"Making the world's knowledge computable" -- Sounds like a job for parallel computing.
There’s a new website in town. It is called Wolfram Alpha (W|A). I must admit I am fascinated by the concept. The tagline on the site is Making the world’s knowledge computable. Sounds different. However, before I discuss my experiences with the site, I would like to be clear about one thing. W|A is not Google nor is trying to be Google or some form of “Google Killer.” So going to Wolfram|Alpha and searching for recipes is not going to work as you would expect. I’ll get in to this a little bit more below, but there is a tendency to compare everything to Google these days — and conclude Google is better and will smash these pesky sites in short order. The key difference is “computable knowledge” versus “searchable web”. In case you don’t know, Wolfram is the Stephen Wolfram who created Mathematica. He also wrote A New Kind Of Science. I never met Wolfram, seems like a bright guy, I wondering if he went to the Donald Trump school of branding.
Here is what the W|A website tells us:
As of now, Wolfram|Alpha contains 10+ trillion of pieces of data, 50,000+ types of algorithms and models, and linguistic capabilities for 1000+ domains. Built with Mathematicaâ€”which is itself the result of more than 20 years of development at Wolfram Researchâ€”Wolfram|Alpha’s core code base now exceeds 5 million lines of symbolic Mathematica code. Running on supercomputer-class compute clusters, Wolfram|Alpha makes extensive use of the latest generation of web and parallel computing technologies, including webMathematica and gridMathematica.
There were several words in this paragraph that grabbed my attention. Of course the parallel computing cluster stuff is always interesting, but I was attracted to the over 50K algorithms and symbolic computing. I admit I am somewhat of a closet AI (Artificial Intelligence) dweeb, so when someone is talking about symbolic processing I always take notice. When they throw in parallel computing I’m all ears. I also have another interest in AI. I believe it is going to be where the killer HPC applications will arise. Note, I did not mean a “killer as in death” application but “killer as in wildly successful.” Just in case you have seen too many Terminator movies.
In addition, I am not suggesting “pie in the sky AI” which in the past was often oversold and under-delivered promises. I am talking about using AI methods to solve real problems and move us further away from the hardware and closer to what is it we really want to do — finish our work faster so we can go home early. For instance, suppose you ran a series program using a matrix of input parameters (often called parametric processing) What if an application scanned your results and told you “Run 17 seems to lack the same consistency of the other runs”. What exactly does this mean, who knows, but if you are the researcher, I would guess you would be giving the run 17 data a closer look. In any case, you get the idea. Symbols allow a higher level of manipulation. Which brings us back to W|A.
The first query, not sure if that is the right word, I entered was Beowulf Cluster. I thought something that runs on a cluster might have clue about clusters or parallel computing. The response, Wolfram|Alpha isn’t sure what to do with your input. Okay, I thought that is not fair, think “computable knowledge.” Some text on the side of the site had some suggestions, entering a date for instance. I thought my birthday is coming up, I’ll enter that, 5/29/1956. Aside from learning that I have been alive, 52 years 11 months 20 days. I was suitably dismayed that I am soon to be 53 and that the only other information W|A was able to glean is I share the same birth date with La Toya Jackson.
Another suggestion was a city. Very well, I entered bethlehem pa (where I live) and there was a nice hit of information. The location, population, surrounding population, local weather, nearby cities were displayed. I noticed that the independent (parallel) information streams seemed to arrive at different times as indicated by the progress bars on the page. Let’s take it a step further I thought. In a few weeks, I am planning to attend my college reunion at Juniata College in Huntingdon, PA. A good “computable query” would be how long is the drive from bethlehem pa to huntingdon pa. No dice. Maybe I could have rephrased the question better, but still this is a somewhat normal “computable” query that gets asked all the time. (By the way, Google did not have any luck in providing a direct answer either.)
Moving on, I thought, well this seems to be a “web enabled” version of Mathematica, well then let’s give it a softball question, is 123456789 a prime number. The response was false and to my surprise it gave me the two nearest primes (123456761 and 23456791). I noticed that like the Bethlehem City query, there was some “similar” or “close” answers given in addition to the “exact” answer. That can be useful, I thought. Time for another softball question, how many people on earth. The answer include more than I wanted, life expectancy, population density, median age, which is helpful I suppose within the context of the question.
While deciding on my final question, I had noticed a pattern in the answers. First, if W|A did not know the answer, it simply stated “Wolfram|Alpha isn’t sure what to do with your input.” If it did know the answer, it provide very neat and organized results with, as I mentioned, some peripheral data as well. In both cases, there was not extraneous “almost” answers. This may be unsettling for some as I believe we are conditioned to expect some kind of answer from web queries. Personally, I find it better to have something tell me I don’t know rather than try and give me any answer.
I like the idea of W|A and I expect it to get better and more robust in the future. I think it is an indication of things to come, but trying to be a general computable search tool on the first go round it a tall order. My hope is it does not turn off too many people who do not understand this tool. I plan to spend more time with it as I did not have a chance to play the chemistry and physics domains. I believe you may see more similar domain specific tools in the near future. One of the most interesting features would be historical context based computation where my past queries help shape future queries. I understand Google may do some of that already. Of course, I would have been overjoyed if my final query, is La Toya Jackson a prime number produced the response No, and that is a stupid question or better still, Yes, want to know why? Unfortunately, I may have to wait or ask La Toya at the birthday party.
Douglas Eadline is the Senior HPC Editor for Linux Magazine.