Things with Wolfram|Alpha are going well. Really well. So well that I’m now incredibly keen to scale them up dramatically.

When I started the Wolfram|Alpha project, I was not even sure anything like it would be possible. But over the last two years we’ve proved that, yes, with the tower of technology we’ve created, one can in fact take large swaths of knowledge, make them computable, and deliver them for everyone to use.

From the outside, it’s easy to see that there’s been steady growth in the domains of knowledge that Wolfram|Alpha covers. And over the next few months there’ll be some big additions, notably in everyday and consumer areas. But to me what’s most dramatic is what’s happened on the inside. Because what we’ve done is to build a giant system of technology and management processes that allows us systematically to make any area of knowledge computable.

The catch is that it always takes effort. We rely on a huge tower of automation. But in every new area we tackle there are new issues, new opportunities—and new ways that resources and human effort have to be used.

I’m very pleased with how broad and deep the coverage we have already achieved is. But we have an immense to-do list, assembled not least from all the feedback we’ve received from users of Wolfram|Alpha. And the good news is that at this point it’s a straight shot: given enough effort, we can complete the to-do list. We have all the systems we need to scale the knowledge in Wolfram|Alpha up all the way.

There’s another thing that’s happening now too: we’re steadily understanding more and more about the platform we’ve created with Wolfram|Alpha. And we’re realizing that there are an incredible number of different ways this platform can be used and deployed.

At first Wolfram|Alpha was just a website. Already it has spawned two dozen other products, and the pace of new products will accelerate dramatically this fall.

We are also just beginning to understand Wolfram|Alpha as a new computing paradigm—and that too is spawning all sorts of new directions. What if one takes input that isn’t text? What if one could deliver results preemptively? What if Wolfram|Alpha ideas are fundamentally integrated into a programming language? And so on.

These directions are more speculative—and someone less confident than I might barely consider them. But I have no doubt that all of them will bear fruit, often in much richer ways than we can yet imagine.

There is so much to do. So much further to scale up the Wolfram|Alpha project. And the exciting thing is that we now have the technology and the organizational systems necessary to do it.

And we’ve also developed a loyal following of people who use Wolfram|Alpha every day—on the web, on mobile devices, through other systems, and elsewhere. At the beginning, we had lots of tourists, bringing us web traffic—and checking out Easter eggs and the like. But over time that has given way to a consistent base of users who rely on us for all sorts of things.

Realistically, though, we haven’t been “discovered” yet in even a small fraction of the areas we cover. We don’t know the whole process by which that happens (and it would make a fascinating study), but somehow, gradually, different areas of Wolfram|Alpha functionality do seem to be “discovered”, and progressively build up larger and larger followings.

At a business level, one strategy we could follow is to wait for more areas to be discovered, build up partnerships, app sales, and so on, and use revenue from this to fund the expansion of the project. And certainly I myself have no fear of very long-term projects: in a month or so, for example, I will have worked on Mathematica for 25 years.

And indeed in my experience, there are some projects that inevitably require the passage of many years. But with Wolfram|Alpha today the story is different: we know that we can scale up, and that if we do, we can in a modest number of years cover an immense territory. There will always be more to do—more computational knowledge to add. But with what we know how to do today, we are in position quite rapidly to vastly expand what can be made computable, and, I hope, by doing so, to make some great advances in what is possible in the world.

I have been fortunate enough to run a group of companies that have been consistently profitable for the past 23 years—and indeed this is what made it possible for me to create Wolfram|Alpha in the first place. And today, with the current state of Wolfram|Alpha as a commercial entity, we can certainly continue to grow its capabilities.

But at this point I feel we almost have a responsibility to push the development of Wolfram|Alpha at the maximum possible rate, and to deliver to everyone what is technically possible absolutely as fast as it can be done.

To be able to do this, however, we need not only great technical and systems innovations, but also business innovations. We need to learn how to take Wolfram|Alpha, and not only scale up its content and capabilities, but also scale up its revenue and its business reach.

So this fall, along with all sorts of new features in Wolfram|Alpha, we will be trying out a whole range of business innovations for Wolfram|Alpha. Inevitably there are tensions and compromises to be made, and in the course of the next few months we will be working hard to navigate these. We need more revenue to scale up what we are doing—but whatever happens, we are committed to ensuring that the basic Wolfram|Alpha website remains free for anyone anywhere to use.

It is, I think, in everyone’s interest to see Wolfram|Alpha developed as quickly as it can be. And we are working hard to figure out the best possible way to do this—with the best models for how value for users can translate into commercial support for the development of the project.

I must admit that I personally have spent much more of my life working on strategy for technology than strategy for business. But there are such wonderful opportunities today for Wolfram|Alpha that I feel compelled to put whatever powers of innovation I have toward business as well as technology.

Whatever happens, there is a wonderful path ahead for Wolfram|Alpha. But I look forward to finding ways to deliver to the world a large chunk of our current to-do list not in a quarter of a century, but in a small number of years. And I look forward to working with our users and others to find the best ways to make that possible.

12 Comments

The last few weeks I’ve been using WA more and more. Every once in a while I can’t help but think that Liebniz must be having a party in his grave! Keep up the good work. Looking forward to a personal version of WA that would allow me to use my own data. “Piping” the results of a query as input for another would be fantastic as well.

Don’t hesitate to contact me if you need some beta testers. ;)

Posted by Jose August 30, 2011 at 10:08 pm Reply

You mention that WA has ‘Not yet been discovered’.
The way to be discovered ,is to succeed. People are very resistant to new ideas but quick to copy success. So concentrate on succeeding in business and make it clear that this is owed to WA.
The key improvement needed is to get WA actively integrating all it’s knowledge. It seems to me that the curation of new data is limited to one area at a time whereas it should be curated so that it is integrated in WAs existing knowledge.

Posted by Brian Gilbert September 1, 2011 at 4:41 pm Reply

!please! do not go preemptive – I believe this is counter-productive to well functioning society (someone tell google)

I love what you’re doing, my keyworded bookmark is proving proving more and more fun and effective than google lately

An idea I just had while using your machine is that (when I ask a math question, like 4289 times 92) wolfram|alpha could include (an optional) animation showing the (right!) way of equating the answer

Posted by Drew September 3, 2011 at 9:42 am Reply

    It does show steps. In the output you will see a red link ‘Show steps’ Click on that and it shows the steps.

    Posted by Brian Gilbert September 7, 2011 at 10:50 am Reply

In the ‘About’ page of WA http://www.wolframalpha.com/about.html you set out the Goals for WA. To start with there can only be one goal, and it needs expressing more clearly as your whole organisation depends on it for guidance.
So what are the weak points?
- The definition buried amongst that text is ‘make all systematic knowledge immediately computable and accessible to everyone. ‘ This is inadequate’.
- It assumes that all knowledge is up to scientific standards and even scientific standards do not ensure that data is 100% reliable. A relaibility attribute is needed for every piece of data so that it can be included in computtions. Some data such as that on global warming includes contradictions which could be coped with this way.
- The claim to accept free-form input gives the user false expectations as WA often misunderstands. It needs to be emphasised that the user can only rely on the answer as expressed in the Input Interpretation in Mathematica form.
- There is little sign that during the curation of data it is is checked for contradictions with existing WA data. This is vital to ensure that WA does not give contradictory answers. As the body of data grows WA will reveal during the curation of data, massive errors in received wisdom .

An earler definition of the goal by you was ‘Computable knowledge engine’.
I suggest ‘Create a computer system that will accumulate a complete and consistent body of computable data such that it can answer any question. ‘
The data includes methods of computing data. The data must include a reliability factor so that the answer can compute the reliability of nthe answer.

Posted by Brian Gilbert October 5, 2011 at 4:41 am Reply

One more step gets you to the Holy Grail, a human language that is unambiguous.

The Mathematica language is unambiguous but its computer friendly attributes such as the use of parameters is not human friendly. Copyable plaintext is human friendly but restricted to one statement at a time. I suggest you institute one of your automation projects to add the function to convert a whole Mathematica program to copyable plaintext and make it possible to input a Mathematica program/WA Query in that unambiguous plaintext.

Once familar with plaintext we can use plaintext to communicate with other people unambiguously. It could replace English in its present form for serious/scientific communications. It could become the Lingua Franca.

Posted by Brian Gilbert October 6, 2011 at 1:35 am Reply

    Ambiguities in language are what makes it so rich! If there were no shades of meaning, even in serious communications, the language would be so drab!

    Posted by Patrick October 10, 2011 at 2:39 pm Reply

Thanks for your role in Siri! Great to see W|A partnered with Apple!!

Posted by Alec Ellsworth October 14, 2011 at 11:27 pm Reply

I think it would be great if one of these days ANDROID users were not left out for choosing ANDROID. I love Wolfram, but I wish it every app for Iphone and ipad were available for android and as soon as each new app is released android app for that would be released simultaneously. I additionally think that an app for engineers should be developed. It should include a drop down menu allowing you to choose your branch of engineering and each should include precise and helpful tools that would help engineering students like myself in these engineering classes

Posted by Josh December 1, 2011 at 4:51 am Reply

PLEASE, make a PALETTE of symbols available for input, just as in Mathematica

Posted by Roberto Roy December 5, 2011 at 12:36 am Reply

A potential innovation in the business model is to have an Intelligent Agent with strong language understanding abilities access Wolfram Alpha in the process of an ORGANIC conversation with a person. Siri is an IA accessing WA, but it is not really able to carry a meaningful conversation with a person: as Siri focused on handling requests in a dozen or so areas (weather, calendar, itineraries, etc.), it calls the respective web services, and delegates to WA things like unit conversions, or questions that are very clearly factual.
These web services don’t even scratch the surface of what is available in WA.

We are building an Intelligent Agent that can carry multi-step conversations and has stronger disambiguation, co-reference resolution and world-knowledge foundations, and is used for an entertainment application (as part of the forthcoming Bot Colony online game). One of the game’s objectives is to help improve the verbal ability of people. A UNESCO study has shown that improved verbal ability is directly correlated with an improvement in the standard of living.

The point is that such an application will draw on much wider knowledge, as well as the computational abilities of WA, to reason about day to day situations and carry on conversations with people. These conversations must be natural, multi-step and in a context where people are entertained, and in the process they also learn something.

This may be a good business solution for WA.

Posted by Eugene Joseph January 20, 2012 at 8:02 am Reply
Leave a Comment

(required)

(will not be published) (required)

(your comment will be held for moderation)