Wolfram Alpha, the latest alleged Google killer, is not a search engine. Rather, physicist and software entrepreneur Stephen Wolfram’s long-awaited new Web project is a “computational knowledge engine.” The difference is important. A search engine tells you where to find stuff on the Web. Type “box office goodfellas” into Google, and you get a list of links; it’s up to you to figure out which page to click on to find how much money the Scorsese film made in theaters. Wolfram Alpha, which launched over the weekend, tries to give you a direct answer— ask it the same question, and you get the result without having to look on any other site: $46.84 million. The site’s best trick is its sometime ability to make calculations on the fly. Type in “box office goodfellas + box office casino” and it’ll tell you that the movies made a combined $89.35 million. Think of it as a calculator attached to a reference library—or your own personal HAL 9000.
If only it worked. Wolfram Alpha is a neat concept, and some of the posted sample queries—you can calculate the payment table for a mortgage or how many calories a 40-year-old male who’s 5-foot-10 and weighs 160 pounds would burn if he ran at 4 miles per hour for 30 minutes (272)—are quite impressive-looking. But in my few days of using it, I’ve found Wolfram Alpha almost completely useless.
Sure, the engine can tell you the musical notation for D# minor and the life expectancy for a 65-year-old woman in Sweden. But there is so much more that Wolfram Alpha doesn’t know. Once you start conjuring your own searches, it’s clear that the samples offer a misleading impression of the site’s depth. Ask how many calories that male runner would burn if he were swimming, cycling, playing tennis, cross-country skiing, or golfing—it’s clueless. Say you wanted to know how life expectancy differed by state in the United States—what’s the life expectancy of a male in California, and how does that compare to the life expectancy of a male in Kansas? “Wolfram Alpha doesn’t know what to do with your input,” the site tells me. And on and on it goes. Wolfram Alpha doesn’t know the homicide rate in South Africa or Baltimore, it doesn’t know how many copies M.I.A.’s last album sold, it can’t tell you the per-capita GDP of the San Francisco Bay Area, and it’s got nothing about the top speed of the Bugatti Veyron.
It may sound like I’m nitpicking, but I was careful to construct questions that emulated Wolfram’s own examples. As it kept coming up empty, Wolfram Alpha came to seem less like HAL 9000 and more like a chatbot. It’s been trained to respond to some kinds of queries, but any variations leave it stammering. It’s an idiot savant, smart about a few things but profoundly ignorant about large swaths of human knowledge. *
Of course, Wolfram Alpha is just starting out, and you can expect it to improve over time. When you consider how it works, though, it seems unlikely that it will ever be as useful as an ordinary search engine. According to reports, the site gets its data from a huge list of reference works—Stephen Wolfram told Wired that the site’s brain is built on things like the CIA World Factbook, U.S. Census reports, Wikipedia, and “about nine-tenths of what you’d see on the main shelves of a reference library.” There’s a lot of information in those works, and Wolfram has done a pretty good job of mashing all the data together to make them useful. But Wolfram’s sources pale in comparison to the library that search engines rely on—the World Wide Web, the largest database of information known to man.
Wolfram has discouraged comparisons between his new site and Google, insisting each serves different purposes. But that doesn’t make any sense; you’ll use both sites to look up stuff you don’t know, so it’s irresistible to compare the two. On a side-by-side test, Google wipes the floor with Wolfram Alpha. Using the search engine, I found life expectancy information for California and Kansas, the murder rate in South Africa and Baltimore, M.I.A.’s album sales, economic data for San Francisco, and the top speed of a Veyron: 253 mph. When you ask Google how many calories you’d burn on various sports, you find this page on NutriStrategy.com, which lists caloric measures for all those activities and dozens more, including playing drums, horse grooming, and raking the lawn.
That example illustrates the difficulty Wolfram faces in trying to match Google. Much of the data that we look for online aren’t found in formal, structured tables like the CIA Factbook. Even Wikipedia misses mountains of facts. But there are countless one-off pages, like the one at NutriStrategy, that collect information we don’t know we need until we search for it in a moment of whimsy. Wolfram misses all of that stuff; Google catches it. Wolfram’s other problem is speed—it clearly relies on some databases that aren’t updated as quickly as stuff you’d find on the Web. When you ask it for the U.K.’s unemployment rate, it reports a 2008 estimate of 5.5 percent. Google makes a similar mistake—at the top of the page, it reports the World Factbook’s 2007 estimate of 5.3 percent. Right below that, though, you get results from news reports for that same query, with a link to this BBC story with the accurate unemployment rate: 7.1 percent.
Wolfram Alpha will prove helpful to some people. Because it’s based on Wolfram Mathematica, a software package that can do complex calculations, mathematicians, engineers, and scientists will find much to love in it. Wolfram Alpha solves difficult equations and makes nice graphs for lots of specialized inquiries. The query “NACA 4351 15 degrees” will tell you, among other things, the lift co-efficient, the critical Mach number, and the center of pressure for an airfoil with the described shape and angle of attack. It’ll also draw a very pretty picture.
Plus, the site is kind of fun in the same way a chatbot is fun—it’s a thrill to test the limits of Wolfram’s intelligence, to see what it knows and where it fails. Geeks have been busy compiling its many Easter eggs. Ask it the average speed of an unladen swallow, and it indicates it knows the Monty Python reference. Or try, “How many roads must a man walk before you can call him a man?” There is also some evidence that Wolfram can save your life. Type in “frostbite 2 degrees f 25 mph,” and it tells you that if you’re stuck outside in 2-degree weather with a 25 mph wind, you’ll get frostbite in 22 minutes.
Google can’t do that. But it already matches some of Wolfram’s best features. For many years, the search engine has been tweaking its results to give you immediate data for common queries—type in “nba scores” or “current time london,” and it’ll give you the information right at the top of the results page, without you having to click anywhere. And last month, Google announced that it would be adding more “structured data” to its results; now when you type in “population new york,” Google returns a graph from the Census Bureau. The site even does math: Type in “2 + 2” or even something much harder—”speed of light/square root 3“—and the search engine solves it.
This is the eternal problem for any wannabe Google competitor. Wolfram Alpha doesn’t revolutionize search; at best, it adds a marginally useful new layer on top of it. But Google can easily co-opt such improvements—and suddenly everyone’s got a better Google. Maybe Google’s the real HAL 9000, sending the same message to anyone who dares to take it on: I wouldn’t do that, Dave.
Correction, May 21, 2009: This piece originally referred to the Web site Wolfram Alpha as “a savant, smart about a few things but profoundly ignorant about large swaths of human knowledge.” The correct term is idiot savant. Return to the corrected sentence.