On May 16, Google introduced the most significant development of its search engine since its redesign: Knowledge Graph. It is an intelligent model that has the ability to understand the meaning of words to improve search quality. The question we will try to answer in this article is: Is Google’s Knowledge Graph really intelligent?
It is curious that, in order to be more intelligent, the search engine of Google has undone the path of abstraction, which humanity did with so much effort. Knowledge Graph re-associates the word with its meaning, leaving to treat it as a string of letters. As a Latin quote says, “légere et non intellígere est tamquam non légere” (to read, and not understand, is not reading). Thus, understanding, the Google’s search engine becomes intelligent.
As Amit Singhal (SVP, Engineering) presents it on the Google’s blog, “This is a critical first step towards building the next generation of search, which taps into the collective intelligence of the web and understands the world a bit more like people do “. We’ll see if this is so.
According to Google, the Knowledge Graph intelligence lies in the following three skills:
1. Understanding what we are looking for. Knowledge Graph knows that words can refer to things, people or places.
2. Gathering relevant information about the thing, people or place that we seek.
3. Making connections between things, people and places.
To check how Knowledge Graph operates, I made an experiment. I entered the keyword “Leonardo da Vinci” in the search box, and from here, I’ve seen how far it is possible to go.
First step: “Leonardo da Vinci”
As shown in Figure 1, entering the words “Leonardo da Vinci”, the image of the painter appears to the right of the screen. Knowledge Graph understands that these words refer to a person. Next to the painter’s image, Knowledge Graph provides information about the artist and some details about his life.
However, the first thing that catches the eye are the sections called “Works” and “People also search for”. This is a function similar to that used by many libraries on their websites, which includes information such as “people who bought this book also bought …” This is undoubtedly where we will be able to see if Knowledge Graph really knows how to use the collective intelligence of the web and to approach to human reasoning.
Second step: “Mona Lisa”
I click on the painting of Mona Lisa and this brings me to Figure 2. Knowledge Graph now relates the Mona Lisa with other paintings and works of art, among them The Last Supper by Da Vinci.
Step Three: “The Last Supper”
I click on the picture of The Last Supper and this brings me to Figure 3.
When I see the painting of The Last Supper, the first thing that comes to mind is … The Da Vinci Code! However, Knowledge Graph only offers me more works by the same author. It seems that we entered into a closed circle, beyond which it is not possible to make connections.
Knowledge Graph is very efficient combining concepts from the same area. However, it is not able to make the leap that I have done in less than a second, from the picture to the film. The Google’s search engine cannot find the relationship between two concepts from different areas.
Perhaps the association between The Last Supper and The Da Vinci Code was too radical. We will give a second chance to Google. Now we will search directly for “The Da Vinci Code“.
As shown in Figure 4, by searching for the words “Da Vinci Code”, Knowledge Graph proposes us two possible options: either the book or the movie. But … look at the bottom of Figure 4, in “Images for da vinci code”. Surprise! The image search engine did make the link between the film and the painting.
Without fully understanding the differences between the search processes of the image search engine and Knowledge Graph, I venture to think that the advantage of the first lies in a better use of the collective intelligence of the web and less use of logic.
To explain more graphically the difference between Knowledge Graph and human intelligence I will use the next example. If we showed a raincoat to Knowledge Graph, it would reply, “jacket, coat, jersey”. If we showed a raincoat to a person, the person would reply, “Casablanca, Humphrey Bogart.”
My conclusion is that Google’s Knowledge Graph is not intelligent, but it has taken the first steps. Knowledge Graph is really great in logical reasoning. However, the Google’s search engine has not the ability of divergent reasoning, one of the basic ingredients of creativity and innovation.