Artificial Intelligence is one of the most sought after features for any program or a system. The computer that can play games like Pong and Space Invaders easily, will be considered by many as the most artificially intelligent computer. The program deep Q-Network (DQN) is considered the most artificially intelligent systems, as the program has already mastered more that 50 Atari 2600 games. But on the other hand, the program still struggles with games like Pacman, which shows the existing scope for improvement.
The program has been funded by technology entrepreneur Elon Musk and physicist Stephen Hawking. Both believe that Artificial Intelligence is of extreme importance as the machines also need to know at times what they are being used for. The experts believe that in case machines can be imparted with a thinking ability for more complex situations, there will be an automatic boost in various industries across the globe.
The first success in the field of Artificial Intelligence came in when in 1997, IBM’s Deep BLUE defeated the then World Chess Champion Garry Kasparov. Since then, a machine’s intelligence has been worked upon to yield better results. Although, the expertise has grown to a much wider spectrum, machines are still way behind human intelligence.
This is where, DeepMind Technologies’ deep Q-Network is better than its remaining counterparts. It has the ability to learn and the program has learnt to play all the games from the scratch. According to Dr. Hassabis, they only provided the program with the information about the raw pixels of the screen and the idea of making a high score. The program has learnt all the other things by itself. As a result, the program can learn how to play a new game after a few hours of providing it to the system. The success rate varies with every new game.
DQN basically uses two techniques of artificial intelligence to achieve its success: Deep Neural Networks and Reinforcement learning. The Deep Neural Network was developed in the year 1980 by Dr. Kunihiko Fukushima and the latter were developed based on Human psychology. For DQN, both the methods work in collaboration, allowing the program to learn the unknown steps.
The experts believe that DQN has many potential applications, mainly in the field of Robotics. Although we are not talking about the robots in Sci-Fi films, but a machine with the ability to adapt can yield much better results that the modern day high tech ones.