AI Learns to Play Atari Games 6,000 Times Faster than Before, Better than Deep Q-Network

A new development from the Carnegie Mellon Univerisity debuts a new artificial intelligence that beat its previous record in learning to play and beating Atari games that it was asked to solve. This new system is 6,000 times better and faster compared to before, particularly with Deep Mind's Deep Q-Network which it trained to finish games on its own.

Developers and researchers are adding AI to many industries and enforcing the technology to bring significant solutions and advancements to respective platforms.

AI Learns to Play Atari Games 6,000 Faster than Ever

A report by Singularity Hub details the recent study and achievement by researchers from Carnegie Mellon University over at Pittsburg for their latest achievement of developing a more powerful AI. This new artificial intelligence is powered through "reinforcement learning," which also takes into account game manuals to further help the system finish it.

"Our work is the first to demonstrate the possibility of a fully-automated reinforcement learning framework to benefit from an instruction manual for a widely studied game," said Yue Wu, lead researcher.

Carnegie Mellon University (CMU) is at the forefront of this innovation with its groundbreaking research in AI learning to play Atari games faster than humans.

In their pre-published research, the team said that it can achieve up to 6,000 times faster processing and solving when taking into account game instruction manuals before attempting to finish it.

New AI from CMU is Better than Deep Q-Network

In 2013, CMU researchers developed a new AI system called Deep Q-Network (DQN) that could learn how to play Atari games by trial and error. The AI system was fed raw pixel data from the game and was given the objective of maximizing its score. Through thousands of attempts, DQN gradually learned which actions led to higher scores and which ones led to lower scores.

The result was astounding. DQN was able to outperform human players in many Atari games, including classic titles like Breakout, Space Invaders, and Pong.

AI in Solving Games

Among those iconic AI companies in the world is DeepMind, and the company is known for tapping into different industries in which their artificial intelligence technology may help to solve or power. One of the most noteworthy partnerships of DeepMind is with Google, particularly as its tech was used by Big Tech to help detect breast cancer, among many collaborations.

The Alphabet subsidiary is also known for beating games using its AI technology, centering mostly on Atari games famous in the console. The team used its AI to train itself in solving and finishing all 57 games from the Atari console, which it was able to do and showcase to the public.

However, analysts claim that artificial intelligence is not "versatile enough" to take on the many features and capabilities it should have on games.

The new reinforcement learning algorithm helps it extend more of its machine learning function, to accomplish tasks faster and more efficiently, game-wise.