, but they’re also not all the time behave intelligently. So, the 2nd thought is;

2. Machine should behave in the best possible manner, which is the ideal or the most rational intelligence. So, it should be able to overtake the human intelligence.

So, we talked about something “behave”. What sort of behaviour we’re talking about? (two main types)

1. Thinking - reasoning properly to come up with a solution.
2. Acting - how the system actually acts or behave.

So, based on those, can see the different ways of defining AI.

Thought - (processing/reasoning) vs behaviour (act)

Human like performance vs ideal performance.

Example for thinking humanly

Turing Test - Proposed by father of AI : Alan Turing , 1950

• There is a close room it can be a computer (or a human) inside.
• There is an interrogator outside, doesn’t know who is inside. It asks/types questions, the thing inside will process questions and return answers to outside, interrogator receives answers on the screen.
• From the answers, the interrogator outside have to make-out who’s inside the room, a computer or a human?
• If a computer is inside, it tries to claim that it’s actually a human being, in the way it answers to the questions.
• It is the task to the interrogator to decide who’s human what is inside the room.
• 
  
• If the interrogator cannot reliably distinguish the human from the computer, then the computer possesses intelligent which is artificial.
• By turing test, to see whether the machine can come up with the right amount of intelligence to match human intelligence by answering questions.

The following capabilities should be in the computer to pass the turing test.

1. Natural language processing - enable to communicate successfully in English (to understand the question)
2. Knowledge representation - to store what it knows or hears
3. Automated reasoning - to act on its knowledge (to use the stored information to answer questions and to draw new conclusions)
4. Machine learning - to adapt to changing circumstances (to detect and extrapolate patterns)

Turing test deliberately avoided direct physical interaction between the interrogate or and the computer. The questions were sent in written form. (because physical simulation of a person is unnecessary for intelligence)

Total Turing Test

Some of the physical interaction is also included. Ex : interrogation sends a video signal. So then the interrogator can test the subject’s perceptual abilities. Interrogator passes physical objects “through the hatch”

To pass the total turing test, the computer will need,

Computer vision - to perceive objects

Robotics - to manipulate objects and move about.

Example for acting  humanly

Cognitive Science

Expert System

https://www.youtube.com/watch?v=GLCares2LRA

Knowledge Engineering

AI system is composed of an agent and its environment. An agent acts in their environment. That environment can contain other agents too,

What is agent and environment

1. Human agent - has sensory organs : eyes, ears, nose, tongue and skin parallel to the sensors. And other organs : hands, legs, mouth as effectors.
2. Robotic agents - sensors : camera, infrared range finders. Effectors : motors, actuators
3. Software agent - has encoded bit strings as programs and actions

Agent terminology

1. Performance measure - how successful the agent is
2. Behaviour - action that agent perform after any given sequence of percepts.
3. Percept - perceptual inputs at a given instance
4. Percept sequence - history of all it perceived till date.
5. Agent function - where percept sequence maps action

Rational agent

They are intelligent agents

Capable of doing actions to maximize its performance measure, based on:-

Percept sequence

Built in knowledge

Always performs right action

Structure of an intelligent agents

Agent = Architecture + Agent Program

(Intelligence in an artificial device)

Categories of intelligent agents

1. Simple reflex agents

• Choose actions only based on the current percept
• Rational only if a correct decision is made only on based on current percept
• Environment is completely observable

2. Model based reflex agents

• Use a model of the world to choose their actions
• Maintain an internal state
• Model - knowledge about how the things happen in the world
• Internal state - representation of unobserved aspects of current state depending on percept history

3. Goal based agents (problem solving agent)

• Choose their actions to achieve a goal
• More flexible than reflex agent because the knowledge supporting a decision is explicitly modeled, thereby allowing for modifications
• Goal - the description of desirable situations

4. Utility based agents

• Choose actions based on a preference (utility) for each state.
• Goals are inadequate when;
• There are conflicting goals, out of which only few can be achieved
• Goals have some uncertainly of being achieved and you need to weigh likelihood of success against the importance of a goal

Nature of environments

Some program operate in the entirely artificial environment. Confined to keyboard input, database, computer file systems and character output on a screen.

Most famous artificial environment is the Turing Test Environment, where one real and other artificial agents are tested on equal ground.

From this onwards, we are talking about goal based agents, which is problem solving agents.

Searching is the universal technique of problem solving in AI.

Going to define

• Problem and solution
• General purpose of search algorithms and compare them
• Analysis of algorithms, O() big O notation etc
• Uninformed search algorithms (brute force or blind)
• Breadth-first
• Depth-first
• Iterative deepening (ID)
• Repeated states
• Informed search algorithms (heuristic or intelligent)
• best -first
• A* search
• Greedy best-first search

What are these search algorithms for?

To solve a problem, or plan, or play a game you often have to search. So, search algorithms are essential to AI.

Search :-

Algorithm

Performance

So, what is a problem? Or a search problem?

For an example, take “finding a path”, imagine you have to find the shortest path between two locations on a map. Here, you have to “search”.

How to look at a problem?

• To systematically explore the search problem we can use a graph or a search tree.
• Can represent the map using a graph.
• Nodes to represent locations and edges represent paths between nodes
• The estimated travelling time or the distance is often attached to edges
• (can be as a label of an edge)

“Solution is abstract”