Fortnightly Mailing

The Unreasonable Effectiveness of Data

Here is a current article by Google's Alon Halevy, Peter Norvig, and Fernando Pereira [376 kB PDF], from the IEEE's March/April issue of Intelligent Systems. It is intelligible to a lay person, and it contrasts lucidly two broad approaches to extracting meaning from information. Though the article does not use the terms, you might describe these two approaches as taxonomic and statistical. These excerpts - the second is the concluding paragraph - give you a flavour:

"The biggest successes in natural-language-related machine learning have been statistical speech recognition and statistical machine translation. The reason for these successes is not that these tasks are easier than other tasks; they are in fact much harder than tasks such as document classification that extract just a few bits of information from each document. The reason is that translation is a natural task routinely done every day for a real human need (think of the operations of the European Union or of news agencies). The same is true of speech transcription (think of closed-caption broadcasts). In other words, a large training set of the input-output behavior that we seek to automate is available to us in the wild. In contrast, traditional natural language processing problems such as document classification, part-of-speech tagging, named-entity recognition, or parsing are not routine tasks, so they have no large corpus available in the wild. Instead, a corpus for these tasks requires skilled human annotation. Such annotation is not only slow and expensive to acquire but also diffi cult for experts to agree on, being bedeviled by many of the diffi culties we discuss later in relation to the Semantic Web. The first lesson of Web-scale learning is to use available large-scale data rather than hoping for annotated data that isn’t available. For instance, we find that useful semantic relationships can be automatically learned from the statistics of search queries and the corresponding results  or from the accumulated evidence of Web-based text patterns and formatted tables, in both cases without needing any manually annotated data."

"So, follow the data. Choose a representation that can use unsupervised learning on unlabeled data, which is so much more plentiful than labeled data. Represent all the data with a nonparametric model rather than trying to summarize it with a parametric model, because with very large data sources, the data holds a lot of detail. For natural language applications, trust that human language has already evolved words for the important concepts. See how far you can go by tying together the words that are already there, rather than by inventing new concepts with clusters of words. Now go out and gather some data, and see what it can do."

Top 5 internet priorities for any next Government

Via Owen Bader, Tom Steinberg from the excellent organisation - see * below - mySociety (link to 50 minute video, and to interesting Gordon Brown phone call on the bus story)  provides a list of the top 5 major things any government of any developed nation should be doing in relation to the Internet, as Tom sees it at the start of 2009.  Three in particular caught my eye:

"2. Free your data, especially maps and other geographic information, plus the non-personal data that drives the police, health and social services, for starters. Introduce a ‘presumption of innovation’ – if someone has asked for something costly to free up, give them what they want: it’s probably a sign that they understand the value of your data when you don’t.

3. Give external parties the right to interface electronically with any government or mainly public system unless it can be shown to create substantial, irrevocable harm. Champion the right fiercely and punish unjustified refusals with fines. Your starting list of projects should include patient-owned health records, council fault reporting services and train ticket sales databases. All are currently unacceptably closed to innovation from the outside, and obscurity allows dubious practices of all kinds to thrive.

5. When people use your electronic systems to do anything, renew a fishing license, register a pregnancy, apply for planning permission, given them the option to collaborate with other people going through or affected by the same process. They will feel less alone, and will help your services to reform from the bottom up."

* mySociety has two missions. The first is to be a charitable project which builds websites that give people simple, tangible benefits in the civic and community aspects of their lives. The second is to teach the public and voluntary sectors, through demonstration, how to use the internet most efficiently to improve lives. Examples of mySociety systems include FixMyStreet and TheyWorkForYou. See also this link to the work of software developer Chris Lightfoot, who died two years ago, and who contributed to many of mySociety's projects.

Macintosh OS X on a netbook for £400 - Guest Contribution by George Roberts

OK, I don't think you can call it a "Mac netbook" but this is how I got Mac OS X 10.5.6 to run on a Dell mini 9. It was easier than I thought.

Blurring the lines between net-books and e-book readers

The main brain behind the wonderful screen on the OLPC laptop was Mary Lou Jepsen. The screen had a low power consumption, high screen resolution, and worked well in direct sunlight. Jepsen's start-up company Pixel Qi is tackling the design and production (by manufacturers not by Pixel Qi) of cheap, but above all, highly functional screens. Here Jepsen provides a brief update:

"We are making screens that are more readable than regular laptop screens, and our studies show they are as comfortable to read as electrophoretics in existing ebook readers. But these screens also update quickly, show video, are sunlight readable and (with the backlight on) show full color. In addition, our display screens integrate with touchscreens of many different varieties (an analyst I know is aware of over 200 different efforts by various companies in new and improved touchscreens)."

10/1/2008 - The 75$ laptop.

Giving a boost to Carol Twigg's work on how to ensure technology interventions are (cost) effective

Donald Clark has been reading and summarising - to good effect -  Carol Twigg's work for the National Centre for Academic Transformation. (See Favourite e-learning research; Carol Twigg's research on cost-effectiveness; Effective learning through e-learning; reducing drop-out rates.)

On 7/9/2005 Carol did a keynote at the 2005 ALT Conference in Manchester. At the last minute, Carol was unable to attend the conference in person, so she delivered the speech remotely. A technical issue between the US and the UK prevented her from using the microphone in the studio she was in, so she did the whole keynote with a mobile phone. Despite this she held the attention of 500 people for an hour. Stunning. Carol's slides from her presentation, and my summary of Carol's answers to questions after her keynote are here on the ALT web site.

As an aside, I think that there is more than a streak of "not invented here" in relation to UK attitudes to Carol's work; ALT^* and others have been regularly referencing  the work in responses to policy consultations; but the underlying "points of fact" that the National Centre for Academic Transformation has established have tended either to be ignored, or marginalised with explanations as to why "it ain't like that in the UK".

^*Disclosure. I work part time for ALT.

Brewster Kahle - the man behind the Wayback Machine

Last week's Economist Technology Quarterly carried a two page spread about Brewster Kahle, the Internet entrepreneur and philanthropist behind the Internet Archive / Wayback  Machine. Below is  a 20 minute December 2007  TED talk by Kahle. Slick it ain't (Kahle looks to be the kind of person who simply does not do slick), but the overview provided into the scale of, and processes behind, Kahle's endeavor to create a free, online collection of human knowledge makes it worth watching in full. Fittingly, one of the three copies of the Internet Archive, is in Alexandria.

One to watch: Yubikey, for an economical method of two-factor authentication

Updated and republished with link to a review of Yubikey security issues, 9/3/2009, with further updates and fixes to broken links on 23/1/2013.

[Added 9/3/2009. Here, with thanks to Positive Internet, is a friendly but quite critical assessment of Yubikey and some of its security flaws by Dr. Fredrik Björck, a security consultant, along with (added 23/1/2013) an update showing the fixes that Yubico was already making back in 2009.]

Yubico is a Swedish start-up that has come up with a clever and economical way of providing three-factor* authentication. The two normal factors are your user name and password, but what if someone or some software system has acquired these? Three-factor authentication involves a device owned by you needing to be part of the process, thereby making it more difficult for someone to access your web-based services. The problem with three-factor systems is their expense, and the YubiKey is cheap enough even in small quantities for it to be realistic to deploy on a mass basis, provided there is cheap enough access to the necessary validation service.  It is also designed to work with OpenID, and would reduce the risks of phishing that are associated with using OpenID for higher-value services.

What if you lose your key? The effect of this would depend on how the service is configured. You could, for example, be allowed "n" two-factor accesses without your key, or, once you've used your credit card to order a new key, or phoned your service provider to be issued with a new key, then be allowed two-factor access until, with your replacement key, you can switch three-factor authentication back on.

Yubico looks to be pitching to be put into use by Google as an add-on for users of Gmail and Google Apps; and you would imagine that Banks must be interested in this kind of service as well (I believe some already provide it).

* Yubico describes its system as a two-factor service, in the sense of something you know (your user name and password), and something you have (e.g. the YubiKey).

With thanks to Ufi's Dick Moore for highlighting this. (23/1/2013 - that was then, this is now.)

Sustainable Energy - without the hot air

Updated 1/1/2010

Sustainable Energy - without the hot air by David MacKay

If Tony Juniper, former Executive Director of Friends of the Earth says "For anyone with influence on energy policy, whether in government, business or a campaign group, this book should be compulsory reading", and Robert Sansom of EDF Energy  says "At last a book that comprehensively reveals the true facts about sustainable energy in a form that is both highly readable and entertaining", then the book is probably hitting the nail on the head. David JC MacKay's is a Professor in the Department of Physics at the University of Cambridge [and, from September 2009, Chief Scientific Advisor to the Department of Energy and Climate Change], and his book is available for purchase or for free download. There is also a podcast from a one-hour talk MacKay gave at the Cambridge Science Festival, on 12 March 2008, for which you can also access a reasonably matching set of slides, in HTML format.

With thanks for the link to Nic Tweddell

Integrating technologies into educational ecosystems, by Pierre Dillenbourg

elearning-reviews continues to provide a first-rate service with this review of Pierre Dillenbourg's Integrating technologies into educational ecosystems. Here is an excerpt from the Chrysi Rapanta's review:

"The article is divided in four basic sections called respectively: ‘Faded Myths’, ‘Convergences’, ‘Examples of Integrated Learning Environments’ and ‘Orchestration’. Starting from the first one, Dillenbourg addresses four basic myths, referring to overstatements about learning technologies that are currently losing their previously high status.

The first myth is about the overestimated effect of the use of media in education, stressing the fact that media itself do not bring pedagogical innovation, in the same way that teachers do not necessarily provoke good learning outcomes.

The second myth is about the futuristic approach regarding the upcoming ‘success’ of e-learning applications. According to Dillenbourg, teachers’ current technological skills are fairly well developed so that lacking skills do no longer count as an excuse for the poor development of TEL practices.

The third myth consists of the assumption that a technological application yields the best results when it is as similar to traditional practices as possible. Recent research in the educational field has shown that this is not the case. In addition, the author suggests that designing TEL environments in a way that imitates traditional campuses or institutions does not help teachers exploit the whole range of possibilities that new technologies offer.

Dillenbourg finishes with the fourth myth considering the teachers’ role. The teacher today is neither a ‘sage on the stage’ nor a ‘guide on the side’. To provide all the meta-knowledge ‘hidden’ in the learning contents he has to be in the middle of the scene, whatever the means of teaching are."

Note to self. Find out if Pierre Dillenbourg wants to speak in the UK.

WikiMatrix - a tool to compare wikis

From Fred Riley at the University of Nottingham, via the ALT Members' email list, here is a tool for comparing wikis, which Fred speaks highly of.