Introduction to AI

 

Definition of AI

Artificial intelligence is the branch of computer science that deals with the design of computer systems that exhibits some level of intelligence within it. Intelligence is the ability to reason to understand, to create, to learn and to plan.

---

Categories of AI

1) System that thinks humanly
- It is cognitive modeling approach in which the system is associated with the way human thinks such as decision making, problem solving and learning
- eg: general problem solver(gps)
it’s objective is to compare the steps of problem solving with reasoning of human subjects solving the same problem.

2) System that acts humanly
- It is behaviorist approach of creating machine that do things that require intelligence when performed by people
- Turing test approach.

3) System that thinks rationally
- System that can think, perceive and reason through the use of computational models.
- Everything can not be formalized.
- Emphasis on correct inference.

4) System that acts rationally
- System that acts so as to achieve the best possible outcome.
- Emphasis on minimal solution

---

Importance of AI

- Ability to create a never ending thought process that could solve our problems.
- The problems can be solved faster than the human efforts.
- Analysis of large volumes of data and to generate important data’s and patterns

---

AI and Related fields

a) Philosophy ->logic, reasoning, learning.
b) Mathematics -> formal representation and proof ,algorithms
c) Probability-> uncertainty modeling
d) Economics ->utility, decision theory
e) Linguistic =>knowledge representation, grammar
f) Computer engineering ->building fast computers
g) Neuroscience ->neurons as information professor

---

Brief History of AI

- Mcculloch and pits (1943)proposed a Boolean circuit of brain.
- Turing(1950) proposed computing machinery and intelligence
- In 1956, Dartmouth conference was held and AI was officially adopted
- In 1950’s, early AI programs were devised such as samuel’s checker ; newell and simon’s logic theorist.
- Between 1955-1965
a) General problem solver(newell and simon)
b) Geometry theorem proverb(gelertner)
c) Invention of LISP(McCarthy)
- Between 1966-1973 ,ai support funds were cacelled as no any significant results are achieved.
- Between 1969-1985, knowledge based system and rule based expert system are devised eg:BENBRAL MYCIN,etc. but, they did not scale well in practice.
- In 1986, machine learning was implemented
- In 1995, AI was implemented as a science used In vision, language, learning ,reasoning and knowledge representation

---

Application of AI

1) virtual personal assistants
- They help find useful information when you instructs it using your voice.
- It also learns about the user over time and develop ability to anticipate users needs
- E.g.:- siri, Google now,cortema, etc.

---

2) Gaming
- AI results in gaming with characters that learn your behavior, respond to stimuli and react in unpredictable ways.
- Eg: chess, call of duty, etc.

---

3) Fraud detection
- AI is deployed to the system to monitor frauds by providing large samples of fraudulent and non-fraudulent activities and making it learn about the activity based on signs and indications.
- Eg: confirmation mail sent by bank on purchase

---

4) Online customer support
- Websites provide real time chat system to their users to interact with customer support.
- Generally, AI is deployed as automated responder.

---

5) Security surveillance
- With supervised training, security algorithms take i/p from security cameras and determine whether it is a threat or not .
- It may respond to threat by itself with some automatic threat control system or by alerting human security officers.

---

6) Recommendation system
- AI system are also used to generate recommended lists based on interests made in the past
- Eg: video recommendation by YouTube.

---

Turing Test

A machine is said to have passed Turing test if it exhibits following factors:-
a) natural language processing
b) knowledge representation
c) automated reasoning
d) machine learning

---

Turing test is an operational test for intelligent behavior. The interrogator communicates with two sources; human and machine. Within a time frame, he must decide which source is human and which source is machine. If he is wrong half the time then the machine is said to be intelligent.