Ai Forums Home Welcome Guest    Monday, May 22, 2017
Ai Site > Ai Forums > The Artificial Intelligence Forum > Common Sense and Machine Learning Last PostsLoginRegisterWhy Register
Topic: Common Sense and Machine Learning

DisAlteraVisum
posted 4/10/2017  04:26Send e-mail to userReply with quote
We humans are born without common sense. When I was a father to a toddler, I had to physically stop my son from running out into the street. In a flash, I understood the ramifications of running out into traffic, but my son did not. It would be preferable to have machine learning that produced a program that could predict the likely consequences of actions, but where to start? So I am going to analyze what would likely go into a data set that could create such a program with the aid of machine learning. The first factor is physics, the second is fragility, and the third is value, the fourth is possibility, the fifth is damage to self, and the sixth is likely human activity. In the toddler example, I knew that physics dictated that the car would likely smash my son into a paraplegic or worse due to his relative fragility if he ran out into the street. I knew that he was valuable enough to save. I knew the possibility of my saving him. I knew that I could block him without damaging myself. If I could not save him without damage to myself, then I would likely still save him due to his value being ascertained to be more than mine. But it was a moot point, since I could save him without harming myself. I knew the likely human activity would be for him to continue running out into the street due to his inexperience, and I knew that the driver would likely try to stop; however he would fail due to the physics.

So those are the factors that led me to knock my little boy backwards. Not very nice. But he is still with us.


keghn
posted 4/10/2017  20:57Send e-mail to userReply with quote
Current machine learning is really bad a unsupervised learning.
The automated process of the human brain are the best device for
unsupervised learning,.

Humans first class of primitive goal goal system
are energy management, eating and hunger.
And preservation of self with the help of ant-reward of pain.
And then copying are parents action.
And then acting out different permutations of actions, play.


The human can deal with issues. A issue is where a person predict of a
action and a new position they will be at in the near future and then it
goes terribly wrong. Since kids are very naive, it can happen a lot.

Issues are stored in the brain in the fastest memory. It is not compressed.
It is prestien. All other common non low stress action will be made
out of issue memory patterns, and compressed to save space.


Human second class of goals come into effect when they become adults.
That patterns of copying parents actions are moved into
subconscious reward system. So that if we follow a certain action
it will be rewarding, interesting.



Last edited by keghn @ 4/10/2017 9:18:00 PM

keghn
posted 4/10/2017  21:18Send e-mail to userReply with quote
Child hood issues will become a LINE. Where one side is safe and
the other is not. Alway testing it. Always watching the change borders with a violent neighboring tribe. Or when parent bring home prey and drops in the middle of the pups. The dynamic line is how close i can get with out getting bit. to take care of my hugger.


Last edited by keghn @ 4/10/2017 9:20:00 PM
  1  
'Send Send email to user    Reply with quote Reply with quote    Edit message Edit message

Forums Home    Hal and other child machines    Alan and other chatbots    Language Mind and Consciousness  
Contact Us Terms of Use