Wolfram Alpha Goes to R Systems for Horsepower
May 16, 2009
CNet’s Stephen Shankland wrote “Wolfram Alpha Gets Supercomputer Boost” here, and I think the story will super-boost the hopes that Wolfram Alpha’s search system will crush the Google. Rooting for a Google smasher is a bit of a pundit trend these days. With Wolfram Alpha supposed to be officially alive as I write this, I think the notion of the world’s 66th fastest supercomputer can make those Google haters experience an adrenaline rush. Mr. Shankland provides a link to the November 2008 list of supercomputers, and when I looked at the list I did not see the Google listed. The reason? Supercomputers are indeed fast, but they by themselves are not exactly what’s required to declaw Googzilla. Search and content processing is an interesting technical challenge and raw speed does not frighten to death the opposition. I learned from Mr. Shankland:
The system, called R Smarr, has 4,608 processor cores using 576 quad-core “Harpertown” Xeon machines, 65,536GB of memory, and high-speed InfiniBand data-transfer connections, according to the Top500 site and a Dell case study on the system (PDF). It also uses both the Red Hat Enterprise Linux and Microsoft Windows HPC Server operating systems, according to the Dell paper. Alpha requests will be served from five co-location facilities, Wolfram Research said. There actually are two supercomputers in the project, with nearly 10,000 processor cores total and hundreds of terabytes of hard drives.
I wonder if this commercial for Dell servers and Intel CPUs will indeed humble the arrogant Googlers? I have to keep reminding myself that dear old Google has been chipping away at the technical problems that keep most competitors from chewing into Google’s share of the Web search market. So what is it now? A decade for Google’s search effort? Wolfram Alpha has been in the search game for what now, a couple of years?
Should be interesting to see if the newcomer can do what Fast Search & Transfer, Microsoft, and Yahoo, among others, have been unable to do. If I were a betting man, I think Wolfram Alpha comes out of the gate with long odds.
Why?
CNet’s Tom Krazit reported here “Wolfram Alpha’s Launch Delayed Amid Glitches”. Hmmm.
Stephen Arnold, May 16, 2009
Universal Answer Engine
May 16, 2009
In the gap created by Wolfram Alpha’s sort of start tonight, I poked through my files. I came across “How to Build a Universal Answer Engine: Ten Vital Principles” here. After a week at a conference and dozens of conversations about whizzy new search systems, I must admit I am a bit jaded. True Knowledge, on the other hand, is excited about the idea of a Universal Knowledge Engine. This is a Web log post about True Knowledge’s comuter system designed to answer users’ questions on any subject. I must admit that I am skeptical. The questions have to be in text. So much for equations? The questions must be in a language I speak, which may or may not make my questions intelligible. Enough of this skepticism. The write up lays out “principles”, and I am not comfortable repeating each. I can highlight two principles and offer a comment:
Principle 4 is “The only truly scalable way to learn envryitng is by allowing users to contribute.” I don’t really disagree, but I think only a small percentage of a user community contributes information. To get around this problem, True Knowledge taps into Wikipedia. I give True Knowledge credit for mentioning that some users are not to be trusted. The problem is that with only the motivated contributing, the system has to find some way to determine the likelihood that a particular contribution and item are likely to be “correct” or “trusted”. This is a pretty complicated task, and I think that it is worth noting that the Google and others are beavering away on this problem.
Principal 8 is “All fqacts need sources and these need to be available to the user.” This sounds like provenance, which is related to principle 4. I saw a demo by a Stanford professor where provenance and uncertainty were query modes supported by the system. True Knowledge seems to be in step with this line of inquiry. Calculating these “values’ will consume a chunk of computer time slices.
To wrap up, the principles are interesting. A number of companies are in the question answering business, and these organizations will need deep pockets to pull off a service that keeps me happy.
Stephen Arnold, May 16, 2009
Arnold’s Life Is Tweet Available
May 16, 2009
Short honk: one of those for-fee write ups that contain real information, not the quacks of the addled goose, is now available in the May 2009 Information World Review. You can find details about the column and the opinion piece “Life Is Tweet for Real-Time Search”. My editor, Peter Williams, has done an excellent job with the article. If you can snag a copy of this UK publication, you can get some chunks of Tweet stew served up. The information will stick to your ribs, unlike the marketing honks of the contributors to this free Web log. Click here for more information. I wrote: Twitter is a hybrid information service. Want to know more. Subscribe to IWR.
Stephen Arnold, May 16, 2009
The Pain in Spain Is Tending to the Inane
May 15, 2009
Read “Recording Industry Tries To Shut Down Search Engine In Spain Without Allowing It To Defend Itself” here. If true, the Internet she be changin’. Search requires content processing. Robots index and point. Software becomes the problem.
Stephen Arnold, May 15’ 2009
Videosurf Update
May 15, 2009
Since VideoSurf’s birth in mid-2008 (Beyond Search reviewed it at http://arnoldit.com/wordpress/2008/09/23/videosurf-video-metasearch/), it’s offered up a beta version (I wrote about it at http://arnoldit.com/wordpress/2008/10/18/videosurf-looking-for-wave-of-new-users/), and now it’s opened up the API so developers can install “visual search” on their own sites. VideoSurf promotes itself as the only video search engine that can search and “see” inside videos to index content rather than depending upon tags and descriptions that can produce spam. The API allows accress to videos that can be selected to relate to site content; site can also tailor the displays to their promotional needs. See more about the API availability at http://www.videosurf.com/blog/search-and-video-lookup-apis-1039/. This could be a smart choice for sites looking to rev up their content while keeping it relevant.
Jessica Bratcher, May 15, 2009
YAGG: Bytes of Clay
May 15, 2009
Googzilla stubbed its paw today. You can read the official explanation here. YAGG means yet another Google glitch. The once alleged online uptime champ turned chump. Millions inconvenienced. Enough said.
Stephen Arnold, May 15, 2009
Brands as Gravity
May 14, 2009
In the online world, there are Jupiters, suns, and asteroids. Traffic sorts itself out in ways that are gravitational. A big brand gets lots of traffic. The asteroids generate less customer pull. Why’s this important? Online sites without pull are not likely to pull the traffic. Sad but true. Quality may be defined as lots of clicks. IT Pro here reported data that underscores the gravitational pull of 10 brands in the UK. The headline told the tale: “Top 10 Web Brands Get Half of UK Traffic.” The top three online brands were Facebook, Microsoft, and Google. The most interesting comment in the report was this remark:
Second-ranked MSN/Windows Live slid nearly a percentage point to 9.2 per cent, but added over a billion minutes in the past year. Third-ranked Google gained 0.4 per cent market share to 5.3 per cent, adding 950 million minutes.
Facebook appears to be the winner, which may have implications for the host of social challengers now available. In the UK, social seems to be the pull.
Stephen Arnold, May 14, 2009
Web Wide API: The Battleground
May 14, 2009
If you wondered what was the driver for the API snowstorm, read “Can Amazon Be the Default Payment API for the Web?” here. The author aaronchua did a good job of explaining the logic behind a single Web API for online payments. The issue is not multiple payment systems. The call is for a single payment system. Assume this happens. Monopoly, right? The APIs are important to Amazon, Google, and others. Winner takes all is logical, right?
Stephen Arnold’ May 14, 2009
Google, Micro-Blogging: Makes Perfect Sense
May 14, 2009
Google, the tarantula of the web, purchased Jaiku in October 2007, a service that allows it’s users to gather micro-blogs from other Web sites. The content can be viewed via the Web or by mobile phone. Google open sourced Jaiku in January 2009, just as Twittermania was gaining momentum.
Google’s decision could be a vote of confidence for open source, or it could be a response to Google’s failure to gain traction among the Twitterati.
Sites like Twitter, Flickr and MySpace each offer their own twists and user-friendly ways of appealing to mass amounts of micro-bloggers and furthermore, potential customers, but using a site that collects each feed and makes it accessible through one’s cell or computer, just makes sense.
In a business world where it’s crucial to keep in contact and notice emerging trends, it would be easy to spend your entire day signing-in and utilizing the sites previously mentioned. Google, despite its success in other search spaces, recognized the importance of real time search in its recent Searchology mini-camp.
The reality may be that Twitter, despite the hype, may be a challenger to Facebook. Facebook’s recent redesign nods in the direction of Twitter. Google, on the other hand, acknowledges the importance of real time search, making a distinction between Twitter’s indexing of tweets and the larger, Google-scale challenge of real time search of Web content.
“Less talk and more indexation” is the goose’s cry.
Hunter Embry, May 15, 2009
Some Google in the White House
May 13, 2009
A month ago, I received a call from a journalist asking about the Obama White House’s uses of Google. I did not answer the question because big time journalists ask me question, and I am not a public library reference desk worker any more.
One insight can be found here. Google said:
App Engine supports White House town hall meeting
In late March, the White House hosted an online town hall meeting, soliciting questions from concerned citizens directly through its website. To manage the large stream of questions and votes, the White House used Google Moderator, which runs on App Engine. At its peak, the application received 700 hits per second, and across the 48-hour voting window, accepted over 104,000 questions and 3,600,000 votes. Despite this traffic, App Engine continued to scale and none of the other 50,000 hosted applications were impacted. For more on this project, including a graph of the traffic and more details on how App Engine was able to cope with the load, see the Google Code blog.
How Googley is the Obama White House? Pretty Googley I hear.
Ste3phen Arnold, May 13, 2009
.
-
- Subscribe to Beyond Search
Feature archive
News archive
-
