Artificial Superintelligence: The Recursive Self-Improvement in NLP

EasyChair Preprint no. 2860

Abstract

Recursive self-improving( RSI ) systems create new software iteratively. The newly created software iteratively generates a greater intelligent system using the current system, then this process leads to phenomenon referred to as superintelligence. In this paper, we provide a formal definition of RSI systems and then we present a recursive self-improving model which takes a program as an argument and return a suggested improvement of the given program. With this concept, we have built a recursive neural network(RNN) which is a word vectorization useful for natural language processing( NLP ) and also which can be accomplished with a recursive algorithm known as Word2vec. We have converted a corpus of words into vectors by both recursive neural network using parse tree & recursive Word2vec method without tree structure which are then thrown into vector space to structure text data hierarchically for outputting parse trees for a sentence and measure the cosine distance between word-context pairs i.e. their similarity or lack of, and the test results are encouraging with a possibility of more comprehensible recursive self-improvement. 

Keyphrases: artificial superintelligence, Recursive Neural Networks, Recursive Self-Improvement, Recursive Word Vectors

@Booklet{EasyChair:2860,
  author = {Poondru Prithvinath Reddy},
  title = {Artificial  Superintelligence: The Recursive Self-Improvement in NLP},
  howpublished = {EasyChair Preprint no. 2860},

  year = {EasyChair, 2020}}