By Yina Moe-Lange
After a successful weekend defeating the Chinese Grandmaster and world number one Go player, Ke Jie, DeepMind’s AI-based player, AlphaGo, is retiring. DeepMind is a London-based subsidiary of Google. Even when Ke Jie used AlphaGo’s own moves against it, the AI system still managed to win three out of three matches. Now that AlphaGo has reached the pinnacle of its career, it is being retired as a competitive Go player. AlphaGo was told it only had to win so one can only wonder what the score would have looked like if AlphaGo had been told to go for the highest score possible while winning. Go is considered one of the hardest strategy games in the world and now DeepMind has proven that AI can excel against humans in that environment.
Google says that the AlphaGo team’s next task will be to apply the same tools used to develop AlphaGo to real-world problems. Their goal is to continue to develop advanced but general algorithms that can be applied in a variety of real-world situations. The research team says they will look at everything from discovering new cures for diseases, figuring out how to reduce energy consumption, and creating new materials. Applying AlphaGo’s drive for breakthroughs and strategy to real-world problems will likely lead to the discovery of a significant amount of new knowledge and insights.
This comes during the same period in which Google’s CEO, Sundar Pichai, stated that Google is moving away from “mobile first” to become “AI first.” This is a clear shift in Google’s strategy. The recent Google I/O developer conference was also heavily themed around AI and the different products and services that Google was announcing. These new position statements are also coupled with the reports that Google has set up a new venture capital program that is focused on AI. It appears that the investment arm will be led by engineers rather than traditional venture capitalists. It will be interesting to follow which startup companies Google will decide to bet on.
After the success of DeepMind’s AlphaGo, it is no surprise that Google is planning to heavily invest in AI both internally and externally. The trajectory of AlphaGo looks to be similar to the one that IBM’s supercomputers have been on. As you may recall, IBM’s Deep Blue defeated chess world champion Garry Kasparov in 1997. This set the stage for Watson, which became the Jeopardy! champion and is now the face and brand of IBM’s AI and Machine Learning initiatives. So, it will be interesting to see if Google’s AlphaGo will go through a similar process.
P.S. As a retirement present to the people, the AlphaGo team has also released a set of AlphaGo vs AlphaGo games. These are a selection of the games that AlphaGo played against itself to constantly improve as its own teacher. Check them out here: https://deepmind.com/research/alphago/alphago-vs-alphago-self-play-games/