Wolfram Alpha - The Software Equivalent of Rainman
I played around with Wolfram Alpha today and I'm mostly underwhelmed. I tried putting in my birthday and that left me with the rainman impression - there are lots of numbers in there, but they are mostly useless and provide no appreciable context.
For instance, we know that MLK was assassinated the week following this date and that Bobby Kennedy was assassinated a couple months later, yet neither of these dates show up in relation to March of 1968. Further, the Tet Offensive in the Vietnam War occurred from January to September of that year and yet Wolfram Alpha has no idea that March of 1968 is in the range of dates defined by the Tet Offensive. That is, Wolfram Alpha's query engine seems to be totally oblivious to the fact that a day in history, without the kind of historical context provided by the significant events of the time, is practically indistinguishable from any other day in history. Wolfram Alpha proceeds to dissect the day in microscopic detail, as if counting ants is a productive exercise, when it should be discussing the broader implications of ant colonies.
Similarly, if I ask how many carbohydrates are in a can of coke, I get the answer 14 grams, which I'm certain is simply wrong. Looking at the details reveals that this is the average carbs in a can of coke where can of is defined as everything from diet soda to the canonical can of coke classic. Nevertheless, the upper bound of the answers provided by Wolfram Alpha is 35.56 grams. The correct answer is around 40 grams. Worse, for all its marketecture, Wolfram Alpha seems to suck at math. Examine the nutrition label in the link above and we find that total carbs = sugar + dietary fiber. Thusly, 14 = 13 + 0. Back to the drawing board.
Finally, if we look up a song title, I Remember, we find that Wolfram Alpha has associated this with the film Amarcord, by Federico Fellini. Why? I have no idea, the search term doesn't even appear anywhere in the film's metadata. Worse, there's no disambiguation page ala wikipedia.
Which brings me to my last point: Wolfram Alpha is a shining example of how not to develop software products - put a bunch of brainiacs in a room and let them go crazy, then wonder why no one can get any useful information out of the product they produce.
Consider that when Wolfram Alpha returns a data set associated with a query, the metadata are displayed in a form in which every word on the page is a link to - wait for it - a popup box allowing you to copy the linked text to the clipboard! It's not like I might have wanted to, oh, let's say, drill down into the data from the macro to the micro perspective or link to another authoritative source on the data.
While I find the mathematical aspects of Wolfram Alpha to be interesting, Wolfram Alpha's use as an information query and visualization tool is hampered seriously by a complete lack of vision as to how to go about exposing the data in meaningful ways. It's somewhat ironic that these are treated as problems with no apparent solutions given that the research is both deep and aged.
For instance, we know that MLK was assassinated the week following this date and that Bobby Kennedy was assassinated a couple months later, yet neither of these dates show up in relation to March of 1968. Further, the Tet Offensive in the Vietnam War occurred from January to September of that year and yet Wolfram Alpha has no idea that March of 1968 is in the range of dates defined by the Tet Offensive. That is, Wolfram Alpha's query engine seems to be totally oblivious to the fact that a day in history, without the kind of historical context provided by the significant events of the time, is practically indistinguishable from any other day in history. Wolfram Alpha proceeds to dissect the day in microscopic detail, as if counting ants is a productive exercise, when it should be discussing the broader implications of ant colonies.
Similarly, if I ask how many carbohydrates are in a can of coke, I get the answer 14 grams, which I'm certain is simply wrong. Looking at the details reveals that this is the average carbs in a can of coke where can of is defined as everything from diet soda to the canonical can of coke classic. Nevertheless, the upper bound of the answers provided by Wolfram Alpha is 35.56 grams. The correct answer is around 40 grams. Worse, for all its marketecture, Wolfram Alpha seems to suck at math. Examine the nutrition label in the link above and we find that total carbs = sugar + dietary fiber. Thusly, 14 = 13 + 0. Back to the drawing board.
Finally, if we look up a song title, I Remember, we find that Wolfram Alpha has associated this with the film Amarcord, by Federico Fellini. Why? I have no idea, the search term doesn't even appear anywhere in the film's metadata. Worse, there's no disambiguation page ala wikipedia.
Which brings me to my last point: Wolfram Alpha is a shining example of how not to develop software products - put a bunch of brainiacs in a room and let them go crazy, then wonder why no one can get any useful information out of the product they produce.
Consider that when Wolfram Alpha returns a data set associated with a query, the metadata are displayed in a form in which every word on the page is a link to - wait for it - a popup box allowing you to copy the linked text to the clipboard! It's not like I might have wanted to, oh, let's say, drill down into the data from the macro to the micro perspective or link to another authoritative source on the data.
While I find the mathematical aspects of Wolfram Alpha to be interesting, Wolfram Alpha's use as an information query and visualization tool is hampered seriously by a complete lack of vision as to how to go about exposing the data in meaningful ways. It's somewhat ironic that these are treated as problems with no apparent solutions given that the research is both deep and aged.
posted by David Watson at
9:24 PM
0 Comments
Links to this post
