"{Search is Dumb}"

Last night I attended a tech talk by Marissa Mayer, VP of Search Products and User Experience at Google (and Google's first woman engineer!). The event was organized by Stanford IEEE. What I walked away with? A slice of pizza and the provocative statement that, "Search is dumb," the title of Mayer's talk. What's it called again when you assign human traits to objects...personification?

I often feel that the world's information is at my fingertips, so it's hard for me to completely get it when I'm told that only a small fraction of search (not even 10 percent) has been solved, that the most interesting stuff in search is still to come, and that, as Mayer put it, search is like a three year old child. Here's why.

Search doesn't understand you. For example, when you search "nice cafe," it doesn't know what you mean by "nice" (a value judgment), and when you search "jaguar," do you mean the animal or the car (type differentiation)? Search doesn't understand the context you bring to your searches, i.e. it doesn't understand user intent. This is the next big challenge in Computer Science -- the journey to engineer the perfect search engine. In a world of perfect search, each search you do would be returned with a perfect answer that takes into account context and intentions...and it'd conveniently be in the right media form (maps, images, videos, web pages).

Search is growing up in a bad environment. Search can only be as good as the Internet is, and right now the Internet has gone negative because the current infrastructure doesn't encourage people to create websites that house high-quality information/data. What we see on the Internet today is:

1) Unmanageable/unmonitored growth: Every minute, 15 hours of video content is uploaded to YouTube. Every day, 120,000 new blogs are created.

2) Anonymity/lack of responsibility: 60 percent of bloggers don't list their full first names. 31 percent of users on social networking sites lie about their identities.

We need to build a more useful web. In the past 10 years, Internet content has grown a thousand fold, but the percentage of that content that can be called useful has stayed constant at about 15 percent (based on web crawling research). Until more of the web becomes useful, improvements in search technology will only have marginal effects on the quality of information users can access. (A useful website is one that contains reliable, relevant information that can be used for productive purposes. It doesn't show up as the 2000th result of a Google search, thus offering duplicate information; it doesn't drive traffic to its site for the main purpose of generating ad revenue; and obviously it isn't porn.)

I'll end with some final thoughts...

a) If Google is really heading toward basing its search results off a database of intentions (as opposed to mere indexes), how will Google collect information on user intentions? If it's through analyzing and tracking user clickstreams and inferring intentions based on past trails, then my question is: Are past trends a good indicator of what users want in the future? To a certain degree, I believe that users don't always know what they want or what's best for them and that part of the innovation process involves exposing them to ideas (or in this case search results) that they may never have stumbled upon otherwise.

b) I'll get over the fact that my blog isn't useful yet. ;)