What is Wolfram Alpha?
Wolfram Alpha promises to be a “new kind of search engine”. Developed by British physicist Stephen Wolfram, best known for his work on the computational mathematics system known as Mathematica. Wolfram Alpha is set to premiere in May 2009 as an “answer engine”, meaning it will answer questions rather than searching for generic terms. Users may be somewhat familiar with the concept of the “answer engine” as Google can handle some basic question and answer formats — try typing in “What is 2 plus 2?” in Google and you’ll see what I mean. Wolfram Alpha promises to answer “How”, “What”, “Why”, and many other types of questions on top of providing facts and data related to user’s searches.
Wolfram Alpha differs from traditional search engines in that it goes beyond simply returning a list of results a la Google or Yahoo. Working outside of the realm of “keywords”, Wolfram Alpha will compute answers and provide “relevant visualizations” from the world wide web’s massive collection of info and data. While there have been other new search engines and attempts at creating “answer engines” — most notably TrueKnowledge and Powerset (developed by Microsoft). These types of search engines are known collectively as “semantic search engines”. To put it in layman’s terms, “semantic search engines” index a big batch of answers, and then try to match the user’s question to an answer. Its like the search engine is playing Jeopardy.
Wolfram Alpha is very similar to a project called Cyc, which itself is aimed at developing a common sense inference engine and has been under construction since the 1980s. Unfortunately, after all this time, there is nothing to show for Cyc, as there has been no commercial release or beta test. The founder of Cyc, Douglas Lenat, was one of the few outsiders given access to an early version of Wolfram Alpha before its major release next month. According to Lenat’s article at Semantic Universe, a blog dedicated to semantics research, Wolfram Alpha ” . . . handles a much wider range of queries than Cyc, but much narrower than Google.” Lenat goes on to say he was very impressed with Wolfram Alpha, but that it has its limitations, as is to be expected. Lenat says “[Wolfram Alpha] understands some of what it is displaying as an answer, but only some of it.”
So how does Wolfram Alpha work, and what do its results look like? It is difficult to say, as only a handful of people have been allowed access so far. Looking at interviews with the BBC from Lenat, Wolfram himself, and a few other cyber geeks, we can get a basic idea of his this “new search engine” might work.
Here’s an example. If a user were doing research on the battle casualties in the second World War, they may type in something like “WWII deaths France / Germany”, Wolfram Alpha would calculate and return a graph or chart of the basic fraction of France’s war deaths to Germany’s own deaths during that war. If you just type in “world war II deaths”, without specifying a country, Wolfram Alpha establishes what country you may mean by looking up your local web host and displaying the information related to your own country — it assumes you’re asking about that which is most important to you — information about your own country. Not only would you get a nice chart on the topic of world war two deaths, but Wolfram Alpha would include various pieces of information about deaths during the second world war, and other related mathematical information — you see, all of Wolfram’s knowledge comes from a mathematical perspective and not from a more “semantic” one. It is important to remember that Wolfram Alpha does not have a database of knowledge, so what it knows about death during WW II is limited to the math involved.
This distinction is important — don’t expect Wolfram Alpha to (immediately) compete with Google for basic web search functions. Google will respond to just about any query, “like a faithful puppy bringing in the morning newspaper” as Doug Lenat puts it. Google returns results without any kind of “understanding” of those returns — Google simply spits back what it finds without attempting to figure out if the results match your question. This leads to sometimes hilarious results — Google will display search results and “ads” that may include all kinds of false “answers” and leads. I remember one personal example — I was searching for a recipe for Pralines on Google, and I got a ton of results for a German exotic dancer calling herself “Praline” and offering all sorts of services and photos that were more than inappropriate to my query. At the other extreme of this issue are other “semantic engines”, which only can answer a very specific line of user searches, but provide mutiple answers that are directly related to that query. Wolfram is attempting to fill the void between these two extremes — relevant results for a wider selection of queries.
The problem with web searches is that they require something computers have a difficult time with — “common sense”. Users want search results where every piece of the search query and every piece of all of the answers provided by a search engine are as deeply understood as mathematics. Unfortunately, this becomes difficult — try teaching a computer to differentiate between a food recipe and a stripper’s personal ads and you’ll end up banging your head against the wall. Wolfram Alpha is an attempt to create a search engine that can tailor results to an individual search query, not just to all queries involving the word “praline”. Doug Lenat provides an example of this that he used to ‘test’ Wolfram Alpha’s “common sense” — he says if you ask for things that happen at 10 centimeters per year, Wolfram can tell you the fact that hair grows at this rate. But if you ask Wolfram “how fast does hair grow?” it will not be able to tell you. Unfortunately, Wolfram Alpha still depends on the user asking “the right question” — which is often the problem with Google. To put it another way, if the connection between two facts isn’t or can’t be represented by a mathematical equation, Wolfram Alpha is convinced that this connection simply doesn’t exist at all.
Not yet, anyway. Remember that Wolfram Alpha is very much in its “Alpha” stage — Dr. Wolfram is aware of the limitations of his system, and is sinking plenty of itme and money into future plans for addressing these issues as Wolfram Alpha continues to be developed.
Will Wolfram Alpha compete with Google for search traffic? Not right away. Is it a “new search engine” — probably. Until Wolfram goes online to the general public sometime in May 2009, many of our questions will be as up in the air as my search for the perfect Praline recipe.