A Look at Wolfram|Alpha: An Answer Engine, not a Search Engine

A Look at Wolfram|Alpha: An Answer Engine, not a Search Engine
Page content

Stephan Wolfram has been a physics prodigy, a successful designer and creator of the innovative software Mathematica, an author of a book that

turns the complexity of the way we view nature on its head by reducing it to component calculations, and now, building on all of these things, he is introducing a new search engine. Or is it? What he calls it is a Computational Knowledge Engine. It gives answers.

Picture of Stephen Wolfram from Wikimedia Commons/Source.

What is Wolfram|Alpha Meant For?

Wolfram|Alpha, as the new engine is called, is not a search engine in the sense of Google or Yahoo. It does not go through the millions and millions of pages it has indexed and give you what its algorithms say is most likely to be the best match for the keywords you typed into the query box. Instead, it can parse the query you input in natural language, and answer that query- if the elements are in its curated database.

Google doesn’t make judgments on the quality of data in the matches it gives you to a search. If it indexes information written by someone who is ignorant of the facts, the results are still going to pop up in your search. Wolfram|Alpha has had about 100 people working on validating the data and equations they put into the engine for nearly ten years, and recently the number of data curators jumped to 250. Wolfram estimates it may take a thousand to evaluate and feed data to the engine in the future, to start making inroads on the areas the engine does not cover and to deal with new information.

Unlike Wikipedia, not everyone will be able to go in and enter data, but there will be web forms to request material be added. And unlike Google, not everything published on the WWW that can be indexed will find its way there either. You also will probably not find 300 different answers to the same question, because the material is curated, not popular opinion.

Chances are that someone interested in pornography will not get results. I have no idea if he intends to put genealogical material into Wolfram|Alpha. It actually might make it the engine of choice for genealogists, amateur or professional if he did, because their answers would be so much more precise than Google. On the other hand, Google is likely to keep indexing every ancient document newspaper and microfiche that it comes into contact with, so some offhand reference to an ancestor will still be more likely to turn up somewhere on Google- if you have the patience to go through the 1000 possible matches Google gives access to for every search.

On Page 2, we will look at some examples.

Examples

Ask: What is the calorie count of 6 pieces of asparagus with a teaspoon of butter? Google is likely to find material that gives the calorie count for a portion of asparagus, and then you need to find out what a portion of asparagus means. If you are lucky, the same document also mentions the calories in a teaspoon of butter. Wolfram|Alpha, if it works as advertised, will take the calories for a single piece of asparagus, multiply by six, and add the calories for one teaspoon of butter.

Front page of Wolfram|Alpha site

Wolfram Alpha front page

Or to use an example they have given out: What is the distance from the earth to the moon? Google will find you multiple documents, giving the average difference, and Wolfram Alpha will look at the moon’s position at the time you ask today, and calculate the exact distance at this moment.

This is somewhat like the difference between a buffet dinner with 35 selections put out for you to choose from, and a stir-fry at the Mongolian barbecue, where you select the ingredients and it is cooked in front of you, giving you exactly the combination you requested. They each have their advantages. Google is a net, pulling up the matches for the keywords you selected, and Wolfram|Alpha is a spear gun aiming for a particular fish. Of course, sometimes the spear gun has blanks in it.

What is in it?

Stephan Wolfram admits that there is a lot of information missing from his engine. He estimates that he has 90 % of reference material (in English, I assume) curated into his engine. From examples available, I think he only means certain types of reference material. It doesn’t mean he has every textbook on anatomy and physiology there. He says anatomy is currently one of the areas where the engine is weak. However, I expect the entire contents of the Handbook of Chemistry and Physics are there, and very possibly the Merck Manual. What he has is a vast collection of hard data, and a vast number of formulas.

He does have some built-in evaluations for ambiguous queries. He uses the Wikipedia popularity index. Wolfram|Alpha should be able to tell fifty cents from Fifty Cent. However, those who have had advanced access to it do say it just couldn’t figure out what they meant for a number of their queries.

Wolfram has said that Google can find an answer to a question, if someone has asked it, and the answer is published on the Internet. Wolfram|Alpha can take a question no one has ever asked before, and if the discrete data for the question is in his engine, and he has the formula, it will give you an answer that is a new piece of knowledge. He also has spent a lot of time with presentation of his answers. You may get a line graph or a histogram. He also makes it easy to drill down in the answers, or go off on a tangent without creating a whole new query.

Perhaps there may be little applicability for everyday life, especially with the current holes in the data he has now. However, I think the potential is there, when the data is there, for it to become a handy tool.

Wolfram|Alpha went live, briefly, before May 2009. The news shot around the globe. Here is a screenshot on their Blog of where queries were

Wolphram Alpha Blog live test world picture

coming from around the world.

Picture from Wolfram|Alpha Blog

On Page 3 we look at possibilities for using Wolfram|Alpha.

Everyday possibilities for use

A person on a 1850 calorie a day diet will be able to figure out their own meal plan, with foods they enjoy, and find out how much of what they can eat to meet that goal. I’d find that an attractive alternative to a canned plan from Weight Watchers or Jenny Craig.

Right now, I’d like to know if his engine could tell me how many Jelly Belly jelly beans could fit in a quart glass jar. When I search on Google, I find that people have asked the question before- on answerbag.com, among other places, but no one has posted an answer. Other matches on Google give ideas for formulas to find the answer- and there are answers for how many Jelly Bellys in an ounce- but a simple answer does not pop up.

If you ask how many Jelly Belly beans would fit inside the Statue of Liberty, you’ll find Google doesn’t seem to have a match with anyone who has asked the question before. (Strange mind, I know…) All of the matches in the first 100 choices that came up were mashups of pages. Of course, I had trouble trying to find data on what the interior volume of the Statue of Liberty is, so Wolfram|Alpha may not be able to answer either. I wonder, though, how many Jelly Belly beans would it take to fill the interior volume of the Great Pyramid at Giza?

Maybe the average person will find themselves using it most often for trivia. But sometimes they could find it handy to just ask “How much will it cost to ship a 16” x 20” x 6” package weighting 7.5 pounds from an address in zip code 10001 to 27 Main Street, in zip code 13607 by UPS ground? That sort of data is likely to end up in Wolfram|Alpha, and it means you don’t need to go to the UPS site, enter the weight into a field, figure out if your package dimensions move you into a different class, and then have them calculate the answer.

It will let you do certain useful calculations in one easy step. If you are building your child a tree house, you can find out whether you should get half inch plywood or three quarter inch plywood to safely hold the weight of 5 100 pound kids. Not only that, it will probably give you a graph showing how much weight either thickness of plywood will hold.

Possible Academic and professional use

But the real value of the answer engine is currently for students, scientists, professionals and technical people. The engine already holds a vast amount of information about diagnoses and disease. Once it can take input form a physician about symptoms; If you are a GP and have someone come into the office with flushed cheeks, easy bruising, high blood sugar, twitches and a severe ache in the lower left quadrant of their abdomen, you won’t need to wish you had access to Dr. Gregory House. You could type in the information to Wolfram|Alpha, and find out matches it has for those symptoms, and the additional tests to get an exact answer. You can also get suggested treatment. It might make a real difference to how well someone will recover, based on how quickly you can begin treating them, with an indexed medical database at your command, and do almost instantly what might be hours of tedious searching for matching symptoms.

Civil engineers and architects could feed in information about un-built structures and desired building materials, and find out what materials have the tensile strength for their needs. Chemistry professors could find out the exact supplies and amount of each chemical they need on hand for a particular class to carry out an particular experiment.

And students can relate information they understand to other pieces of information.

Wolfram|Alpha today

Well, Wolfram|Alpha still does not know much about plywood, disappointingly enough.

However, there are queries which give answers that make information come alive for the searcher.

I asked about the gravitational pull of the moon.

It gave the standard answer, 5.328 ft/s^2 (feet per second squared)

It also gave some comparisons that make those numbers seem real and immediate.

~~ 0.43 x acceleration from 0 to 60 mph of a 2001 Jaguar XK8 (~~ 3.8 m/s^2 )

~~ 0.44 x maximum acceleration during the launch of a Saturn V rocket (~~ 3.7 m/s^2 )

~~ 0.85 x acceleration from 0 to 60 mph of a 1983 Buick Century (~~ 1.9 m/s^2 )

Many of us have driven or ridden in a Buick. We may not have ridden in a Jaguar, but we have seen one on TV or in a magazine.

We can now relate that to the force the moon has on the ocean- and how that causes tides.

Even more interestingly, it seems that a 2001 Jaguar XK8 has more power than a Saturn V rocket did. That is awesome.

Using Wolfram|Alpha today

Wolfram|Alpha, as Jonathan Wylie points out in his article the Top 10 Search Engines for Students, can do things other search engines cannot. It brings computational power into the hands of students, and can correlate multiple pieces of information in its answers, giving them new avenues of research.

So- will it be another Cuil, or not?

I think, if enough facts and equations are available in the engine, it will insinuate itself into daily life in much the same way Google has, but for different purposes. They can be a complement to each other- and perhaps even allies. Wolfram Alpha can pull in certain types of current information from Google. And Google could use Wolfram Alpha to give a best match to some of the queries it gets.

Not Cuil, but not yet cool

Wolfram|Alpha has developed quite a bit in the nine months since it went live. However, it still has a long way to go. It can dazzle with facts, but I still believe its real strengths are not being utilized. It needs to be able to make more connections. It need to pull the information in it into combinations created by the user, and the links are not there yet. Still, it holds a lot of promise. It is a new way of providing information- just not all the information I want to see it give.

Wolfram|Alpha and the Semantic Web

The Semantic Web is a popular term now, and it involves all data on the WWW having tags to fit into ontologies (a formal way to organize a particular type of data, in a hierarchy of relationships). It is going to take a long time to do this, even if all new data in the WWW starts being tagged as it is entered. In some ways, Wolfram|Alpha, with its curated data, has done an end run around that method. It gives you much of the value of a semantic web immediately, at least for the data in Wolfram’s engine. Stephan Wolfram has said that if the Semantic Web was here, now, it would make the task of Wolfram|Alpha much easier. Of course, the Semantic Web is expected to tag a number of things that won’t be in Wolfram|Alpha. When more of the WWW has become part of the Semantic Web, we may need a third type of engine to use it efficiently.

May 18th, 2009

May 18th was both a success and a disappointment to many users. Upon going live, there were server tie ups, and many people could not use the engine immediately. Since, then, the hype has died down, but the engine is chugging on.

Wofram Alpha Blog shows servers set up

Wolfram|Alpha could be a tool for Science Fiction writers. How better to design a plausible new world for your next story or novel than to plug the pieces into Wolfram|Alpha.

If your world has .8 times the gravity of Earth, Wolfram|Alpha will be able to tell you how high a 150 pound man should be able to jump, and how heavy your ninety pound spacesuit will feel while you wear it while you are jumping.

Want to know the tidal patterns for a world 1.5 times the size of Earth, with one moon larger than Luna, and the same distance away, and the second moon smaller, and closer? Wolfram|Alpha should have knowledge of the equations to figure out those answers, and it probably does- but it cannot connect the dots yet. I expected it to calculate that information from the beginning.

Mystery writers may also eventually find it an asset. And it may settle more bets in bars than OTB. It continues to benefit students, often of younger ages than might have been expected.

Picture from Wolfram|Alpha Blog of their YouTube video of the servers for the site getting set up.