I am yet to read your subsequent articles. So this comment might be out-of-sync with your complete perspective. Overall what you seem to suggest (focusing on NLU and symbols and explicit representation of properties to differentiate etc.) as far as I can tell, has already been tried by GOFAI.
I fail to understand what's different in the paradigm that you are trying to develop anew?
Ofcourse, as I said, am yet to read your other articles. So may be this is jumping the guns from my side. Thanks for writing!
Thank you for your attention and comments. I am afraid that my explanations are not clear enough. My theory is different from GOFAI - I have not seen anywhere how they explain what symbols are and how to differentiate them or the hierarchy of concepts, they do not explain the core algorithm suitable to any tasks, they do not explain how language uses those mechanisms, they overestimate the role of logic.
I will try to explain those differences better. Do you have good links to the core GOFAI?
Perception is recursively inserted into perception. Symbol after symbol. We use language symbols to organize our lifes and the language organzines us.
That we can learn by words without having to experience something ourselfs, is a very good indicator that language plays a key role in creating AGI systems. By the way, I dont believe english is a good choice for AI ...
"Learning by words" is only efficient if we know what words refer to and it fits better rules than appearances. Try to describe a peacock to a person who never saw them. On the one hand, it is easy to explain a recipe, but on the other, it is hard to describe the taste of that meal without references to "perceptual memories and experiences".
Any natural language is wonderful. English with its order of words is pretty easy to demonstrate some principles, so please don't be too hard on it. :)
I am yet to read your subsequent articles. So this comment might be out-of-sync with your complete perspective. Overall what you seem to suggest (focusing on NLU and symbols and explicit representation of properties to differentiate etc.) as far as I can tell, has already been tried by GOFAI.
I fail to understand what's different in the paradigm that you are trying to develop anew?
Ofcourse, as I said, am yet to read your other articles. So may be this is jumping the guns from my side. Thanks for writing!
Thank you for your attention and comments. I am afraid that my explanations are not clear enough. My theory is different from GOFAI - I have not seen anywhere how they explain what symbols are and how to differentiate them or the hierarchy of concepts, they do not explain the core algorithm suitable to any tasks, they do not explain how language uses those mechanisms, they overestimate the role of logic.
I will try to explain those differences better. Do you have good links to the core GOFAI?
Perception is recursively inserted into perception. Symbol after symbol. We use language symbols to organize our lifes and the language organzines us.
That we can learn by words without having to experience something ourselfs, is a very good indicator that language plays a key role in creating AGI systems. By the way, I dont believe english is a good choice for AI ...
Thank you!
"Learning by words" is only efficient if we know what words refer to and it fits better rules than appearances. Try to describe a peacock to a person who never saw them. On the one hand, it is easy to explain a recipe, but on the other, it is hard to describe the taste of that meal without references to "perceptual memories and experiences".
Any natural language is wonderful. English with its order of words is pretty easy to demonstrate some principles, so please don't be too hard on it. :)