How we will search the web in the next decade.
We must welcome the future, remembering that soon it will be the past; and we must respect the past, remembering that it was once all that was humanly possible, George Santayana
When Google started in 1998 many fell in love with it. It had a cute name, fast results and there were no ads. What was not to like about it? And then a few years later, this guy comes and rewrites their core algorithm, from scratch! As if on purpose Google takes a technological leap right at the turn of the century, leaving its competitors back in the 20th. Of course nobody noticed as nothing visual changed. We were still greeted by the same logo, search box and ten results per page. To this day many of us still perceive Google in exactly the same way, even though we’re essentially using a completely different search engine under the same brand.
For fifteen years now, a team of engineers at Google has been adding layers upon layers of new features and algorithmic tweaks, most of which were subtle enough to go unnoticed. Heck, many didn’t notice when they introduced ads in their search results. Algorithmic updates and feature experiments tend to graduate from Google’s sandbox to a limited user base. Once the confidence level is high enough, more geographic regions get to see it. Eventually it may go worldwide.
The culture of experimentation and testing is so ingrained in Google’s DNA that they might easily run some twenty thousand experiments per year which will help them decide what works and what doesn’t.
We would try out anywhere from 10,000 to 20,000 ideas. Of that, many more thousand, 8,549, we would send to these blind side-by-sides, and then a smaller fraction of that actually get sent out to real users and to see whether users tend to click on the newer results or tend to click on the older results. And the final number changes that we launched last year was 585.
Once the ad money started pouring in Google was sufficiently cashed up to ignite an “experimental renaissance” of sort. This stage in Google’s history reminds me of Jurassic era. Most of their attempts to break outside the pure search market resemble clumsy evolutionary experiments which ended up with some funny looking life forms. “The mass extinction” happened when Larry Page came back as the CEO of Google in 2011 and cut everything that didn’t fit his vision. The big cull affected thirty Google products including Knol and Sidewiki. The whole process was fascinating and it showed everyone that Google’s not mucking around. They’re serious about Google+ being the central point of all Google products and a device for understanding true identities in the anonymous web.
Google right now consists of the three building blocks:
a) Link Graph
b) Social Graph
c) Knowledge Graph
Link graph has been there since day one and basically runs on the random surfer model (or more accurately reasonable surfer model) where pages accumulate and pass PageRank around. The amount of accumulated ‘link juice’ determines the importance of the page on the web. Naturally it’s not all so simple and there are many other factors which influence how Google sorts their results.
Social graph is a recent thing, but not quite new. Similarly to document connectivity on the web Google looks at individuals and their explicit and implicit social connections trying to make sense of who we are and how to customise our search results accordingly.
Knowledge graph is quite new and exciting really. It’s Google’s way of connecting various entities on a semantic level. So right now we’re witnessing first glimpses of something that will soon be true artificial intelligence. It’s not quite there, but a few neurons are starting to fire up in Google’s attempt to understand its own index on a whole new level. When I say neurons I am not kidding.
Neural networks are not a new concept. They’ve been around for a long time now and have practical application in machine learning including natural language processing, image and speech recognition.
This means that between now and about 2020 we can expect to see Google which starts to understand things much better. We’ll be able to have a basic conversation with the search engine and it will be giving us more and more useful information. Google will know us better and will offer answers and suggestions in anticipation of our questions depending on our behavioural patterns, location, time, device, and search history.
From 2021 onwards Google will be able to predict, suggest, influence and aid our future choices. We’ll be able to send it on a complex task to research and summarise findings for us. It will be able of performing background investigative work and assist in self-paced learning. Google will be our virtual assistant and we’ll be able to ask question such as:
“How long do I have to wait between taking antibiotics and malaria tablets to avoid unpleasant side effects?”
“I want to visit France, Spain and Italy. Suggest an ideal itinerary based on weather, optimal travel rote by car and weather in each country. Show me rental car options and find the lowest price for an Audi A4.”
The search engine will then be able to ask clarifying questions in a conversational manner. It will even chit-chat with us, just to extract that extra information from us and lean more about who we are.
Google will have a separate search index dedicated to each and every single user on the planet. Each such ‘mendex’ will be a new node in the social graph.
We can expect our personal Google to work on complex tasks such as content curation and even act as our representative in business and social life. It will be able to join the discussion, create new content and teach a subject.
Eventually, all search will stop and one day we will simply know things, and we will take it for granted.