Over the next decade, many governments will face a series of choices with regards to the deployment and regulation of machine learning technology. These choices are likely to be influenced by public opinion and interest in this technology. The extent to which citizens of different countries are interested in AI can thus be of relevance to understand how these choices will unfold.
Interest in a topic in a target population has historically mostly been ascertained with surveys. More recently, researchers have had access to Google Trends, a tool that analyzes the popularity of search queries in Google Search. The advantage of this tool is that it is an intuitive and straightforward way of measuring the general public’s interest in a topic: it is plausible to assume that people who search for “machine learning” also are interested in machine learning. However, it is worth pointing out that Google Trends also has its drawbacks. Most notably, as Google usage in China is very limited, the data from this important country is too biased to use. Also, the data does not represent the actual fraction of searches but relative popularity.
The countries with the highest interest in AI topic are from East Asia (the technical details of this study can be found in the methods section). Top 10 countries with the largest search for AI topic during the past year.1
As we can see from this list, four out of the top five locations are in East Asia, and Japan is number 7. While these countries differ with regards to language, history, GDP/capita and demography, they can all be described as being part of the “Sinosphere”, a term the denotes a region that was historically influenced by Chinese culture, including literary traditions and philosophical thought.
|9||United Arab Emirates|
In this cultural context, the board game Go is not only very popular, but also carries deep and ancient cultural significance. Mastery of this game is mentioned as one of the essential skills of a gentleman in early Chinese literary sources. Thus, we hypothesized that the unusual interest in AI was a consequence of the highly publicized and widely seen match between DeepMinds Go-playing AI “AlphaGo” and the top-ranked professional Go player Lee Sedol in March 2016. This hypothesis was confirmed by reviewing data of searches over time. The effect is especially strong in countries where searches for AlphaGo were very popular.
As you can see from the graph below, there is a spike in the interest in AI at the time of the match between Lee-Sedol and AlphaGo in countries where “AlphaGo” was a common search term, but not in other countries. Thus, we believe that there was a real and considerable effect in the Sinosphere of this particular match, and that this effect could have major social and technocultural consequences in the future diffusion of AI technology.
Figure 1. AI interest trends among countries with high and low interest in Go. The figure shows the relative popularity of the topic “AI” in Google searches for individual countries (faded lines) with high (red) and low (blue) interest in Go, as measured by searches on “AlphaGo”. The thick line shows the average trend. Similar trends were observed for searches related to machine learning.
Figure 2. ML interest trends among countries with high and low interest in Go. The figure shows the relative popularity of AI topic in google searches for individual countries (faded lines) with high (red) and low (blue) interest in Go, as measured by searches on “AlphaGo”. The thick line shows the average trend.
As we lack search data from China, we cannot say whether interest in AI was also influenced here. However, others have argued, relying on qualitative data, that it indeed was the case. Ironically, the somewhat idiosyncratic efforts of DeepMind’s founder to create a machine that could be the best humans in his favorite board game, could have altered the techno-historical trajectory of AI technology on a grand scale.2