By Sterling “Chip” Camden
Contributing Writer, [GAS]
Today marks the official launch date of Wolfram|Alpha, a “computational knowledge engine” from Wolfram Research. The stated goal of Wolfram|Alpha is “to make all systematic knowledge immediately computable and accessible to everyone.” Simplistically, you ask it a question and it gives you the answer – along with a lot of related information.
The engine stores curated knowledge of more than 10 trillion pieces of data. It contains more than 50,000 types of algorithms and models, using more than five million lines of Wolfram’s Mathematica symbolic language – and plans are to expand this knowledge base continuously. It’s no surprise, then, that Wolfram|Alpha excels at queries that are heavy on numbers.
For instance, typing fibonacci 43 gives you the the 43rd Fibonacci number. As with all queries, Wolfram|Alpha first displays how it interpreted your input, followed by the result. Then it gives you more information, such as alternative representations and truly useless stuff like how that number compares to the world population. By contrast, Google provides links to various sites that will give you some of that same information. Wolfram|Alpha apparently doesn’t know what to do with Perrin numbers, though, while Google provides relevant links for perrin number 12 (including one from Wolfram MathWorld, ironically).
Asking Wolfram|Alpha for mortgage $700,000, 5%, 30 years gives you a breakdown of monthly payments, total interest paid, and a full amortization schedule. The same query on Google gives you links to sites that might be able to answer your question, but a lot more who’d like to sell you that loan.
Unfortunately, a lot of everyday search queries aren’t based on numbers. Google ruby vs python, and you’ll get a ton of links to comparisons between the two languages. Oddly, Wolfram|Alpha thinks that you want to compare the two movies with those titles. At first glance you might think that the comparison of two programming languages would lend itself better to a quantitative analysis than the comparison of two movies. But the level of abstraction in programming renders the comparison of languages more of a qualitative discussion, while Wolfram|Alpha’s comparison of the movies sticks to the quantitative facts (cast, runtime, directors, etc.)
Forget about vanity searches on Wolfram|Alpha unless you’re someone important. My name was interpreted as a comparison between Sterling, Colorado and Camden, New Jersey. Did you know that it’s 1480 miles from my first name to my last? On Google, on the other hand, you have to get to the third page of results to find something that isn’t about or by me.
My Dad used to work for NSA, and he told me a (possibly apocryphal) story about someone who once asked one of the NSA computers, “Is there a God?” To which the computer replied, “Now there is.” Google naturally provides links to the many sites that discuss this question. Wolfram|Alpha humbly (or ominously) replies, “Additional functionality for this topic is under development.”
At least Wolfram|Alpha recognizes a request for 42 when it sees it. Google’s first two results give you essentially the same answer, but it follows those with discussions on the meaning of life outside the scope of the works of Douglas Adams.
Wolfram|Alpha’s ability to understand and parse queries is based on known vocabularies of more than a thousand domains. This, too, will grow – perhaps to some point in the future where it will understand most anything you can throw at it. For now, it has a limited ability to grok.
But Wolfram|Alpha has a bigger deficit than that, in my opinion. All of its knowledge, though curated for accuracy, is contained within its own database. You can often click on a link for source information, but that only lists a bibliography of sources, not direct links to the source for that specific answer. Thus, for information that is less than mathematically necessary, you don’t have the opportunity to evaluate its authority the same way you do with Google results. Google is therefore more appropriate for researching topics for which the jury is still out on some details. Perhaps as Wolfram|Alpha grows and learns, it will be able to incorporate notions of authority and opinion – or develop opinions of its own.
Here’s a screencast of Stephen Wolfram demonstrating some of Wolfram|Alpha’s capabilities.