Introduction to AI by nvmoyar https://padlet.com/nvmoyar/w0l5ur4ggoc4 Contextualise the problem en-us 2017-11-13 19:13:38 UTC 2020-07-15 13:44:36 UTC hello@padlet.com Definition of Intelligence nvmoyar https://padlet.com/nvmoyar/w0l5ur4ggoc4/wish/206438018 Intelligence can be defined in the context of the task. 

DEFINITION OF INTELLIGENCE: an intelligent agent is the one that takes actions to maximize its expected utility given a desired goal. This definition attends to RATIONAL BEHAVIOUR. This definition requires agent to behave optimally, however there are constrictions in real life, like limited computation resources, costs, rules that can be applied like deadlines, etc. OPTIMALLY BEHAVE: Given those constrictions we can't expect an agent to behave always optimally, but we can come up with a level of performance  or bound that we desire the agent to meet: example win 60%, find a route no 2 miles longer than the optimal route. This is called BOUNDED OPTIMALITY and it's a feasible way of quantifying the intelligence. ]]> 2017-11-13 19:15:11 UTC https://padlet.com/nvmoyar/w0l5ur4ggoc4/wish/206438018 Defining your problem nvmoyar https://padlet.com/nvmoyar/w0l5ur4ggoc4/wish/206438641 how would you characterise the task of driving a car in a road? 


]]> 2017-11-13 19:16:03 UTC https://padlet.com/nvmoyar/w0l5ur4ggoc4/wish/206438641 Fully observable / Partial observable nvmoyar https://padlet.com/nvmoyar/w0l5ur4ggoc4/wish/206438879 Like tic-tac-toe or chess vs Battleship board


]]> 2017-11-13 19:16:23 UTC https://padlet.com/nvmoyar/w0l5ur4ggoc4/wish/206438879 Deterministic / Stochastic nvmoyar https://padlet.com/nvmoyar/w0l5ur4ggoc4/wish/206439443 Do you know the results of your actions or there is some uncertainty or randomness]]> 2017-11-13 19:17:11 UTC https://padlet.com/nvmoyar/w0l5ur4ggoc4/wish/206439443 Continous / Discrete nvmoyar https://padlet.com/nvmoyar/w0l5ur4ggoc4/wish/206440159 Continuous like the real world where the number of states are infinite, or there is only a number of states possible]]> 2017-11-13 19:18:11 UTC https://padlet.com/nvmoyar/w0l5ur4ggoc4/wish/206440159 Benign / Adversarial nvmoyar https://padlet.com/nvmoyar/w0l5ur4ggoc4/wish/206440818 The agent can be the only one taking actions that intentionally affect its goal or adversarial, which means that it's competing against other agents]]> 2017-11-13 19:19:05 UTC https://padlet.com/nvmoyar/w0l5ur4ggoc4/wish/206440818 Driving a car in a road problem nvmoyar https://padlet.com/nvmoyar/w0l5ur4ggoc4/wish/206462716 Problem of driving a car on the road is:
  •  partially observable (consider "blind spots", driving at night, fog, etc.), 
  • stochastic (unpredictable events) and 
  • continuous (you're interacting with the physical world).
  • It can be adversarial too - depending on where you are driving!
]]> 2017-11-13 19:55:34 UTC https://padlet.com/nvmoyar/w0l5ur4ggoc4/wish/206462716 GAMES nvmoyar https://padlet.com/nvmoyar/w0l5ur4ggoc4/wish/206465155 Games are based on skill or chance (stochastic, radomness). Chess for instance is a game of skill, since all the relevant elements of the game are fully observable on the board. Since the board is finite, it is not continuous either. Tic tac toe is a deterministic game as well. ]]> 2017-11-13 20:00:25 UTC https://padlet.com/nvmoyar/w0l5ur4ggoc4/wish/206465155 Intelligence can be defined in the context of the task nvmoyar https://padlet.com/nvmoyar/w0l5ur4ggoc4/wish/655106414 * Is it a classification problem?
* is this task something that happen in the real world?
* do we know the result of our actions?
* are we the only element affected by the result of our actions?

An intelligent agent is the one that takes actions to maximize its expected utility given a desired goal. Must be notice that this definition attends to a RATIONAL BEHAVIOUR. This definition requires agent to behave optimally, however there are constrictions in real life, like limited computation resources, costs, rules that can be applied like deadlines, etc. OPTIMALLY BEHAVE: Given those constrictions we can't expect an agent to behave always optimally, but we can come up with a level of performance  or bound that we desire the agent to meet: example win 60%, find a route no 2 miles longer than the optimal route. This is called BOUNDED OPTIMALITY and it's a feasible way of quantifying the intelligence.

HOW WOULD YOU CLASSIFY THE TASK OF DRIVING A CAR IN A ROAD?? ]]> 2020-07-15 13:39:35 UTC https://padlet.com/nvmoyar/w0l5ur4ggoc4/wish/655106414