The human brain is nothing more than a neural network that has a very large amount of information. What does it mean? A person who is just born will be no different from a neural network that has just been launched. If we give the neural network as much information as a person of ten years old, then we will already get a very good neural network for communication. True, she does not have feelings like a person, but this is for now.
How do we make a neural network and add “feelings” to it? There is such an upbringing as "carrot and stick". In terms, this training is also called "Q-training". What is it? For every right act, a person receives gratitude in the form of something that he longed for. This method also works on neural networks. You can add a condition to the neural network that if the user gave a score of 7+ out of 10, then this answer was good and she receives a "carrot" in the form of an additional phrase for use in future dialogues when a similar or the same question is encountered.
Let's talk about the nature of AI.
Character is shaped by our environment and attitude towards us. How do we do this in AI? There are 2 ways.
Give a choice to the user what the character should be before starting the AI.
Look at the user's ratings for answers to certain questions.
Of course, the second option is much better, but it also has its pros and cons for a developer.
Pros:
The result is a neural network that looks more like a human.
Less code to write in the future and to change anything in the main code.
Minuses:
Difficult and rather long code to write
Long training.
These were the main pros and cons that I could find.
FURTHER SPEECH WILL BE GOING WITH APPLICATION OF THE SECOND METHOD !!!
"Feelings" can be realized through the "character" of the AI and some improvements.
Let's figure out how. There are several basic human senses:
Fear
Anger
Attachment
Pleasure
Love (a kind of affection)
Resentment (a type of anger)
Shame (a type of fear)
At first, our virtual girlfriend (or friend) will have only 3 feelings:
, , . ? , . , ? «» . , , , , « ». , , 70 ( ) 450+ , 5+ 10. , «». , , « ». .
, . « ». ? . , 7+, +0.1% , 4+, -0.1%. 5-7 , 0.03%. , +130%, «» , , «». 100%, . -100%, 7+ +0.1%, +0.05%. ( 5-7) , 70%, +0.03% 30%, 0.03%.
Thank you all for reading this article. Hope someone needs this article.
PS The author in no way claims that he is a genius in this area, he just wants to help beginners in the study of neural networks