Join us as we explore the AlphaGo Moment, when Google’s superhuman AI conquered the world’s hardest game.
Our speaker Hajin Lee served as World Go Champion Lee Sedol’s official liaison during the historic Go match between Google Deepmind's AlphaGo and Lee Sedol in 2016. Hajin, who is a ranked professional Go player herself) will speak about Go culture and the AlphaGo competition. Next, Dan Maas will dissect how the AlphaGo program works, how it compares to IBM’s DeepBlue chess-playing program from 2 decades ago, and more.
The AlphaGo Moment
In March of 2016, the world changed: a software program developed by Google defeated the world’s champion player of Go. The game of Go is considered by many to be one of the most complex and nuanced games: compared to chess, there are 1,000 times as many moves at each moment. Go has its origins in Asia and has been played continuously for at least 3,000 years with a strong culture of training and competition. Against this backdrop, Google’s DeepMind used a branch of artificial intelligence (AI) called deep learning to tackle the game of Go.
In 2014, the best Go program could only play at an amateur level. Just 2 years later, Google DeepMind had grown so confident about their new Go program, named “AlphaGo”, that they challenged 18-time world Go champion Lee Sedol to a 5-game match in March 2016. AlphaGo prevailed in 2016 winning the match 4-1. Then one year later, Google’s DeepMind developed a new Go program called AlphaGo Zero that trains by playing itself, with no input from human Go experts. After only 3 days, it achieved a level where it can beat the 2016 version of AlphaGo 100-0. We live in exciting times.