Stephen Wolfram’s new knowledge engine, Wolfram Alpha goes live today, providing a new kind of search for Internet users. As a result, the Internets are alive with comments about how Alpha compares to Google. But that’s something you shouldn’t do.
When I previously wrote about Wolfram Alpha, I described its potential as a “fact computer” and described how it intelligently trawls through a vast database of knowledge using algorithms to calculate its responses. It works like the computer from Star Trek: Ask it about an orbital calculation, and it’ll respond–ask it if football fans like watching games on TV and it flounders.
Check out some example searches. I typed in the zip code for Beverly Hills into Alpha and Google–the results are below.
Google provided a quick map of the area, a link to city-data.com for statistics, a link to Forbes‘ “Most Expensive Zip Codes” list and other such results. Wolfram Alpha provided a basic map, reeled off a hundred different stats down to the number of congressional districts in Beverly Hills and the total number of mail receptacles. And right at the bottom of the results, it reported the current local time and weather: 16ºC and overcast.
I also typed in a date: May 31, 1979–long before the Internet was alive, so the search wouldn’t just trawl up simple weblinks written on a specific day.
From Wolfram Alpha I could find out that that date was 29 years, 11 months and 18 days ago, that it was the 151st day in the 22nd week of 1979 and so on. Google’s first response was a Kakophone.com link which could tell me what was number one in the Billboard charts that day, or what won the Best Picture Oscar that year. There was a link to the New York review of books edition issued on that day, and many news-related links after that.
And here’s the killer query. I actually did ask both services whether football fans like watching on TV.
Wolfram Alpha just balked at it, politely reporting it didn’t know “what to do” with the input, and suggested possible onward searches relating to eight other topics, including “National Football League,” “Culture and Media” and “Socio Economic Data.” I suspect if you were diligent you could do some advanced data analysis using those onward tools and come up with a viewpoint in answer to the question.
Google’s first response was a link to a story headlined “People like to watch Alabama football on TV and in person.” And the results list was rich with opinion-based stories. By reading through the linked sites, probably finding some viewer stats embedded in a few, and doing some of your own thinking, you’d probably come up with the conclusion “Yes, they do. Very much.”
That question is difficult, since it embodies a problem that is neither particularly statistical or fact based, and its response is highly subjective. To solve that query is, in fact, beyond the ken of current technology, since it requires what’s called “soft” analysis to synthesize different data into a coherent whole, and that’s currently a skill reserved for the gray stuff between your ears.
That is, in fact, the point. Wolfram Alpha and Google just aren’t the same–even with Google’s recent and upcoming advanced search tweaks. One is a precise and fact-based rich data source that can provide all the details on a fact you need–and more–very quickly, whereas the other is essentially a smart list of things that match the text of your query, covering a variety of angles from hard data to opinions, to imagery. Alpha will be invaluable to scientists, technology writers, and when you need to know something detailed about a specific topic, event, or otherwise. Google is useful for fuzzier answers, and when you just want to know what people are saying. You have to do your own fuzzy grey matter-based analysis to determine which service will best provide the answer you need.