/Interview/ Q&A: Jon McLoone

WolframAlpha created a mixture of excitement and confusion when it launched in May this year. Senior developer Jon McLoone explains how the ‘knowledge engine’ delivers its definitive results

.net: How did Wolfram Research come up with the idea for a “computational knowledge engine”?

JM: In a sense, it is a natural extension of the technologies that we have been developing over the last 20 years. Our Mathematica technical computing system has been focused on automating the mechanics of computation and modelling, allowing scientists and engineers to focus on asking the questions and interpreting the results, rather than doing the computations. WolframAlpha does that on a much broader and more intuitive level letting people ask questions but not have to do the work of sifting search results and collating or transforming them to make sense of them. Mathematica is optimised for carefully specified precise questions and processes and for adding your own knowledge, WolframAlpha is optimised for more ad-hoc and fuzzy queries.

.net: What distinguishes it from a search engine like Google?

JM: There are several key differences. Firstly, we are not trying to do the central purpose of search engines – to point you to other places on the web, which seem related to your query. Our aim is to provide you answers directly from a trusted set of curated data. Secondly, we are focused on facts not opinion. A large fraction of the web is focused on discussion and opinions. A search engine rightly helps you find relevant discussions, but we aim only to return factual data. Lastly and most importantly, a key feature of WolframAlpha is that we build computation on top of the data- knowledge and expertise that uses the data. This allows us to give answers to questions that have never been asked before because rather than looking for a page where someone has already written the answer, it can use algorithms to generate an answer. From simple comparisons like “height of Snowdon/height of Eiffel tower” (3.62 times higher) to complex differential equation solutions like “distance to moon Christmas day 2012” (252130 miles).
.net: Could you give us a few practical examples of what WolframAlpha can be (and is being) used for?

JM: Looking at some of the more common queries, there are a lot related to school homework, for example WolframAlpha’s ability to solve maths equations, and calculus problems showing the working steps, and plotting graphs. Nutrition and exercise queries are popular, looking up how much fat is in a particular meal and whether 20 minutes of brisk walking burns it off again. Finance queries are also popular, comparing different stocks and shares, mortgage calculations etc.
.net: How has WolframAlpha developed since its launch back in May?

JM: We have a very efficient pipeline for coding, testing and deploying improvements that has allowed us to update the live system approximately once a week. So far that has included over two million lines of additional code and over 50,000 data updates since launch. These behind-the-scenes updates have extended WolframAlpha’s knowledge base by about 10%, and improved some of the linguistic analysis to reduce the rate at which queries are not understood by around 10%. Aside from the technology, a whole community has developed around WolframAlpha with hundreds of thousands of messages much of it in the public forums and some privately. This has helped us prioritise directions and target the right functionality, and we are not starting to help organise some of that community for collecting future data.
.net: There’s been a lot of initial excitement but things seem to have quietened down a bit and traffic has dropped. What’s the feedback been like over the last months, and how difficult do people find it to get their heads around the engine?

JM: The buzz that it generated at launch was amazing and showed that there is a real demand for a new way of getting answers. The response has been extremely positive. Of course there have been those who were expecting a traditional search engine and didn’t “get” what we were about, but on the whole people want what we are offering but they want more of it. Of course this was always going to be a long-term project, the question of when to launch was not about when it would be finished, but when we had done enough to be useful. As time goes by we will fill in the main data sets that we are missing, and with field we cover we will pick up additional users.
.net: What are some of the main technologies under WolframAlpha’s bonnet?

JM: WolframAlpha is almost entirely powered by the latest versions of technologies that have been the basis of our business for the last 20 years. The data handling, computation, visualisation, and document construction are handled by Mathematica, which is the same tool used by engineers and scientists for their R&D and students for their calculus homework. The parallelisation over our computing cloud is handled by grid Mathematica and the website is delivered using web Mathematica. We did have a head start on the web technology as we have been using web Mathematica 3, which was only released to the public in mid-September.
.net: Where does the data come from and how do you verify/check it?

JM: The data comes from a very wide range of sources. Some of it is the output of government departments or academic research, some is licensed from data suppliers, some of it comes from open APIs from different online projects and some of it is collected sometimes one datum at a time by our curators. One thing that they all have in common is that they must all pass through one of our expert curators to get into WolframAlpha. The curator acts partly as a gatekeeper, who must check the reliability of the source, and corroborate values. But they also work to standardise data, for example using standard date conventions, merging multiple sources, and then adding the computations that are specific to that type of data. This is a very intensive task, but delivers very reliable data.
.net: What are the main challenges in WolframAlpha’s development?

JM: The major technical effort is scaling up the data curation, so that we can get more data in to the system. Because this is an expert task, needing understanding of the subject matter that the data relates to, it requires some quite challenging recruitment.

Also challenging, but quite different, is changing people’s perceptions of what information they can and can’t get. We all have lots of experience with page-indexing search engines. We know the way to ask questions and we know the kinds of questions that will fail to find anything useful. Those rules are different with WolframAlpha, we answer lots of questions that search engines cannot, but don’t try to answer others that they do. In a sense it is a great success when we see people posting comments that “WolframAlpha couldn’t answer the question….” followed by a complex query that one could never solve with a search engine. We have raised that users aspirations, we just need to keep up with them now.
.net: When will you release the WolframAlpha API and what are your hopes for it?

JM: The WolframAlpha website is only one view into the larger project of “making the world’s data computable”. The API is another way to access the project which should, in time, be more important than the website. By providing the API, we are making it possible for all kinds of other projects to add computational knowledge to their services. Building the infrastructure for such features is very expensive, but by being able to access WolframAlpha behind the scenes, services will be able to do more interesting things very cheaply. For example, supermarket recipe sites, even wiki based recipe sites, could generate nutritional contents labels, travel booking sites could tell you what the average weather has been for the exact dates of your booking, in previous years, for your exact destination. In both cases, the API call would be a few lines of code for the website developer.
.net: What’s in the pipeline for WolframAlpha?

JM: As far as the user will notice, it will be more data, and being able to do more interesting things with the data. Probably largely unnoticed by the user will be partnerships that use the WolframAlpha API. But there will be a few surprises along the way!

Jon McLoone
Job Senior Developer
Age 39
Education Degree in mathematics from Durham University
Previous career With Wolfram Research since 1992, working on software development, system design, technical writing and strategy
Online www.wolfram.com



1 Response to “/Interview/ Q&A: Jon McLoone”

  1. January 11, 2013 at 10:54 am

    I’m really enjoying the design and layout of your website. It’s a very easy on the eyes which makes it much more pleasant for me to come here and visit more often.
    Did you hire out a designer to create your theme?
    Superb work!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s


%d bloggers like this: