The July 2011 GTUG meet took place in Gordon House last night. There were two distinct and quite different components to this meeting: first up, Martin Omander, Developer Programme Manager, talked about the Google Prediction and Storage APIs; this was followed by a panel discussion on Hackerspaces and Formal Education.
Martin’s talk opened with some thoughts on focusing on your strengths leading into some ideas around letting Google focus on core IT capabilities and building interesting and useful application on top of these. The first of these base capabilities is storage – Google has built a large, reliable storage and distribution infrastructure and provides an easy to use interface to it. Martin demonstrated gsutil which is a command line tool which can be used to manage files in Google Storage. More specifically, gsutil provides an interface which is very natural to *nix user to perform operations such as list, copy, delete and rename files, etc. Also, it was remarkable how responsive it was – from the demonstration, it seemed to differ little from local file manipulation. Martin also made reference to the Data Liberation Front, which is a Google initiative to make it easy to move data in and out of Google products and services. To highlight how serious Google are taking this, he demonstrated how easy it is to use gsutil to copy files out of Google Storage and into Amazon S3. Google Storage is free to use (with a quota of 5GB) until the end of this year.
The second section of Martin’s talk focused on the Prediction API. This is a set of machine learning techniques which sits behind a very easy to use API. There are basically two ways to use the API – the first is in a classification mode in which training data is classified in some sense and the predictor can determine how to classify new data; the second mode is one in which there is a single scalar output value which is driven by a number of input variables. To demonstrate the first mode, Martin trained the system with a set of sentences which were either English, French or Spanish – the training set was not so large and the system was trained in a matter of seconds. Then, he used an English sentence as input and the system was able to automatically determine that the input was in English. To demonstrate the second set, Martin chose property price data from California: the input data set comprised of the property price and a number of parameters which influence price including date of build, number of bedrooms, number of bathrooms, total area, etc. Martin trained the system with this data and then was able to choose quite arbitrary values for the variables and the system produced an estimate of the value of the property. Google Prediction API is free to use up to some modest quota; more intensive applications incur some costs.
Martin finished off the talk by noting that these are two powerful and quite fundamental tools which Google are offering, on which many interesting and useful applications can be built. [Slides here.]
Martin’s talk was followed by a quite different event. Based on some feedback from a recent survey we did and some feeling that talks are sometimes not the most engaging format, we decided to run a panel discussion. This first panel was largely borne out of a discussion in the Schoolhouse after a previous GTUG and it focused on Hackerspace and the Formal Education system; it arose from the observation that hackerspaces are interesting, creative places which seem to be quite divorced from education and we wondered if this is as it should be, or can there be benefits from bringing them closer together.
We were delighted to have Bill Liao chair the event. The panellists were John Looney, Engineering Manager, Google, James Whelton, school leaver and entrepreneur, Mark Cunningham, student, member of TOG, Martin Mitchell, Software Engineer, IBM, founder member of TOG and Mark Deegan, Lecturer, DIT. Bill set the scene by making a few initial comments on hackerspaces, their history, how many there are in Ireland, etc. and then he talked about what the education system is doing. Each of the panel members then gave a brief intro and their view on the topic. As it turned out, all of the panel saw significant deficiencies in the education system and understood that the hackerspaces can much more easily be places of learning and innovation. A panel in concord is never going to be as much fun as one which is strongly divided, however, Bill still managed to extract an interesting discussion from the panellists on issues ranging from Google’s hiring policies to how a hackerspace approach could be constructed in a way which is compatible with the formal education system to whether hackerspaces should be self-financing. The floor was invited to give comments/questions and many people wanted to have access to the microphone. Some were commenting on their experiences in hackerspaces, some were commenting on the student societies in universities and some were interested in whether there is any point trying to bring these two worlds together. Clearly, it was a topic that people were interested in and had opinions on.
The discussion itself was somewhat inconclusive – it was clear that all involved believed that hackerspaces are good and interesting places and most believed that they learn more there than in the formal education system. However, there was no consensus on whether it makes sense to have hackerspaces more closely aligned with the education system, what (if any?) benefits this might bring and what problems would arise. Fodder for another discussion, perhaps…
The next GTUG event will be the HTML5 hackathon in AOL on Aug 6/7. The next talk session will be on Aug 30th, subject tbc.