Some AI Terms

AI Applications:

  • Speech Recognition
  • Q&A
  • Automatic Dialog System
  • Text-to-Speech Converter
  • Machine Translation
  • Web Search
  • Spam Filtering,¬†Text Classification
  • Face Recognition
  • Scene Segmentation
  • Image Classification
  • Robots
  • Theorem Provers
  • NASA Fault Diagnosis
  • IBM’s DeepBlue, the chess player
  • Google’s AlphaGo, the GO player
  • Fraud Detection
  • Recommendations

 

Rational Agent: An agent that can act and perceive in a way that maximizes utility.

Reflex Agent: Works on current state of world, uses mostly if-else.

Planning Agent: Asks what-if. Has a model of the world. Captures future.

Search Problems: 
[INPUT]: Abstract world state i.e. search space,
A successor function, updates and calculates
A start and a goal state
[OUTPUT]: Sequence of states, from start to goal state.

State Space Graph: Have set of search states connected with edges having cost or other related entities. No state is repeated. It is exponentially big.

Search Tree: Has start state as root and then each node is explored by the next possible move. No need to build the whole tree. A cycle in the graph means an infinite tree.

Introduction to AI/Computational Rationality

What is AI?
The definition is never definite. From the name, it is->Human-made Intelligence. Then question arises; what is Intelligence? I am going to answer this at the end.

There are four ways to arrive at the output of an AI based agent,
1. Think Like people
2. Act Like People
3. Think Rationally
4. Act Rationally
In the edX course, our objective is to learn the last method.

Rational means achieving your pre-defined goals to fullest by maximizing expected utility. Rationality is the outcome not the thought behind it.

Key Terms:
Act Rationally [That I am gonna read in the course]
MAXIMIZE EXPECTED UTILITY