Alternate World Algorithm

 If we had the time and patience to teach an agent grounded knowledge , knowledge from experience and historical context of concepts, we would have more success in giving it a model of language to work with. We could create a hypothetical world like our own as a simulation, and in this environment it learns how to use language as tools to achieve goals within it. E.g. there would be characters that teach the agent how to relate to objects by naming them and by the agent observing how our language and activities we do when using it correlate. 

 So one character would be seen to pick up a cup when another character says "please pick up that cup" then it correlates the two and realizes what those words mean because of all the correlations it observes. There are problems with this , namely that the characters training the agent are agents in their own right and need that training first also. i propose another solution.

 We create a hypothetical world that is nothing like our real world and place the agent in it. Then the agent explores this world and develops discrete behavior of the sort that would cause it to label each action with a different label that it comes up with by itself , while experiencing this world. So we wait until it has developed enough concepts to do with this world. Once that is achieved we replace each of these concepts with a word from natural language like say English. Now the agent has grounded knowledge of some sort that uses English words to "describe" it. But of course since we arbitrarily assigned the English words to its concepts, a series of actions that this agent performs in its world would produce a sequence of unrelated words. 

 To fix this we use gradient decent to alter the hypothetical world that this agent lives in, till there is a correlation between how it operates in this world and how we use language in our world.Despite the fact that the worlds are totally different. E.g. when the agent performs actions that correlate to "the cup is red"...that phrase will have a totally different meaning to it (as a permutation of actions) but we would have arranged the happy coincident that the agent will be communicating information to us.

Comments

Post a Comment

Popular posts from this blog

ABSTRACT submited

BACKPROPAGATION ON A NOVEL NETWORK (NON ANN) FOR HATESPEECH DETECTION AND MORE This paper introduces a novel type of network that can accurately classify "hate speech" and differentiate it from normal content. It develops a network centered around a unique POS (part of speech) tagging type of system whose specifics are learnt by the network. Each word is assigned variables that indicate the polarity, (sign) of the effect a particular word has on every other word in a sentence, the magnitude of that polarity and also to what extent the word filters itself from the effect from other words. When choosing a set of properties, we would like to properly understand what we want the network to do. We know that both posts classified as hate speech and those not contain samples of words from a common pool (they share words). But we still want that the results for opposite classes to polarize. On top of this I believe that this represents a complex system, in that small chang...
Read more
This involves a "hate speech" or "spam" detection algorithm. (depending on the training data). This will be based on ANN theory , and involve the backpropagation algorithm while not based on an actual neural network. The training data will be labeled into two classes, "hate speech" and "non hate speech".   Hate speech will have an expected value of 1 , representing certain hate speech, while non hate speech will have an expected value of -1.   So what we would like is for the software to be input a post and gives it a value of 1 or -1, or something much closer to one or the other. During training The program will compute a value by performing the equivalent of forward propagation through the parameters we assign to each of multiple structures we will embed in the input ...then using this value and the epected value we will compute a cost function. We will then backpropagate through the heirachical structure of parameters in the direc...
Read more