Posted on May 18th, 2009 by David Bayon

The people vs Wolfram Alpha

Since Wolfram Alpha launched at the weekend, I’ve lost count of the number of articles I’ve read in which the author asks it inane questions and laughs when it falls flat. Even our own Darien Graham-Smith (along with several others in the office) seems almost delighted to prod and poke at it to find instances where Wolfram’s big pre-launch claims can be mocked – usually by comparison to Google or Wikipedia.

Unfortunately, this is something that was bound to happen given the publicity the site has received in recent weeks from the mainstream press. The big problem occurs because most people are attempting to hastily test the new engine without any real reason to be using it.

Without a specific scientific task in mind, most are instead racking their brains for the everyday queries they usually ask a search engine – or, worse, for deliberately obscure trivia facts – and are thus failing to use Alpha as it will surely be used once the first few months of fuss have passed: i.e. probably not by them.

The team have been quite open about the early stages of the project. The majority of critics would do well to note FAQ comments such as this: “We’ve first emphasized areas where computation or mathematics have traditionally had a more significant role, or where knowledge is more readily quantitative.”

Or this: “How much data does Wolfram Alpha have on popular culture? Basics, particularly about more computable issues, such as movie box-office grosses. More is being added.”

Of course, the winners of Eurovision are unlikely to have been high on the ‘computation or mathematics’ priority list, even less so as it’s an American team compiling the vast amount of data. To scoff and cite a Wikipedia page full of trivia as evidence of Wolfram Alpha’s failings is to miss the point of the tool completely.

It’s not meant to be Google. It’s not meant to be Wikipedia. If you can find out the biggest selling single of the nineties with a quick Wikipedia search, why on Earth wouldn’t you? On the other hand, if you search for the relationship between Microsoft and Apple on Wikipedia and Google your immediate result won’t look anything like this: (click to enlarge)

Be realistic about what Wolfram Alpha is. Instead of transplanting your day’s Google searches into it and mocking the lack of success, stick to those tasks for which, without all the media fuss, you would ever have even thought of using Wolfram Alpha. Google and Wikipedia will still be there for everything else.

Tags: Google, wikipedia, Wolfram Alpha

Posted in: Rant, Software

Permalink | Trackback

Follow any responses to this entry through the RSS 2.0 feed.

9 Responses to “ The people vs Wolfram Alpha ”

Daniel Says:
May 18th, 2009 at 4:52 pm
Thank you for this! I couldn’t agree more. Why would anyone want to compete with Google when they already do an acceptable job as it is. This is a new kind of tool- one that I think appeals more to those who are looking for quick quantitative information (like the nutritional value of an egg). For my math/science needs, I’ll be using Wolfram|Alpha. For everything else, I’ll probably stick with Google.

However, I saw on Newsy that although this is no competition for the original Google, there may be some competition between Wolfram and the not-quite-in-beta Google Squared, which makes more of an effort to organize their results. So we’ll see. I still think I’ll be using Wolfram|Alpha.
Robert Says:
May 18th, 2009 at 6:50 pm
I will point out that Wolfram Alpha says that the series n^(-1.5) diverges.
David Says:
May 18th, 2009 at 7:25 pm
Thank you for posting this intelligent observation. I would also recommend perhaps sharing this insight with some of your colleagues. This other PC Pro article, for example, is an embarrassment: http://www.pcpro.co.uk/blogs/2009/05/18/9%c2%bd-things-wolfram-alpha-doesn%e2%80%99t-know/
Bobbie Says:
May 19th, 2009 at 1:18 am
“If you can find out the biggest selling single of the nineties with a quick Wikipedia search, why on Earth wouldn’t you?”

Because I would like to use the API. For example that query could be used to make a youtube mashup to play music videos for the biggest selling single of the nineties
Alex P Says:
May 19th, 2009 at 9:26 am
Could google list results for Wolfram Alpha? That would make it a one-stop-shop…
Ant Evans Says:
May 19th, 2009 at 3:03 pm
Just ask the pope.

benedictxvi@vatican.va
Lise Says:
May 20th, 2009 at 10:12 am
You have an excellent point, but then I’d question why the launch has taken place in the way it has – lots of public fanfare on general news channels rather than a quiet release into the scientific community? Headlines in the daily broadsheets like “Can Wolfram Alpha take on Google?” and “‘Google killer’ Wolfram Alpha gets public demo” surely aren’t going to help the public perception of WA as a scientific tool.

I’m looking forward to WA becoming a useful thing, but at present I’ve found some of the same data-shortage problems reported elsewhere. It’s certainly fine if you want to compare stocks and shares – WA knows more about Revlon sales figures than I could ever have imagined – but try to compare other data sets and it does repeatedly run into the problem of not knowing what they are yet. Which makes me wonder why it was launched, so publically and noisily, through the channels that it was.
Steve Cassidy Says:
May 21st, 2009 at 8:33 am
Nicely put, everyone. I would say that the right test for Wolfram Alpha isn’t giving it Wikipedia queries about Britney Spears, Poschiavo, or Thrush(es). The test is, take Wolfram’s example queries, and put them into Google. What happens then?
Bill Maslen Says:
May 21st, 2009 at 10:00 am
Hm. I’ve given Wolfram Alpha a pretty good thrashing, using a wide variety of questions, and I’d have to agree that the interpretation algorithms need a lot of refinement. Certain types of question it deals with very well (e.g. “distance between Earth and Mars”), others it answers ‘intelligently’ (e.g. “not enough information is available”), but an awful lot of fairly straightforward questions get the “not sure what you mean” response. I suppose you could say it’s the opposite philosophy to the Google “if you’re not sure what they want, just give them the kitchen sink” approach, so from that point of view, represents an evolutionary step that won’t necessarily succeed on its own, but points the way to a more refined search model that will eventually include parameters that can be tweaked in far more sophisticated ways than the Boolean options available on Google will currently permit. From that perspective, it’s actually rather exciting. From the jaded, overhyped consumer’s perspective, it’s sadly rather disappointing.