Dumb machines and verbs

When Google acquired YouTube, most people thought it was simply expanding its empire. But the truth is that it was another step in the research on how both the search engine and the place with countless videos will become an advanced servant of anyone connected to the Internet. Right now, Google, first and foremost, and the other search engine owners are trying to find ways to get into our minds. To make their engines work as close as possible to the way the human brain perceives the world. The phrases that researchers and managers exchange with each other are of the type: “The future of search is verbs”, “The appearance of a link is not the answer”, “Machines should understand concepts”, “Search has gone from searching to find something (that we knew existed), to searching for something we don’t even know (its existence)”. Phrases that when you want to analyze will take you to the research sites of the most advanced institutions in the world.

Valuable information in the hands of fools

As Peter Denning and Tim Bell wrote in an article published in "American Scientist" (November - December 2012, p. 470), since the time of the first computers, information (in essence, the data from which the information is derived) has been circulated in a thousand ways and in all directions, but without the machines understanding what they are circulating. Something that does not happen to us humans, when we are of course in a good mental state. When we transmit something to others - knowledge, a threat, a secret, a confession -, we first have a sense of what it is that we will transmit. Computers currently remind us of that slave whose master wrote a message on his shaved head, let his hair grow and sent him somewhere, certain that even the slave himself was unable to read the message. Now, however, this is changing.

Information Theory has been very concerned with how to transmit a lot of data in a short period of time and securely. Now it is also concerned with the meaning that these have or can take. Machines to "realize" what we ask them to transmit. We are currently in the era of interactive systems. They are fed with data and produce results. From GPS to Facebook. However, it is already considered that a problem that has not been solved in the best way is how these systems could automatically acquire a logic, that is, a digital label that would accompany a series of photos or a collection of many videos and would be consistent with their content. In 2010, Paolo Rocchi, an Italian-born researcher at IBM, resorted to the tricks of semiotic theory and borrowed from there the signifier and the signified.

We see the red traffic light on the road and in our minds this is associated with the command to brake. In the computer that will make it easier for a robot to drive a car on the roads, the pair (red light - stop) must be registered. What could be easier, someone might think. Yes, but the volume of associations that a person has made with the help of his brain, from infancy, where it begins first of all with the faces of those who approach him, is enormous. Researchers understood this when they tried to make machines recognize faces. Today, however, the means available are greater and the amounts on the table that await the winner are mythical.

Google is currently closer to the goal, and even if it is not advanced in a particular field, it can buy the one that is ahead. As was the case with the company Metaweb, which was founded in 2005, was bought by Google in 2010 and has now released the Knowledge Graph package for public use, a search engine that has moved from searching for a sequence of characters, with or without meaning, to searching for entities (https://www.google.com/insidesearch/features/search/knowledge.html). From character sequences to entities, as they say. “We searched and cataloged everything that could exist in knowledge on the planet,” said John Giannandrea, somewhat arrogantly, in the magazine “Wired” (January 2013), “and we are marrying it with what a search engine can do today.

Pizza. What could be simpler than a word that one types in to request information? And yet, the challenge now is for search engines to be able to distinguish, and as quickly as possible, whether we want to learn about the history of pizza or if we are hungry and looking for the best pizza in terms of price and taste in the area where we are.

The phrase "the link is not (anymore) the answer" can be perfectly understood in the case of pizza, but it also perfectly reflects the orientation of the research program by a team at the University of Washington, which has ended up giving revminer.com a first form. A search engine capable of bringing to the mobile screen information about the various shops where food is offered, e.g. pizza, the opinions of those who have already visited them, but also acts on its own and then gives you the text and the summary in "keywords", e.g. parking, waiting, drinks, dessert, behavior, etc.

The future of search is verbs, Bill Gates had said, meaning that at some point people will want search to suggest things and not simply bring to our attention what is there. Of course, in this mad race for even more data and the tendency for the computer itself to produce new information, after relevant processing of what is already stored in its memory, there is always the fear of overconcentration (and even forbidden) data, as happened a few years ago with Google Earth. Let's be careful. Sometimes too many and inappropriate conveniences can be harmful.