The launch of the “computational knowledge engine” Wolfram Alpha was one of the most anticipated product launches of early 2009. Since then, it’s been rather quiet around Wolfram Alpha, even though the company continues to add new features and data on a regular basis. Today, we had a chance to talk to Wolfram Research’s founder Stephen Wolfram about the first year and the company’s plans for the future.
Looking Back
As Wolfram told us, the most basic question he tried to answer when the company started development was simply to see if it was even possible to take all this data and make it computable. Now, a year later, his answer to that question is an emphatic “yes.”
Wolfram, however, also acknowledged that right after the launch the user experience for first-time users wasn’t necessarily ideal, as Wolfram Alpha didn’t yet have data for a lot of knowledge domains. The choice at that point, he said, was to either delay the launch and get more data, or to release Wolfram Alpha and be able to learn how its users would use it, and then enhance the experience over time.
Wolfram says that today, most users are aware of the difference between a search engine and Wolfram Alpha, and the experience for first-time users has become far better. He stressed that the team (which consists of about 200 employees and 500 volunteers) is currently adding new data at an increasing pace. That’s gotten easier as the team has learned how to import information from a large variety of knowledge domains and sub-specialties.
Getting the Data is Just 5% of the Work
Unlike Google, Wolfram thinks that the Web “isn’t useful for getting raw data.” Indeed, whenever the Wolfram Alpha team experimented with this, the data simply wasn’t up to par. Instead, the company will continue to mostly work with data from primary sources. Getting this data, however, is only 5% of the work. The real difficulty is to understand how to compute this data and to understand how people talk about this data: What kind of questions do they ask? What are the alternate names for a specific chemical element? In addition, the Wolfram Alpha team and volunteers also check for anomalies in the data they receive. If there are major outliers, the team will track down more information to verify the original source.
Sadly, though, not all data is free and Wolfram Alpha has to pay if it wants to include some databases. To make Wolfram a viable business and still offer this data, the team is considering a subscription plan that will give paying users access to deeper datasets from subscription databases.
Challenge: Bringing Wolfram Alpha to More Users
The question now, however, is how to get more users and how to bring Wolfram Alpha to more users through more channels. As we noted earlier this year, the company’s newly minted managing director Barak Berkowitz thinks that the team’s “number-one priority is to get Wolfram|Alpha in the hands of everyone.” To get to this point, they will soon release more and better tools for third-party developers who want to use the company’s APIs to integrate Wolfram Alpha’s functionality in their own sites and services. It’s also worth noting that Wolfram Alpha now offers an appliance that companies can install behind their firewall to curate and compute their own data.
Looking Ahead: Analyzing Your Own Data, More Knowledge Domains, Programming with Natural Language Queries
Besides looking back, we also asked Wolfram about his plans for the future. In answering this question, he stressed that this new approach to computing is just getting started and it usually takes him about 10 years to develop his projects before he fully understands what’s possible once this new paradigm has arrived.
For the near future, however, Wolfram hopes that Wolfram Alpha’s users will be able to upload their own data and perform complex computations on this data and use Wolfram Alpha to find correlations within Alpha’s vast database. The usage scenarios for this could include anything from analyzing sales data to doing personal analytics on data from devices like the Fitbit. In addition to uploading data, Wolfram Alpha will soon make it easier for users to download data to use in presentations.
Wolfram also wants to bring Wolfram Alpha and Mathematica closer together. One development that Wolfram is especially excited about is using Wolfram Alpha’s ability to understand and compute natural language queries in order to create Mathematica programs. By building on this capability, Mathematica users may soon be able to write and manipulate their code using natural language queries just like in Wolfram Alpha.
Obviously, the team behind Wolfram Alpha will also continue to add more data across an every-growing number of knowledge domains. Today, for example, the team is launching real-time space weather data, 12 complete genomes and local maps, as well as numerous other knowledge domains related to math, biology, physics and geography.
