In a recent development, a new intelligent system has been designed by researchers at the Australian National University. The system based on deep learning can learn to provide solutions for decision-making problems, which include problems that are more complex than what it has been trained to crack.
Deep intelligence is a favored artificial intelligence technique for pursuits such as making captions to describe images, learning to play board games or video solely from images, and transcribing speech to text.
New Intelligent System a Game Changer for Problem Solving
Action Schema Networks, as the system is called could hypothetically allow a robot for learning how to navigate a floor with 10 rooms, following which it can roam on a floor with several thousand rooms. Action Schema Network’s (ASNet’s) ability to crack much larger problem is a game changer. Employing ASNet-based system, new cyber-security applications could be created that help find system vulnerabilities, or create new robotics software to carry out specialized pursuits in unmanned space missions or automated warehouses.
The new intelligent systems depends on automated planning technology for decision making. Nevertheless a planning algorithm is needed to choose the top course of action whether it may be for a Mars rover picking where to click pictures, or a smart grid learning to isolate a fault. The new intelligent system involves some deep-learning-laden systems, which includes AlphaGo has been employed to provide a solution for decision-making problems.
AlphaGo – is a recent and popular example which attained superhuman like skill in Go. However, AlphaGo has the limitation that it can only be played on a fixed board size.