Question

Excuse me if my English is not too great, it is not my native language. I would like to know, what exactly are rational agents and what is their importance to AI. I am familiar with the knowledge of AI research today is focused on the idea of building rational agents. My queries are, how do you define such agents? What are the needed criteria that need to be met or provided to build a 'rational agent'? I would prefer a basic explanation, something a high schooler can understand.

Answer 1

Hello,

It is no question the quest for artificial intelligence, is now mainly driven by the idea of building rational agents. Examples of agents are the Martian rovers, or driver-less cars, such agents that have been designed to operate in 'specific ecosystems'.

Now that we briefly touched the surface, we need to know the formal definition of what we mean by 'rationality' in artificial intelligence. The basic definition can be simply defined as the 'right thing to do', with the maximum pay-off for the agent, considering the information available in the ecosystem. Now the 'right thing' is dependant on what the objectives of the agent are, note to the first paragraph for what agents are, in artificial intelligence. The agent given here is an autonomous entity, that has the ability to adapt to changing situations. A rational agent perceives its ecosystem and acts accordingly to achieve a plus outcome that is rationally the best choice, to maximize the success of its endeavour or the next possible closest plus outcome.

To expand on the prior mentioned, the rationality of an agent needs to be clearly defined by a performance measure (which you can think of as rules) and is necessarily limited by its 'experience' (the sequence of inputs it has from its ecosystem till date) and the data it has at end. Due to this, there is a must to define a clear performance measure for every rational agent, which will define the degree of success it has achieved. Also, one needs to develop a table of possible percepts, and corresponding action sequences for the agent. A percept is a specific input from the ecosystem to the rational agent.

In Artificial Intelligence: A Modern Approach (AIMA) a textbook on Artificial Intelligence, written by Stuart J. Russell and Peter Norvig this is how they define a rational agent in their text on the subject, in which I think exactly captures the essence:

“For each possible percept sequence, a rational agent should select an action that is expected to maximize its performance measure, given the evidence provided by percept sequence and whatever built in knowledge the agent has.”

Peace.

Answer 2

I apologize before hand if there are any spelling or grammatical errors.

Answer 3

What is a plus outcome?

Answer 4

Generally a plus outcome is when full potential is met, or the next possible best thing is met.

Answer 5

I see :)
Thank you very much!

Answer 6

You are welcome.