DeepMind is a London-based artificial intelligence lab that was recently acquired by Google for an estimated £400 million. In an official statement this past Wednesday, the company stated that it has improved upon its most notable AI software, AlphaGo, to form the newly created “AlphaGo Zero.”
AlphaGo is an algorithm that is notoriously known for its self-learning process. It made headlines around the world for its ability to master the ancient Chinese board game “Go” and has defeated some of the best human players from all across the world. After a lot of work and improvement, DeepMind has upgraded the software into AlphaGo Zero, which is much faster and well-rounded than its previous version. The main objective of AlphaGo Zero is to learn every possible outcome of self-play games in order to defeat its opponents. The new software was so effective in three days of trainings that it was able to pick out certain advantageous positions and sequences to defeat AlphaGo Lee, DeepMind’s software that defeated top Korean “Go” player Lee Sedol in 2016 by 100 games to zero. After 40 days of training and 29 million self-play games, AlphaGo Zero was able to defeat AlphaGo Master, the software that defeated world champion Ke Jie, 89 games to 11.
AlphaGo Zero is a reflection of how artificial intelligence can be manipulated to learn from its mistakes in order to master a process. Of course, it is limited to a simple game at the moment, but this is only a glimpse of the true potential of artificial intelligence in the near future. If the systematics of AlphaGo Zero were applied to a real-world “game,” we may see incredible results. This also shows how human involvement with artificial intelligence is a mutual relationship that only helps our cause in the very end.
Though some view artificial intelligence as an unnecessary investment, it may be the technology that could revolutionize our society for the better. The negative stigma of AI should be reevaluated, as the benefits for artificial intelligence seem to be potentially life-changing.