Wolfram|Alpha: The Use Cases
At the recent SemTech conference in San Jose, I sat down with Wolfram|Alpha's Russell Foltz-Smith. Wolfram|Alpha bills itself as a "computational knowledge engine," a nerdy and unfortunately not very intuitive description. Because it's hard to grok, most people have categorized Wolfram|Alpha as a new type of search engine. The site got a lot of press when it launched in May, as many pundits saw it as a challenger to Google. However in our own extensive tests of the product before launch, we concluded that it isn't a "Google Killer" and that it has more in common with Wikipedia.
Even now there is still confusion about what Wolfram|Alpha is and what its main use cases will be. In this interview with Russell Foltz-Smith, we discuss what people are using Wolfram|Alpha for now; and more importantly what its uses will be in the near future.
Wolfram|Alpha: What is it Good For?
Wolfram|Alpha is a product that was built on top of founder Stephen Wolfram's Mathematica product, a software tool for mathematicians that was initially released in 1988. The aim is to allow users to type human-like statements and have computations done on those. Wolfram|Alpha was first conceived and started development about 4 years ago, and just 6-8 months ago the team gave serious consideration to taking the product to a wider consumer audience.
I started out by asking Foltz-Smith what the Wolfram|Alpha team thought of all the media hype around their product, particularly about the "Google Killer" theme which many media outlets reveled in. Foltz-Smith replied that they were expecting to be compared to Google, but not to that extent. Their team was a little surprised there wasn't more discussion around Wolfram|Alpha's similarities to Wikipedia and Freebase (although he noted that ReadWriteWeb certainly covered that!). Regarding the Google comparisons, Foltz-Smith said that they didn't give into the hype - they stuck to what their goals were.
I remarked that many people still seem confused about what Wolfram|Alpha does and what it can be used for. Foltz-Smith said that people will use it for different things. The crux of the product though is that it allows people to compute and calculate things.
But will mainstream people use Wolfram|Alpha? Right now, it seems to be focused on mathematicians. Foltz-Smith replied that yes, eventually Wolfram|Alpha will find a mainstream audience. It has started specific, but it will go broader. First, he said, it has to "pass a test" with "serious users" - by which he means academics and computational users. If it's useful for them, claimed Foltz-Smith, it will then go mainstream.
Use Case: Education
One real-world use case we talked about was using Wolfram|Alpha in education. Russell Foltz-Smith said that Wolfram|Alpha could be used to automatically generate problem sets for students, and then research those sets.
A recent article in education website Chronicle.com argued that Wolfram|Alpha may have a less desired effect: encouraging cheating and laziness in students. This is because Wolfram|Alpha not only solves complex math problems, it "also can spell out the steps leading to those solutions."
Stephen Wolfram told Chronicle.com that computer-algebra systems like Wolfram|Alpha actually improve education - because they allow students to explore complex problems on their own and intuitively determine how functions work, rather than just learn rote processes. Wolfram claimed that "it's better to let them [students] stand on that platform and go further."
Either way, it's clear that Wolfram|Alpha and similar computational software will force the education system to adapt and change. Students now have a new (and certainly easier to use, as it's on the Web) platform on which to compute things. There's no point in the education system pretending it doesn't exist. If you're interested in tracking the progress of Wolfram|Alpha in educational settings, there is a wiki devoted to 'Teaching Undergraduate Math with Wolfram|Alpha.'
Use Case: Computational Journalism
This one was described to me as "anomaly spotting." For example with the current interest in swine flu news, Wolfram|Alpha could be used to fact-find and compute interesting trends. As Foltz-Smith described it, Wolfram|Alpha could "automatically enhance news."
Foltz-Smith noted that CNN and other major networks do this already (analyze data), but that it's expensive to do. The end results on CNN are added value things like interactive maps and fancy diagrams. Wolfram|Alpha could make this type of data gathering and analysis presentation inexpensive and common place amongst all kinds of news operations - including good old blogs.
Use Case: Sports Watching
Imagine sitting in your sofa in the lounge, remote control in one hand and your favorite beverage in the other. You're watching the Friday night game on TV, it's a close game and you're curious about which team has the better chance of winning. Why, check Wolfram|Alpha of course! In real time, Wolfram|Alpha could compute statistics about not just the history of the two teams - but the history of the location of the game, the weather, the season so far, etc.
As Foltz-Smith explained it, Wolfram|Alpha would be able to do "chained queries" - queries made up of multiple parts. For example: which quarterback had the best winning record in games played in the rain during the 1970s.
Other Use Cases
We also discussed medical and scientific use cases. Although there are early examples of Wolfram|Alpha in health, such as a nutrition label generator, Foltz-Smith was generally cautious about medical uses - because a lot of health data "can't be wrong." He noted that in use cases like medical research, the issue of data fidelity is key. For example with the human genome, you have to take great care of that data and associated algorithms. Also he explained that as something like the human genome scales, how do you do QA?
Foltz-Smith admitted that the Wolfram|Alpha team is still working on these and similar issues. But they have a lot of people devoted to solving this problem. Some types of data could be crowdsourced, e.g. in linguistics, but other data needs different approaches.
Conclusion
It was interesting to hear about some of the potential uses of Wolfram|Alpha. We at ReadWriteWeb think this product has a promising future. If Web 2.0 was about creating data (user generated content, to use the most familiar term for this), then the next generation of the Web is all about using that data. Wolfram|Alpha is premised on using and computing data.
Let us know in the comments what use cases you see for Wolfram|Alpha, and whether you're aware of similar computational web apps.
See also:
Share
Facebook
Comments
Subscribe to comments for this post OR Subscribe to comments for all ReadWriteWeb posts
I don't know whether it would be funny or slightly tragic if they spent millions of dollars to develop this tool and it ends up only being used to predict the outcome of Fantasy Football match-ups.
Wolfram Alpha can be a very useful tool in many applications. At first, I was confused what to do, but as I spent more time with Wolfram Alpha, it became clear. I recently wrote an article in the same vein detailing how I use Wolfram Alpha for web design.
There are several very interesting ideas in play here.
1. Stephen Wolfram seems to have had a change of heart (getting one, actually) since he has always charged a fortune for the Mathematica program, aimed at penniless students (what's half of $2000?) as well as rich corporate researchers.
Wolfram providing something useful "for free" must mark the end of time.
2. I have always been impressed by how computers can remove the tedium of "mechanical" operations, especially in symbolic algebra, so that people can focus on the concepts, which IMNSHO are more important than the tedium.
3. In those cases where "the tedium is the message," (learning the steps) then I see some value in presenting people with several alternative derivations or expansions, so that they can actually learn, rather than memorize steps rote-style (which only suits exam-takers, "certificate chasers" and making loads of money in the so-called real world)
4. For what it's worth, it is often worthwhile to see other people's solutions to learn problem-solving, much as it's often valuable to read other people's source code in order to learn programming. What's important is to actually *learn* from this, rather than just memorize.
5. Of course, if the computer or the internet simply "airdrops" the answer, you'll learn very little unless you have the curiosity and integrity to use this to tell you that your own steps end up with the right result.
6. Of course, some schools are still struggling over the impact of calculators. Not to mention that schools will expel you for collaborating with your peers, which is one of the most valuable skill you can have at work.
GO FIGURE.
From the Chronicle article:
As Mr. Freedman, the physics instructor, put it: "The greatest challenges that science and math students face are conceptual, not computational, and neither calculators nor WolframAlpha can do much about that."
I'm skeptical.
Try inputing
"passing touchdowns 49ers, raiders"
or
"passing touchdowns forty niners, raiders"
or
"passing touchdowns forty niners, oakland raiders"
or
"passing touchdowns fortyniners, raiders"
As of this comment, THEY DON'T WORK. I smell the sweet scent of perpetual motion.
Walter - try not to be so parochial. That query doesn't (it's the SF part that it chokes on oddly) but sub in Seattle Seahawks and it does.
And that's my issue with WA - if you phrase things JUST RIGHT it's fine, but it can be a lot of work figuring out how to phrase the question. For example, a discussion with a friend about weather here led to me wondering how many days above 70F Seattle got and I thought WA would be a perfect engine for that. I have yet to figure out how to ask it that though. Oh, I can get averages etc... and see them for multiple years. But if I want to see how many days were over 70F from 1970-2000? Nothing I've tried works. And the issue is that when if fails it just fails - it's an all or nothing approach.
I'd have LOVED this in math classes... I can see it being very useful in some cases. BUt it's wrong to think of it as a general purpose tool that people can occasionally use - you need some practice in how to phrase queries.
So out of idle curiosity, has anyone managed to find a single solitary thing that Wolfram Alpha is actually useful for?
Because I never get any response other than "Wolfram|Alpha isn't sure what to do with your input" - to the extent that if I actually type in the example above "passing touchdowns, dallas cowboys, denver broncos"...
..."Wolfram|Alpha isn't sure what to do with your input"
I'm unclear what is parochial about being unable to repeat an example with different data. Isn't being able to extract knowledge from data what wolfram|alpha is all about?
I'm also unclear what you mean by SF. San Francisco? SF 49ers? Tried those too.
This is not a knowledge engine. It's an interactive almanac with a built in scientific calculator and a poorly thought out trend-chasing user interface. That is being foisted on the world as a knowledge engine, by people who ought to know better. Wolfram wears no clothes.
"Wolfram Alpha can be a very useful tool in many applications. At first, I was confused what to do, but as I spent more time with Wolfram Alpha, it became clear. I recently wrote an article in the same vein detailing how I use Wolfram Alpha for web design."
Checked out your page, Ethan. That is the most excellent use of Wolfram|Alpha I've seen to date.
It's almost as good as my Web design and CSS desk reference book, though not quite as useful as doing a simple web search and parsing the results to get what I need.
Ground breaking stuff, this Wolfram business. Colour me abjectly disappointed.
Something fun...ask Wolfram|Alpha "How old are you?" and "What are you?"...it will tell you in it's own words.
For me, Wolfram|Alpha, as useful as it already is, points to the future. It gives us a window into what kind of things are possible, and will be common place in the not too distant future.
For literally decades, computer scientists have dreamed of a system that was able to accept roughly naturally language queries and return an exact answer, even in cases where computation and reasoning were involved.
As far as I know, W|A is the first really useful system that demonstrates this kind of thing.
So the excitement isn't just about what W|A is, but what it may become.
WA contains a lot of interesting data and algorithms. If you can find the right incantation, you can learn some things.
How about some examples of those obscure sports queries like "passing touchdowns in the rain" or whatever they suggest. I have yet to be able to configure the more interesting chained queries that have been suggested in articles on WA. Mostly it is frustrating like people above have mentioned.
Thanks a subject that truly move