Two months ago Wolfram Alpha launched its question answering engine in one of the most eagerly anticipated product launches of the year. Although sensationalists wanted to call Wolfram Alpha a Google killer, here at the Hub we have consistently viewed Wolfram Alpha as a valuable complementary tool to Google rather than a direct competitor. Just a few weeks ago at sci foo 2009 I was able to sit in on a presentation from Wolfram Alpha Co-founder Theodore Gray in which he gave an update on the status of the service. My takeaway was the same as before: Wolfram Alpha is the real deal. This is a serious company backed by some really smart people and a formidable technology.
As I noted earlier, the presentation from Theodore Gray was quite an experience given that more than half of the 12 or so people in the room joining me to watch the presentation were serious big hitters at Google (VP’s and higher) that I know personally. So we’ve got the co-founder of Wolfram Alpha basically revealing some pretty juicy details of his service to a very interesting crowd – quite a setting.
One of the most interesting insights I gathered from Theodore is that Wolfram Alpha is not simply sitting on its laurels letting its service stagnate and languish. Quite the opposite is true in fact – these guys are modifying and innovating their service at light speed. With hundreds of thousands of users now pounding on the service everyday, Wolfram Alpha is able to learn about bugs and kinks in the system at a blistering pace. In fact this is one of the primary reasons that the service was launched when it was, even though it wasn’t as polished as Wolfram Alpha might have hoped. Wolfram Alpha wanted to leverage the power of millions of users to largely automate the detection of the service’s myriad flaws. Each day the Wolfram Alpha engine is improving as tiny problems are identified either by users directly or by analyzing the logs, and then fixed. The website’s interface itself, just like Google, remains ever the same with only a deceivingly simple search box, but behind the scenes the service is constantly evolving.
Not only are bugs being worked out, but the amount and quality of data that the Wolfram Alpha engine is able to draw upon is rapidly expanding. As it stands now, the only data incorporated into the Wolfram Alpha engine is data that has been acquired though trusted, one on one relationships between Wolfram Alpha and the creator/owner of the data. This includes all sorts of data, from Morningstar financial data to chemistry data from the CRC handbook, and this trove of data is growing rapidly. The Wolfram Alpha service is becoming more comprehensive and powerful as the amount of data it is able to draw from expands.
Although incorporating data from trusted sources through direct relationships is a valuable first step in Wolfram Alpha’s data acquisition strategy, the people at Wolfram Alpha fully understand that this paradigm has severe limitations. The true power of the Wolfram Alpha question answering engine will only be unleashed when it is able to draw upon crowdsourced data that can be submitted by anyone all over the world. A large corpus of more current data, or recently changing information, such as the death of a person, achievement of a new world record, and various other day to day information updates will then be available to the engine. For Wolfram Alpha, the challenge is to incorporate this data in a way that properly labels it so that users can decide for themselves how much they can trust the data supporting the answers to their questions.
Even though the Wolfram Alpha engine has a long way to go towards question answering nirvana, the service is already impressively capable and it deserves respect from the public at large as a new and valuable tool in our information gathering arsenals. Wolfram Alpha is not Google. It is not a search engine that helps you find information on the web. Wolfram Alpha is a computational question answering engine – it can compute factual answers from simple human readable questions. It is a true paradigm shift, a revolutionary service that can convert a regular human being into a super human with stunning computational capabilities at the tip of their fingers.
Theodore noted that there are several calculations that are common to professions that although straightforward, still require large amounts of time and effort for even the skilled professional. Take the preparation of a molar solution in chemistry, a daily task for practicing chemists and students in high school and college. These calculations require one to look up various pieces of information in tables and charts and then several mathematical conversions to achieve an answer. Now with the Wolfram Alpha engine, both trained professionals as well as the common man can make this calculation instantly by inputting only the most basic characteristics of the problem. Simply type “0.4 molar potassium chloride” into Wolfram Alpha and bam! – out comes the list of properties of the solution and the quantities of solute and solvent required to make it.
Wolfram Alpha opens the possibility of completely changing the rules of education. Wolfram Alpha can answer the questions to an hour’s worth of hard chemistry homework in a matter of minutes. Just as the calculator has revolutionized the capabilities and the learning expectations of students in mathematics, might Wolfram Alpha do the same for chemistry, physics, engineering, and other fields?
Impressive as the Wolfram Alpha service may be, one has to ask the question “is anybody using it?”. The answer thus far is a pretty resounding no. If I look at my own usage of the service, I have to admit that yesterday was the first time I have used it in many weeks, and that was to perform research for this story. Now that the initial euphoria and interest has worn off since its launch 2 months ago it seems like nobody is talking about Wolfram Alpha. A look at Wolfram Alpha’s traffic stats from Compete is telling:
From the graph above we can see that during the month of May Wolfram Alpha only saw a paltry 1.5 Million visits to its site in May, followed by a precipitous drop to 750,000 visits during June. The data from compete is extremely unreliable and can easily be off by a factor of two or more in my experience, but still I think the trend is clear. Traffic has declined since the initial launch, and even if the Compete data is off by a factor or so, we have to admit that Wolfram Alpha’s traffic is likely on the order of just a few million visitors per month or less rather than the tens of millions that might be expected.
So what does it mean for Wolfram Alpha when even a self proclaimed fan of their service like me is not using it? I think the point is twofold. First, it may turn out that many of the factual answers that Wolfram Alpha is good at providing are not answers that we need as often as we thought. If I was in school right now or in an engineering job I might find myself using Wolfram Alpha a lot more, but for everyday use perhaps not so much. Second, human behavior is hard to change. It will take time and perhaps more than a website to get me to use the Wolfram Alpha service as more of a daily part of my routine.
But don’t let Wolfram Alpha’s limited penetration into our lives fool you into discounting the promise and the usefulness of their service. Whether it be Wolfram Alpha or one of their competitors that brings it to us, the convergence of several technological trends means that the time has come for a reliable, automated question answering service to become a reality. Wolfram Alpha offers a fantastic technology and it is going to get much better. It will take time for the company to find more effective ways of injecting its service into our lives through open api’s and integration with other tools (even with Google perhaps), but still the service is a great one and I think it has a bright future.