AlphaGo Proves the Future of AI is Bright
As algorithms have advanced, technology has gotten smarter. Case in point, last month in China, Google’s AlphaGo –AI specifically developed to tackle the ancient strategy game – defeated Go world champion, Ke Jie. In the three game series, AlphaGo beat the Chinese grandmaster by a slim margin. This isn’t the first time the revolutionary machine has proven its superiority in the game. In March in Seoul, South Korea, AlphaGo made history as the first machine to beat one of the world’s top players, Korean grandmaster Lee Sedol.
Created by Google’s AI lab, DeepMind, AlphaGo has made waves in the artificial intelligence realm. From the beginning, it was met with fear and skepticism. In its early stages, AlphaGo was received with doubt as many students and professionals questioned its skill and capacity to challenge the world’s best players. When the machine proved its skills and succeeded against Sedol, many were worried that a machine had exceeded the powers of humanity. As DeepMind continued to evolve the system, top Go players became increasingly fascinated with its mastery of the game, some going as far as positing the machine as the future of the complex game.
AlphaGo wasn’t born the master of all grandmasters. Like any other AI – and people for that matter – AlphaGo learned how to succeed by playing game after game against itself. Some matches were conducted under time limits with each move taking mere milliseconds. Other matches unfolded over several hours. In a bid to train and navigate the system through the mysteries and extensive complexities of the game, DeepMind recruited Fan Hui, the European Go champion, to help train it. While DeepMind has only released 50 games the system has played against itself, the machine has played hundreds of thousands of times – if not millions – against itself.
DeepMind continued to advance AlphaGo and the system was redesigned to train from games played against itself without the help from human moves. The new incarnation proved to be a resounding success when it beat Chinese grandmaster Jie.
Another compelling aspect of AlphaGo is its ability to run locally, otherwise referred to as embedded AI. During the match with Jie, the AI system ran on just one of Google’s new tensor processing unit (TPU) chip boards. Without connecting to internet or servers, the system was able to do all of its computing locally using only about a tenth of the processing power used by the original model before it. Powerful and efficient, it not only demonstrated its exceeding computational capabilities, but also unveiled a wide breadth of tasks AI can be applied to. From optimizing power grids and advancing scientific and medical research to streamlining shipping routes and revolutionizing security, AlphaGo beckons innumerable possibilities.
In a bid to improve surveillance and security measures, ULSee is applying its machine learning algorithms to create embedded AI recognition solutions. Using a camera to capture body movements, the company’s behavior recognition is trained to identify a myriad of actions, poses, and motions without the help of internet and external servers. This development adds a new dimension to protection and privacy.
The evolution and success of AlphaGo is just one example of the way AI can not only replace human skills, but enhance them. Using AlphaGo as a learning model, many top players have changed and improved their own strategies. While human ability can never be fully eclipsed by technology, embedded AI holds the computational power and efficiency to push people to places they couldn’t otherwise reach on their own. From healthcare and scientific research to driverless cars and robotics, the technology can be applied to a wide range of tasks and industries. Though the DeepMind team has since retired AlphaGo after its victory in China, the future of AI shines bright. As the exciting evolution of AI unfolds, one thing is for sure – the technology will impact the world in unprecedented ways.