Last month, the big brains of the “computational knowledge” world gathered in central London to explore how “advances in computational technology are unlocking knowledge assets and shaping the future.” The event, dryly named the London Computational Knowledge Summit, was underwritten by Wolfram Alpha, which bills itself as a “computational knowledge engine,” a tagline so catchy, they trademarked it lest it be stolen.
The brainchild of scientist, inventor, author and entrepreneur Stephen Wolfram, the Alpha engine was heralded as a Google-killer upon its debut last year. That was before everyone realized that it was more of a calculator than a search engine. (Those seeking confirmation of this will note the website’s favicon features an equal sign.) A year later, Google is expanding at a rate exceeding that of the universe itself, while Wolfram Alpha remains something of an online curiosity to anyone whose CV includes the phrase “liberal arts.”
Wolfram’s own résumé, predictably, reads like that of a boy genius just one romantic spurn away from becoming a supervillain. “Beginning in his teenage years,” an online bio exhorts, “Wolfram made a number of discoveries in physics and cosmology. In the early 1980s, his now-classic work on cellular automata helped launch the field of ‘complexity theory.'”
But wait, there’s more. After Wolfram grew up, he took the 300-year-old notion that laws based on mathematical equations could be used as a means of describing the natural world and turned it into software for modeling everything under the sun, over the sun and even in the sun. The killer app, known as Mathematica, is used for modeling phenomena in fields as diverse as engineering, biotechnology and finance.
So when Wolfram Alpha was launched in the spring of last year, lazy journalists and high school term-paper scribes rejoiced. Finally, the man who gave the world a plug-‘n’-play way to simulate chemical processes or test financial risk models had made a tool for the rest of us! Wolfram Alpha inhaled the web’s collective information into its own massive database and processes answers with an ever-evolving complement of proprietary algorithms borne of Mathematica software. What does this mean? The truth is out there, and now it’s in your iPhone, thanks to the Wolfram Alpha app.
Phrases like “knowledge extraction” might roll off the tongue of a character in Christopher Nolan’s Inception, but they’re not part of the general parlance. That’s, in part, what Wolfram seeks to fix, if not in name, then in deed. He’s empowering his engine’s users through “natural language processing” or, specifically, by letting them speak in intuitive human terms rather than some sort of computer-speak.
Remember the dark ages of the web when Boolean searches were all the rage? Such qualifiers aren’t necessary here; however, a fair amount of specificity is, especially when the data entered is partial or idiosyncratic.
While most human brains in the Bay Area “know” that Sonoma, Napa and Marin are counties, Wolfram Alpha only understands them as searchable terms and presents results based on a library of internal algorithms. Upon entering the query “Sonoma, Napa and Marin” for a comparative analysis of the counties, the engine assumed Marin was in Spain. Refining the query by adding “counties in California” yielded a comprehensive breakdown of the counties, and their statistical relationships to one another were presented side-by-side. The amount and range of information is beguiling, in fact overwhelming, which makes the aforementioned “knowledge extraction” a bit of a bear.
After a moment’s reflection on the data set, interesting observations begin to effervesce. There are nearly three times as many deaths in Sonoma County as in Napa County, though the population of Sonoma County is only double that of Napa County. Why? And where’s the data on Marin County? Perhaps we can infer that no one actually dies in Marin County, which accounts for its comparatively high real estate prices.
You see, inasmuch as the engine can slice a near infinite amount of information like a Ginsu blade, it can’t tell you what it means. The ability to infer, to extrapolate and perceive meaningful relationships between the data remains a strictly human occupation—at least for now.
Or, as one might say, you can lead a geek to Wolfram but you can’t make him think.
Daedalus Howell promotes knowledge extraction at