The AI Problems – The types of problems and early work done in AI
The early work in AI was done in common sense analysis. Imagine if you are holding a phone with screen facing towards floor, you would not like the phone to hit the floor, as you know it may break the screen. This knowledge that the phone would break if it hit the harder surface, you learnt it not by doing experiments but probably by drawing useful analysis of relationships between two objects (the relationship between phones screen and floor). This useful analysis between different physical objects is termed as common-sense reasoning. The early work on AI was done on solving the problems which have common sense reasoning. General Problem Solver developed by Shaw and Newell was applied to these common-sense problems also.
Second part of AI Problems were concerned game playing, AI Programs were developed which could play games such as checkers, chess etc. The good part was these AI programs learnt from their previous games and got better as they played more. It became difficult to defeat a computer program who has played good number of games. Just like you, you would consider yourself a better player if you have played multiple games of cricket.
We generally say a person is intelligent if that person is good in geometry or in strategy games, like chess or checkers. It was thought that if we can create a computer which can process the the rules of these quickly, we would be able to create an Artificial Intelligent Computer. Let’s understand this way – They thought since only few rules are needed to play a play game or solve theorem – It’s like if you know the rules of cricket – like – an over has six balls, the rules of no ball, wide ball, leg byes , you would be an amazing player of cricket. If a computer can process these rules quickly, will be considered intelligent – They were all proved wrong because of the combinational explosion of games—Like you don’t win any chess game with the same scenario. So, need of handling large data bases was required.
How to create Artificial Intelliegent system?
Imagine yourself as baby few moths old, how did you learn about the world , probably the following way
The learning process –
- Someone taught you to see – Recognize objects–>then you learnt to speak–> Then you learnt to understand a language – respond based on understanding –>Which developed Common Sense Reasoning. In AI Computer Systems develpoing a system which can understand language said to be Natural Language Understanding systems| These all were known as ROUTINE or MUNDANE TASKS. (You learnt if you throw a phone on phone from a significant height it may break).
- Once you learnt all the basic things – Called Routine things or Mundane Tasks –>You decided to learn a game, a subject and developed interest in game or studies –> You leant how to apply logic. –> These were known as FORMAL TASKS. (You developed some interest in a game, subject or adventure or anything which drove you).
- Once you learnt the formal tasks –>Once interest was achieved–>With regular practice you became the master or expert of the same, Just like any famous neighbourhood doctor, a cricket player, painter etc.–> These were termed as EXPERT TASKS (You started learning from your own mistakes and more you learnt more closure you become to be an expert. DENDRAL and MYCIN were famous AI Expert Systems.
So roadmap is : Learn Routine or Mundane Tasks–>Practice a particular routine task –> and become expert, This was the same process followed for developing an intelligent computer system.