The fun in working with Robots is that you get to play “GOD”. Atleast … psued(o) GOD

Me: Hey Robot .. you go there .. Now there .. Then there ..
Robot: Am i being supervised? Dont i have freedom, a choice . Can’t i get to choose what i should do?
Me: Ok. If thats what you want … By all means !
Robot: By the way .. why am i alive … why do i exist in first place?
Me: Every Robot is created for a purpose. You have a purpose .. You have a “GOAL”
Robot: Oh! Really .. Wow .. Thats Awesome 🙂 … And what is my purpose.. my “GOAL” ?
Me: You will know when you reach it !
Robot: #%#%?
Me: You choose your actions .. not me … I don’t control you!

After few iterations. …

Robot:I don’t seem to have reached my GOAL .. I haven’t received that BIG reward. Any tips for choosing my actions?
Me: The problem with you is that you have been completely GREEDY … The classical algorithm might to be greedy about your actions to reach your GOAL . You are always exploiting … You have to explore too at least for epsilon time .. You need to develop a policy to reach your GOAL .. and preferably an optimal one.
Robot: Without having any knowledge of my environment, how to reach my GOAL and that too optimally?
Me: Refer “Reinforcement Learning:An Introduction by Richard S. Sutton and Andrew G. Barto”
Robot: Oh .. Is this subject understandable …
Me: Sort of … YES !

After a lot of iterations …….

Robot: I haven’t still reached my GOAL! I have been moving all around the environment .. hitting obstacles after obstacles ….. Where the **** is my GOAL? [ **** = hell ]
Me: To find your GOAL is your purpose .. That’s is the reason of your existence!
Robot: More over ..at each state .. i am getting more and more negative rewards ..
Me: No Pain .. No Gain 😛 …
Robot:Cant you give some positive reward ? don’t you realize .. negative reward hurts …. We robots also have feelings …
Me: Well … am just helping you to keep away from unwanted loops and not miss out on your GOAL.
Robot: When will i reach my GOAL?
Me: Well … I don’t know .. Few Robots reach soon … Few take a long time … But i can assure you that you will reach your GOAL after infinite iterations.
Robot: %$#$^
Me: Don’t worry .. You see you are learning at each step.

After some more iterations …

Robot: Why are you so concerned for me to reach the GOAL?
Me: Just reach your GOAL …. And please do it faster … I need results … I have to plot graphs .. You see … I need to submit a paper and deadline is fast approaching!
Robot: Aren’t there faster means by which i can reach my GOAL?
Me: Yes … “Transfer Learning Community” is working in that area … You can learn from other Robots if they have solved some related problem you are working on …
Robot: Ok … But … I am different … I am not like other robots …. My actions are different … I may not be able to replicate whatever actions other Robots have done …
Me: Oh .. yea … Am exactly working on that problem ….

I sort of have a feeling if this also could be conversation between me and GOD .. where i am the Robot 😛

Advertisements

The fun in working with Robots is that you get to play “GOD”. Atleast … psued(o) GOD

Me: Hey Robot .. you go there .. Now there .. Then there ..
Robot: Am i being supervised? Dont i have freedom, a choice . Can’t i get to choose what i should do?
Me: Ok. If thats what you want … By all means !
Robot: By the way .. why am i alive … why do i exist in first place?
Me: Every Robot is created for a purpose. You have a purpose .. You have a “GOAL”
Robot: Oh! Really .. Wow .. Thats Awesome 🙂 … And what is my purpose.. my “GOAL” ?
Me: You will know when you reach it !
Robot: #%#%?
Me: You choose your actions .. not me … I don’t control you!

After few iterations. …

Robot:I don’t seem to have reached my GOAL .. I haven’t received that BIG reward. Any tips for choosing my actions?
Me: The problem with you is that you have been completely GREEDY … The classical algorithm might to be greedy about your actions to reach your GOAL . You are always exploiting … You have to explore too at least for epsilon time .. You need to develop a policy to reach your GOAL .. and preferably an optimal one.
Robot: Without having any knowledge of my environment, how to reach my GOAL and that too optimally?
Me: Refer “Reinforcement Learning:An Introduction by Richard S. Sutton and Andrew G. Barto”
Robot: Oh .. Is this subject understandable …
Me: Sort of … YES !

After a lot of iterations …….

Robot: I haven’t still reached my GOAL! I have been moving all around the environment .. hitting obstacles after obstacles ….. Where the **** is my GOAL? [ **** = hell ]
Me: To find your GOAL is your purpose .. That’s is the reason of your existence!
Robot: More over ..at each state .. i am getting more and more negative rewards ..
Me: No Pain .. No Gain 😛 …
Robot:Cant you give some positive reward ? don’t you realize .. negative reward hurts …. We robots also have feelings …
Me: Well … am just helping you to keep away from unwanted loops and not miss out on your GOAL.
Robot: When will i reach my GOAL?
Me: Well … I don’t know .. Few Robots reach soon … Few take a long time … But i can assure you that you will reach your GOAL after infinite iterations.
Robot: %$#$^
Me: Don’t worry .. You see you are learning at each step.

After some more iterations …

Robot: Why are you so concerned for me to reach the GOAL?
Me: Just reach your GOAL …. And please do it faster … I need results … I have to plot graphs .. You see … I need to submit a paper and deadline is fast approaching!
Robot: Aren’t there faster means by which i can reach my GOAL?
Me: Yes … “Transfer Learning Community” is working in that area … You can learn from other Robots if they have solved some related problem you are working on …
Robot: Ok … But … I am different … I am not like other robots …. My actions are different … I may not be able to replicate whatever actions other Robots have done …
Me: Oh .. yea … Am exactly working on that problem ….

I sort of have a feeling if this also could be conversation between me and GOD .. where i am the Robot 😛