Updates
5/5/2009. Perceptive reflection by David Weinberger on the importance (or lack of it) of WolframAlpha.
28/4/2009. Here is some informative and sceptical discussion about WolframAlpha on Slashdot.
29/4/2009. NB. On Tuesday 28 April, the Berkman Center for Internet & Society at Harvard
University will hosted a face-to-face and remotely accessible preview of the WolframAlpha system. Participants included WolframAlpha founder
Stephen Wolfram and Jonathan Zittrain, Professor of Law. Here is David Weinberger's contemporaneous report from the session, and here is a link to a 55 minute interview by Weinberger with Wolfram.
16/4/2009 - "Making 'expert level' knowledge accessible to everyone" - Stephen Wolfram
I signed up for a preview of WolframAlpha, and today I took part in a webinar presented by Stephen Wolfram, in which he put the very impressive WolframAlpha through its paces by feeding its search field with a wide range of brief (English) queries e.g. "next solar eclipse in Chicago" - including from participants - generating, quickly, elegantly presented well-structured answer-screens, with facts, data, graphs, time-lines, and links to sources of data (but not images). Think of it as a "CIA world fact book" that can generate its fact-sheets flexibly and on the fly, about a huge
range of topics, possibly taking account of the location of the person making a query if the GPS coordinates of the user's device are incorporated into the query.
(One query failed e.g. "arctic ice thickness", and Wolfram's certainty that that particular query would fail made me less confident than I otherwise would have been that the demonstration was not focused mainly on queries that Wolfram knew would work.)
According to Wolfram there are four "big pieces" to the technology (this is written live, so please excuse errors):
- curation of data - huge amounts of curated data imported - via a "curation pipeline" that flags and eliminates poor data (with expert human input, and QA, sometimes in real time) - Wikipedia is one of the sources that is used;
- computation - making anything that science and engineering has made computable computable within the system, drawing on the existing mathematical capabilities of Wolfram's Mathematica, which has ~5 million lines of software code (see also MathWorld);
- free-form input linguistic understanding component (mapping short utterances into a symbolic form from which answers can be computed - understanding the "pigeon-languages" that people use for queries in different domains) - contrasted with "big-data" computation of the "Google Translate" kind;
- automated presentation component e.g. how to present data in a way that fits the domain and makes the responses useful.
WolframAlpha will initially be (and is hoped to remain) free, with (a) priced professional version(s) sitting alongside it (which will provide, for example, scope to download data for reuse in various formats - PDF, XML etc); and there will be options for WolframAlpha technologies to be accessed by organisations, privately, for (or used on) their own corporate data.
My overall reaction from the presentation to WolframAlpha is that it:
- has the potential to change the face of web search;
- will be of specific and potentially great value in teaching and learning;
- will put competitive pressure onto Google to offer services that aggregate and present data from diverse sources, rather than "just" search results.
(Added 17 April after sleeping on the webinar.) But is there a downside? Probably. WolframAlpha will act as a closed "knowledge mediation engine", sitting between users and data. Google and its equivalents mediate by greatly influencing what sources searchers find; but they leave searchers to make their own judgements thereafter, and they allow users to dig deeper by following hyperlinks. In contrast, WolframAlpha provides a packaged summary, exposing, it seems, neither the rationale behind its choices as to which facts to include in its answers, nor the semantic relationships upon which the rendering of the presented data depends. So, though I'm impressed, I'm also cautious.
Related post
Freebase: lowering the barriers for participation in the Semantic Web. Guest contribution from Phil Rees, 7 October 2007.
Seb. Is there any public URL that points to the webinar?
I came across several of Wolfram's sites and did not stumble on anything alike ...
==
Mirek. Not that I know of. The webinar finished at 22.30 UK time on 16 April. Seb
Posted by: Mirek Sopek | 17/04/2009 at 06:00