Back in March, Radar Networks CEO Nova Spivack blogged about Wolfram|Alpha, a new Web service from Mathematica creator Stephen Wolfram, and said it could be as important as Google. Nova’s a smart guy, but it was reasonable to greet his enthusiasm with skepticism–new stuff gets favorably compared to Google all the time, and often turns out to be massively disappointing. But Wolfram|Alpha is now live–albeit in a preview mode that’s struggling to keep up with the sudden influx of users–and Nova’s assessment turned out to be cool and collected. The service is a work in progress, but it’s the most interesting new research tool since Wikipedia–and yes, it’s not unreasonable to discuss it in the same breath as Google.

It is, however, a little more complicated to explain. Wolfram|Alpha’s tagline is “Computational Knowledge Engine,” which reflects the fact that it consists of a gigantic database of facts–math and scientific matters are a specialty, but it also knows about world history, geography, movies, music, stock prices, and a lot more–and has a semantic understanding of the information therein. Which means that when you search for something on Wolfram|Alpha, it responds not by providing Google-like links to other sites, but by attempting to understand the meaning of your query and assembling an answer from its own repository of knowledge.

It’s far easier to explain all this with examples. Like these ones:

The less specific your query, the more likely Wolfram|Alpha is to give you a surging sea of relevant information, including things you never knew that anyone knew, which it’s computed for you:

Over at Wolfram|Alpha, there are scads more example queries on topics of all sorts. The notion of a Web research tool delivering results rather than links to other sites isn’t new–I remember a promising, ill-fated startup called FactCity that tried to do just that–but nobody’s ever come close to what Wolfram has pulled off here.

Unlike most Web debutantes, this one is emerging with its own extremely distinctive personality. It’s wildly ambitious, nerdy, and a tad self-congratulatory. (Offhand, I can’t think of another major Web entrepreneur other than Wolfram who chose to name his brainchild after himself–I think I’ll just call it “Alpha” from here on out.) In the words of John Hodgman, it often provides more information than you require–in the above query about LBJ’s death, I can’t quite tell if the fact that it provides the current population of Johnson City means that it’s showing off, or that it’s not quite smart enough at this point to know that you probably don’t care.

Alpha is also elegant (the presentation of results sport some of the nicest Web typography and infographics I’ve ever seen). Most of all, it’s both astoundingly useful and fun to use.

It’s not, however, perfect. Google circa 2009 will give you something on any imaginable query. But Alpha’s knowledge base, though apparently vast, isn’t all encompassing. It often failed to understand my queries, though it sometimes understood them well enough to tell me that relevant information would be added in the future:

And sometimes even when it knows something, it can’t extract that information in ways you might expect it could. For instance, it knows the cast of Raiders of the Lost Ark, but this query failed:

Some people are more bothered than I am by the fact that Alpha’s knowledge of the world is not yet exhaustive. I’m more frustrated by the fact that not all of the tidbits it knows are hitched up to related pieces of information. Search for a movie, and it’ll tell you the cast, but search for an actor or actress, and it doesn’t tell you what he or she has been in:

Alpha also doesn’t make much use of hyperlinking; it would be even cooler if, for instance, the reference to Carrollton, Illinois in the above results could take you to more information about it.

How reliable are Alpha’s facts? Unlike Google, it’ll never send you to a site that’s schlocky and inaccurate, and unlike Wikipedia, it’s not dependent on volunteers getting things right. The “Source information” links at the bottom of results give you some general information on where it gets its data, most of which comes from standard reference works. By Web standards, it should be pretty darn trustworthy, and it’s all factual. (It’s pointless to ask Alpha questions which involve matters of opinion.) I only ran across one area where I’d quibble with Alpha’s sources–when you search for a domain name–such as Technologizer.com–it quotes traffic stats from Alexa. And for every site I’ve ever had personal involvement in, Alexa’s numbers are useless gibberish.

I don’t mean to suggest that Alpha is in any way a disappointment–at least once it stops responding to some queries with this Failwhale-like error message:

Actually, for a service that’s just a few hours old, it feels surprisingly mature. (It’s sure more evolved and polished than Google was when it first started to get good buzz a decade ago.) Along with Google and Wikipedia, it’s going to be one of my daily research tools from now on. And I can’t wait to see where it goes. Stephen Wolfram’s Wolfram Research has been working on Mathematica for more than two decades, so it’s not unreasonable to think that Alpha could go places that this first version just hints at.

 
11 Comments


Read more: Search Engines, Wolfram|Alpha