I Am Not Smart Enough for Wolfram Alpha

You may have heard recently about Wolfram Alpha, a new Web service that some people are projecting (mistakenly, I'm certain) as a "Google killer". With Wolfram Alpha, you type terms into a field on a webpage and get back information about those terms. But the big brains at Wolfram research insist that it's not a search engine, it's a "computational knowledge engine." The company announced today that the service should be publicly available on May 18.

After playing with the very early version of Wolfram Alpha for a few hours, I agree that it's not your average search engine. What is it? I'm not sure I'm qualified to say: Not only did I not understand all the answers I got from Wolfram Alpha, I couldn't even comprehend some of the questions I could ask it.

For instance, this sample request helpfully suggested by the designers of Wolfram Alpha -- Fit a polynomial to given data -- is something I've never even thought of asking Google to do. But who knows, someday it may come in really handy. Some of the answers are just as abstruse. When I did a comparison of Google's stock versus the stock of Intel, WA supplied me, in addition to other data, a chart labeled "mean-variance optimal portfolio." To someone who understands finance, this is probably incredibly useful. For me, it's Greek, which is probably why I'll be eating cat food when I retire.

I don't mean to criticize Wolfram Alpha though. The information it returned that I could understand was interesting and added depth and context that you often don't get from typical search engines. The service's data is incredibly limited at this point, though, which isn't surprising. The father of the project, controversial scientist Stephen Wolfram, says this isn't a finished product. So many of the queries even that the company suggests trying with WA simply don't work yet. The question is whether they ever will - from an outside perspective, it feels like the kind of project that could require immense amounts of human work.

(Wolfram himself is an interesting guy - a former physics wunderkind, he was in 1981 the youngest recipient of a MacArthur genius grant. In 2002, he published "A New Kind of Science," a book that argues that the fundamental laws of the universe can be described as simple programs called cellular automata, another concept I don't really understand. Much of the established scientific community has been skeptical. One review of the book was titled "A Rare Blend of Monster-Raving Egomania and Utter Batshit Insanity." A more even-handed, and exceptionally clear, review came from American Scientist. )

Probably the best way to understand WA is to see it in action. So here's a selection of screens from my exploration of the service.

This screen of a simple request for the population of San Francisco is a good example of how WA is different and in some ways more useful than conventional search engines. Not only does it tell you the current population of the City by the Bay, it also charts it over time, something I wasn't looking for when I typed the query, but found pretty interesting:

-----------------------------------------------------------------------

Given that Wolfram is a scientist, it's not surprising that questions about science and math yield particularly deep answers:

Financial data is also extremely detailed. My comparison of Google and Intel yielded a torrent of information, including the fact that the average Google employee brings in $1.24 million in revenue annually, which made me feel incredibly inadequate:

-----------------------------------------------------------------------

When you enter an ambiguous query in WA, the service tells you how it's interpreting what you entered and offers an alternative. When I entered "Oliver Cromwell," it guessed I wanted biographical information, but also offered information about Cromwell as an historical event, which yielded the screen below:

One unique property of WA is that it is set up to do on-the-fly comparisons when you enter two distinct queries. Here I entered declaration of independence and War of 1812. The resulting chart supplied dates, people and countries involved. If this data ever became as deep as that associated with scientific and mathematical queries, WA could be a godsend to history students:

-----------------------------------------------------------------------

I'm not sure why this feature was a priority for the WA developers, but it is fascinating: enter two first names and the service will compare their frequency in the United States now and over history, as this smackdown of Mary vs. Tiffany shows:

-----------------------------------------------------------------------

Ever careful, WA even checks to make sure that you meant Mary as a female name. Click the "Mary (male)" link and you discover that each year about 67 very unhappy boys are named Mary:

-----------------------------------------------------------------------

I wouldn't expect pop culture to be a top priority for Stephen Wolfram, so it's probably not surprising that the results in that area are pretty mixed. When I asked for Star Trek, I got a table with most of the basic information I'd want about the movie, even down to the name of the actor who played "Burly Cadet #1." And a drop-down menu let me change the results if I was more interested in "The Wrath of Khan" (and, really, who wouldn't be?):

-----------------------------------------------------------------------

My search for Amy Winehouse was disappointingly shallow, though - just her profession and date and place of birth. No professional biography, no discography, no arrest records:

-----------------------------------------------------------------------

Another quirky set of data included in this preview of WA is nutritional information about generic foods. Again, I'm not certain what all these charts mean, but they seem to be implying that five marshmallows is not a healthy breakfast:

-----------------------------------------------------------------------

Subscribe to the Daily Downloads Newsletter

Comments