Not a Search Engine
Google dominates the search engine market. No question about it. It’s the default homepage of billions of browsers, and the first stop for many internet users across the globe. Last week, I was invited to a sneak preview of something that may challenge that, or at least offer a service that Google is only just starting to explore.
Wolfram Alpha is not a search engine. Those were the first words of Gary Price, as we started on a tour of the ‘Computational Answer Engine’. No, instead, Wolfram Alpha analyses the structure of any question asked of it, and provides a response calculated from verified data. It has 20 trillion pieces of data available to it, collected over 2 years by 100 PHD-level participants. The data comes from a huge variety of sources: government census; commercial enterprises and organisations across the globe.
The project comes from the mind of Stephen Wolfram, founder of the Center for Complex Systems Research at the University of Illinois where he developed Mathematica, a “computational software program used in scientific, engineering, and mathematical fields”.
So, how does it work?
When asked “How much fuel does American Airlines use per day?”, Wolfram comes back with the exact answer, plus a collection of useful related data, charts and appropriate references. You can expand the question by adding additional parameters, such as “in US$” and the engine provides the calculated cost of that fuel in dollars, along with other currencies at the current exchange rate.
Typing in a name, such as “Russell” brings back data on the popularity of the name with global census data. The query “Brother’s Father’s Mother’s Sister” returns the answer (Great Aunt, if you were wondering) along with a handy tree diagram to illustrate the relationships.
There are obvious limitations to this system – searches on pop culture, non-celebrities or breaking news may not bring back relevant results, and the system may have trouble understanding your question, if not asked in the right way. But then, Wolfram isn’t trying to answer those sorts of questions – it’s not emulating Google, it’s providing a different sort of search – a more intelligent search with data you can really trust.
Last week, the developers accidentally left the engine exposed for 4 hours. In that time, they registered 3,000 unique visitors and 18,000 searches. They had a 72% answer success rate. 80% of the failures were determined to be vanity searches (i.e. people searching for their own names, and unsurprisingly finding no results).
During my 30 minute webinar, Gary discussed their plans for monetizing: sponsored data/targeted ads; the ability to purchase processing time for more complex queries (“The engine could be used for medical calculations – cancer research, for example”); and an exnterprise model, where companies and organisations can plug their own data into the engine.
Wolfram officially launches on May 18th, so we’ll soon see how it stands up to the myriad of questions web users will undoubtedly throw at it.
