| CPC G06F 8/40 (2013.01) [G06F 8/31 (2013.01)] | 26 Claims |
|
1. A computer implemented method for generating an artificial intelligence (AI) model from a self-creating script coded in a programming language, the method comprising:
receiving an overall objective for the AI model;
generating a sub-objective for a first state of the AI model, wherein the sub-objective is generated based on the overall objective;
generating a plurality of action code scripts based on a specification of the programming language by:
identifying a plurality of code fragments, wherein the plurality of code fragments are identified from a plurality of candidate code fragments, wherein the plurality of code fragments are code fragments of the plurality of code fragments which compile and execute without warnings or errors,
generating events by combining two or more of the plurality of code fragments, and
generating a plurality of procedures by combining two or more of the events, wherein the plurality of scripts are identified from the plurality of procedures using a reinforcement learning algorithm to identify procedures passing advanced assertions;
assigning rewards to each of the plurality of action code scripts using a reinforcement learning algorithm, wherein a higher reward is associated with achieving the sub-objective;
identifying a preferred action code of the plurality of action codes based on the assigned rewards;
inserting the preferred action code into the self-creating script; and
executing the self-creating script including the preferred action code to satisfy the sub-objective.
|