Blog Listing

What Can We Learn from the AlphaGo Movie?

By Yina Moe-Lange  

 

A couple of weeks ago the AlphaGo documentary was released on Netflix, giving everyone an inside view of the DeepMind team and the months leading up to the big Go match in South Korea.

 

The documentary was quite fascinating and gave nice insights into the what happens behind the headlines, the part of the story that we usually don’t get access to. But there are two points that are important to look into as we move forward into the future with Artificial Intelligence (AI).

 

One important point that I would have liked to learn more about is the strategies and learnings AlphaGo achieved. The documentary does not discuss or explore AlphaGo beyond human capabilities. AlphaGo is treated as a human Go player rather than its own species. What kind of strategies does AlphaGo use to beat human players? Are they different from the ones that people use or is it just better at searching for options? I am wondering to which extent the AI actually developed new approaches to the Go game.

 

Screenshot from the live stream of Match 2 from Youtube.com

 

Demis Hassabis, the Co-Founder and CEO of DeepMind says in the documentary, “We think of DeepMind as kind of an Apollo program effort for AI.” But I would disagree and say that this is the Sputnik of our time. AlphaGo’s achievements have kicked China into high gear pursuing AI research.

 

In 1955 both the US and the Soviet Union stated that they would launch artificial satellites in the near future, marking the beginning of the Space Race. Both the US and the Soviet Union wanted to show its dominance and superiority of its technology.

 

 

A replica of Sputnik I in the National Museum of the US Air Force

 

Sputnik 1 was launched by the Soviet Union on October 4th, 1957. One can easily point to the launch of Sputnik as the trigger point for the US, where they realized how aggressively the Soviets were pursuing their task of space technologies. In reaction, the US created the Advanced Research Projects Agency (ARPA) which later became DARPA, an agency dedicated to the development of emerging technologies.

 

The Space Race left behind a number of significant achievements, but most importantly it increased focus and funding for education, research and development.

 

This is all relevant, as in the present day there is a similar “AI race” occurring around the development and implementation of AI. After AlphaGo showed its moves in South Korea, it became even more apparent that AI has massive potential and many applications.

 

In July of last year, China’s government released a plan on how it would work to become the global leader in AI technology. It then named a national team comprised of four companies, Baidu, Alibaba Group, Tencent Holdings and iFlyTek in November.

 

Tencent’s AI Go program Fine Art just recently made headlines as it beat China’s top Go player, Ke Jie. This just shows how quickly and aggressively China is working to get ahead of the rest of the world in AI.

 

What is important to note in this technology race is that the proliferation of information is quicker and broader than it was during the Cold War period. Papers that are published are generally accessible and there is a growing amount of Open Source software. It will be interesting to see if the Chinese are as open with their publications and methods as some of the leaders in AI in the US and Europe.