reSolve: The Way to Improve Your Own Skills of Artificial Intelligence Agents

The development of their own skills of artificial intelligence agents is facing two critical problems today. The researchers have developed a new approach that supports oracle to solve these problems. reSolve is designed as a task-based framework and aims to improve the learning process of agents.

Basic problems include:

• Performance cavity: Zero- evolutionized skills perform lower than human-selected skills. In some cases, it gives better results even not to use skills.
• Newly production problem: A successful trajektori recorded during the training becomes difficult to repeat an inventory agent in the distribution phase.
• Oracle-in-the-loop design: reSolve uses a consultant mechanism that provides guidance throughout the process.

This frame makes it possible for agents to learn dynamically not to stand. It is aimed to create more reliable and portable skills by leveraging the power of rational algorithms and human-pay intelligence collaboration.

ReSolve is seen as a promising step towards becoming more effective in the application of AI systems. Future research will test the effectiveness of this method in different tasks.