Skip to main content
Publishing

Publishing

‘Knowledge engine’ hits the Web

18 May 2009

Here’s a little test: if an electron beam were coming towards you with an energy of 1 GeV, how much lead would you need to stop it?

It’s not an easy question to answer. Even if you knew the formula that links the kinetic energy of electrons to their stopping distance in lead, you would have to do a little maths before you found the correct result. But now there might be an easier way: go to Wolfram Alpha, an internet “computational knowledge engine” that officially launches today.

The brainchild of Steven Wolfram, a British physicist and founder of Wolfram Research in the US, Wolfram Alpha has a goal “to make all systematic knowledge immediately computable and accessible to everyone.” Unlike search engines it gives a single, straight answer to a query, and unlike user-edited encyclopaedias, such as Wikipedia, it is almost totally reliable — or so the makers say.

“We’re not going to claim that we’re perfect, but we want to be a very credible source,” senior developer John McLoone told physicsworld.com. “We won’t allow people to arbitrarily upload data into it…So we don’t have to worry about Wikipedia’s fundamental problem, which is whether you can trust what’s in there.”

Four pillars

The idea is simple enough. All you do is type your question into the input field, press enter, and — usually — out drops the answer. There is information on practically any topic, although there tends to be more on the sciences. So you can ask anything from “What is the median household income in Wisconsin?” (about $46,000) to “Where is the International Space Station now?” (orbiting just over the Philippines). If you’re still interested, the range of a 1 GeV electron in lead is, apparently, 29.72 mm.

The site rests on the workings of four distinct “pillars”. The first pillar comes in the form of curators, who have scoured data that are freely available in the public domain, cross-referenced them and put them in the same format. Then there are experts in different fields, who have coded the data with useful meanings. On top of that are numerous algorithms for presentation and, finally, interpreting the syntax of the input.

One of the key things about Wolfram Alpha is that it will generate answers that don’t exist right now John McLoone, Wolfram Research

But, according to McLoone, none of it would have been possible without “Mathematica”, the computational software Wolfram Research has sold and developed for 20 years. “It handles all of the web-page production and parallelization needed to scale up to the size of site we have,” he explains. “More importantly, there’s all of the data manipulation, analysis and computation that’s layered on top of the raw data.”

‘Google killer’?

Some blogs have already referred to the knowledge engine as a “Google killer”, although McLoone prefers to see the site tackling a unique market. “One of the key things about Wolfram Alpha is that it will generate answers that don’t exist right now, which is where a search engine can’t possible help,” he says. “If there isn’t a page with the answer written on, you can’t search for it.”

The system is not foolproof, however. For example, if you type, “How long does it take to get from Earth to Mars at 100 m/s?” it assumes you are referring not to the Red Planet but to the tiny borough of Mars in Pennsylvania, US, and returns the dubious answer “1 day 1 hour 30 minutes”. Or type “Plot y > |x + 1|” and the plot comes up correct but the axes display some jumbled scales.

Anthony Laing, a PhD student at the University of Bristol, UK, says he tried some specific questions in quantum physics following the instructions on the site but got nothing back. “I think it needs filling up, so to speak,” he adds. “It’s a nice idea though and I’m interested enough to keep trying it from time to time.”

Wolfram Alpha is free online, but there will be “professional” versions which, for a monthly fee, will give users access to greater computing resources and the ability to upload and export their own data. The project is also being funded by partners and sponsors, who are presently anonymous.

Copyright © 2024 by IOP Publishing Ltd and individual contributors