Some have billed it as the next Google, but Wolfram Alpha is much more than that
The best-kept secret in computer science is out. Stephen Wolfram's new search engine, Wolfram Alpha, will launch to the public later this week after years of covert development. It seems certain to make a splash. Instead of searching the internet and retrieving related documents, Wolfram Alpha will attempt to deliver precise answers to a broad range of questions. Wolfram calls the project "tremendously ambitious". Industry experts have already begun hailing its genius. The British physicist says he's been planning the project for many years. "Once a decade I wondered whether it was worth trying, but I was not at all sure that it could work."
Wolfram Alpha does indeed work, but it's far from being the finished article. The version released to the public will be a work in progress, which means many searches will result in the phrase, "Wolfram Alpha isn't sure what to do with your input." Wolfram says his team of around 100 researchers has managed to glean and log around "90 per cent of the shelves in a typical reference library." Much of the data is US oriented, although Wolfram insists his team tried to prioritise data that's available internationally. "All the socioeconomic data is from international sources," he says. "We've not yet started data acquisition from country to country."
Wolfram would like his software to be known as a "computational knowledge engine" rather than as a search engine. It does not actually search the Web for information. The software references in-house databases to work out answers to factual questions. For example, the distance between the earth and the moon at the time of asking, which is variable, or the probability of getting ten heads from 15 coin tosses. Unlike Google, it won't be able to tell you where to buy a suit in central Manchester, but it will be able to compute the value of tailoring to Manchester's economy and compare it to other cities around the world. "This is not delivering things that have already been written down," says Wolfram. "It is not concerned with facts, it's about what we can compute."
Type "life expectancy France UK" into Google and the software tries to find pages on which those words occur together. If it can't find the words together it serves pages on which the words are close, but can't sift out potentially useless information, such as the life expectancy for other countries, or pages that contain all the words but have no relevance to your search. By contrast, Wolfram Alpha "knows" that France and the UK are separate entities and can calculate and display life expectancies in those two countries, plus other potentially useful information, such as population density and the median age of the two countries.
Wolfram's own computational software, Mathematica, is embedded in the program, potentially allowing an unlimited number of questions to be computed from the database. But it's not just the program's computational skills that are causing a stir. It's the engine's capacity to comprehend plain English. "Linguistic prophesy," says Wolfram, "is the program's ability to understand freeform language." It uses a form of common sense, too. Type in "Microsoft Sun", and the software will work out that you mean to compare Microsoft to another IT firm rather than the star: obvious to a human, but a very difficult trick to achieve through a computer.
Google is said to be working on a semantic version of its software that would allow the search engine to comprehend the terms it is searching for (it already offers a degree of understanding with its "related searches" links). But until this is completed, Google remains dumb by comparison. "Wolfram Alpha is like plugging into a vast electronic brain," say Nova Spivak, founder of the Web tool Twine. "It provides extremely impressive and thorough answers to a wide range of questions asked in many different ways, and it computes answers, it doesn't merely look them up in a big database. In this respect it is vastly smarter than (and different from) Google."
Much of the data from Wolfram Alpha is in real time. Earthquake information, for example, will be updated every few seconds. But real time feeds, says Wolfram, contain their own challenges. If you want to know how windy it is in a certain region and you find the wind speed is zero, "is it really that the wind has dropped or is it that the device that measures wind in that area has got stuck?" Problems like these, and more, will be ironed out once the program has gone public, Wolfram says. The team is working to improve the depth of data as well as its quality. "We hope to have a formalised system for people to contribute data," says Wolfram.
Wolfram Alpha's may be very different to Google. The two companies are after all trying to do two different things. But the business model's are close. Wolfram wants to give the software to the public for free, hoping to attract advertising. He also plans to release a pay version for businesses. Stephen Wolfram's moral purpose seems closely aligned with that of Google's founders Larry Page and Sergey Brin. Indeed, Brin spent one summer interning for the UK phycisist. All three are concerned with mankind's pursuit of enhanced knowledge. "Our goal is to bring expert level knowledge to everyone," says Wolfram. "With things that were once only interesting to experts, if you package them right they become of interest to everyone."
Posted 11 May 2009 : Director.co.uk
