Steven Wolfram, the man behind computing-application Mathematica and the search engine Wolfram Alpha, has a short attention span that?s married to a long-term outlook.
Wolfram Alpha is an online service that computes the answers to queries (e.g., age pyramid for the Philippines or glycogen degradation pathway rather than searching for those terms showing up on webpages.
When asked what his favorite query is, the particle physicist and MacArthur ?genius? award recipient says he?s enamored that Wolfram Alpha can tell you about the plane you just saw flying over your town ? in his case ?flights visible from Concord, Massachusetts.?
But Wolfram?s no plane-spotter.
?My life consists of watching all the new domains being put into Wolfram Alpha,? Wolfram said. ?Whatever thing we just finished is the thing I?m most excited about.?
And you might understand Wolfram?s excitement about being able to know the tail number of a plane overhead when you get that answering that question isn?t easy.
For one, there are a lot of planes in the sky. And two, even if you know which planes are in the sky, radar data is delayed, so Wolfram Alpha has to project a plane?s course. And it?s got to take into account that people can?t actually see planes that are very high in the sky.
While that might sound like Wolfram has a short attention span, he?s also taking the long view, as Wolfram Alpha has just passed its second birthday.
?This is my third big life project,? Wolfram said. ?Two is early in the life spectrum.?
Wolfram Alpha?s team is now 200 strong, a mix of programmers, linguistic curators and subject-matter experts.
And their to-do list? It?s decades long.
?If you were to look at our whole to-do list, which is a scary thing to do, to finish it would take 20 years,? Wolfram said. ?That doesn?t scare me too much, since I?ve been working on Mathematica for 25 years.?
Wolfram Alpha may have a search box, but it?s doubtful that it?s the default search box for anyone, except perhaps Rainman.
But traffic to Wolfram Alpha is in the millions of visits per day, according to Wolfram, and the company is ?slightly profitable.? That?s in no small part because high school and college students have figured out at least part of what Wolfram Alpha is useful for ? whether they are working on trigonometry equations, music theory or economic models.
?That?s not the worst place to have a core base of users, given they grow up,? Wolfram says.
Wolfram says he takes encouragement from looking at the streams of queries that people put into the search box. Those show that people are trying to use Wolfram Alpha for complicated things like comparing the economies of two countries. And there aren?t many tourists who just show up to see a funny Easter egg in the software, or to enter junk queries.
But Wolfram is frustrated a bit that users don?t know the full power of Wolfram Alpha.
?The mental model for when to go to Wolfram Alpha is not fully fleshed out yet,? Wolfram says.
One of the company?s solution for that is to create a wide range of very focused apps, such as its app for computer network administrators, and those for classes, including astronomy, calculus and algebra.
Wolfram Alpha has also partnered with general purpose search engines such as Bing and DuckDuckGo. The key there, according to Wolfram, is figuring out which of the queries into a general search engine would benefit from a calculated answer, not just a list of links. One of the challenges is that searchers are used to getting search results in single digit milliseconds ? while Wolfram Alpha takes considerably longer ? say 500 milliseconds ? because it?s calculating answers.
One way to solve that is to cache some popular precomputed answers, and ? for others ? to indicate to searchers that they can get more details on Wolfram Alpha.
?We compute it and do the computation in the background, so by the time they show up, it looks like it was there but it wasn?t,? Wolfram said.
The long-term challenge for Wolfram Alpha is getting more and more datasets into the system. While the process has gotten smoother, each dataset comes with its own unique complexities ? meaning that there?s no cookie-cutter approach that will speed new datasets into the engine.
?Our main conclusion is that there is an irreducible amount of work that requires humans and algorithms,? Wolfram said.
The company is also branching out into datasets that one wouldn?t expect from a high-powered calculator, such as info on sports and pop culture, areas that Wolfram Alpha clearly shied away from at first.
?I thought, ?Gosh, what can you compute about people??? Wolfram said. ?Well, it turns out there?s a lot you can compute, such as what people were born in this city and who was alive at the same time as other people. In every area there is a lot more to compute than you think.?
He?s now thinking about how you can ingest people?s networks of friends (the so-called social graph), how images can be imported and calculated, and what happens when Wolfram Alpha allows people to upload their own data sets.
What?s also becoming apparent is that there are a lot more places that Wolfram Alpha is turning out to be useful than just the website. Makers of software such as spreadsheets and specialized financial applications are turning to the company?s API ,so that they can include computational functions in portions of their software. That means more-diverse revenue for the company, which surprised Wolfram, because when the company launched, he suspected there were only two or three ways for it to make money.
Now he says it?s looking like there are 15 channels or even more.
?People just need what we are doing. It seems like it is a foundational component in so many places,? Wolfram said. ?The big debate internally is which of these channels will be the most lucrative, but I think it is still not at all clear.?
And if you think the word channel makes Wolfram sound like an executive, you?d be right.
?I had thought when I started Wolfram Alpha that that stuff isn?t so interesting, and I was going to hire people to figure that out,? Wolfram said. ?That didn?t work out so well.?
?So I decided I should learn it, and it?s actually kind of interesting,? Wolfram said. ?Now is a fascinating time of platform turbulence, which we haven?t seen since probably about 20 years ago in the rise of PC workstations.?
Wolfram Alpha is also self-funded, as was Mathematica.
And in typical Wolfram style, that makes him both more conservative and more radical than others.
?For 23 years, Mathematica has been a simple private company,? Wolfram said. ?For better or worse, that allows one to do much crazier projects than you can through the traditional VC route.?
But doing crazy things doesn?t extend to adding 300 new employees to try to build even faster, even if there?s not enough revenue to pay their salaries.
?I?ve been lucky enough to run a company that?s been profitable for 23 years, so I developed the habit of doing things that way,? Wolfram said.
That?s a way of doing business, that if you think about it, computes much better than getting tens of millions in funding for an iPhone app.
Source: http://www.wired.com/epicenter/2011/05/wolfram-alpha-two/
No comments:
Post a Comment