Zabaware Support Forums

Zabaware Forums => General Discussion => Topic started by: Zaphod on August 23, 2013, 09:45:12 am

Title: Interesting A.I. article in The New Yorker...
Post by: Zaphod on August 23, 2013, 09:45:12 am

Stumbled upon this yesterday:

http://www.newyorker.com/online/blogs/elements/2013/08/why-cant-my-computer-understand-me.html?currentPage=all

Thought y'all might like to read it :)

Title: Re: Interesting A.I. article in The New Yorker...
Post by: NoamI on October 30, 2013, 01:08:08 pm

Thank you, Zaphod,  (and say hi to Slarty..)
  The Winograd Schema test is more scientific than the Turing test.
ex: Sally scolded Mary because she was angry/clumsy.  She is Sally/Mary.
The antecedant of 'she' can't be determined until an association is made:
1. anger may cause one to scold.
2. being clumsy may cause one to be scolded.
Making the association is hard, not because the info isn't findable, but
because recursion (parallelism) is needed.  The brain finds both possible
antecedants and processes the rest of the sentence, and then ranks
the plausibility of the outcomes.  Then it must discard the losing scenarios.
That is hard (messy)!
In many cases a statistical association might find the answer (IBM Watson),
but only processing of all the possibilities will guarantee a correct (human) understanding.