SlideSpeech @slidespeech-blog-blog - Tumblr Blog

When Words Become Commands

Stephen Wolfram has given notice that "Something Very Big Is Coming" with the introduction of a new symbolic programming language.

In essence, this new language has the potential to allow a natural expression of intent or action. With other computer languages, a great deal of work and thought is required to build up data structures in particular formats so the computer can "process" things or offer users the ability to manipulate or visualize data of specific types. Hence we have separate word processing and image processing and music software. We have separate spreadsheet and database tools. The work of building all these data-type specific software tools has required specialized skills: programming, database administration, user interface design. The Wolfram Language may wrap the lower-level implementation of these computer-oriented tasks in an interpreted higher-level language which just does what we tell it to do. The words of the language become commands for action.

See Spot Run

Imagine the creative possibilities of a language which could take "See Spot Run" as a command to generate an animation of a running dog. Such language already works in our heads. The input triggers processes which cause us to "see" (visualize or imagine) that "Spot", a dog, is moving quickly (running). Computer languages typically don't have this expressive power due to ambiguities of meaning. "See Spot Run" could also be a complaint you make to your dry cleaner about the ketchup stain on your tie: Notice Stain Grew-larger.

The whole point of the first computer programs was to calculate the answer. IBM's Watson has shown how a computer can play the game Jeopardy! by entertaining many possible answers and, effectively, guessing at which one is best. The guesses may not always be correct, but the "thinking" behind the answers has the advantage of being fast, vast and at least somewhat transparent. When the system "explains" how it arrived at the results, misunderstandings can be clarified. For example, with the new Wolfram Language, if "See Spot Run" is too ambiguous then "See Dog Run" might resolve the ambiguity.

MOOCs or MOOQS?

Ubiquitous computing -- the idea that we can now have unlimited, any time, anywhere access to computing resources -- implies learning experiences could be atomized. We might not need a course in anything if we can get answers to all our questions on demand. Thus, the potential exists for Massively Open Online Courses (MOOCs) to be replaced by Massively Open Online Question Systems (MOOQS). Google Search and Stackoverflow are two examples, both of which depend on some human having worked out and posted an answer for others to find. The Wolfram Language may ultimately be able to interpolate between human-generated answers, making it possible to ask the system novel questions which tap into the language's algorithms to compute novel answers. Such computing promises "intelligence amplification" -- rather than "artificial intelligence" -- as it offers convenient and natural access by humans to algorithmic power without requiring artificial expression of the intent in computer code.

#wolfram #IA

SlideSpeech documents learning Node RED in this series of presentations. (Google sign-in required). What is Node RED? A visual design tool for interactivity and connectivity in the "Internet of Things".

Slides Images from SharePoint. Slide Notes from LibreOffice.

The SlideSpeech server has been running on Microsoft Azure® since September 3, 2013 including a new, native slide image conversion process. This makes it even easier to use the Google Chrome Extension or Keynote on the iPad to create SlideSpeech presentations.

Tony Backhouse listens to John Graves sing via Skype. A virtual voice lesson.

#voice #lessons #skype

MOOC Disruption in Perspective: Trains to Cars

Steve Kolowich’s piece “The MOOC ‘Revolution’ May Not Be as Disruptive as Some Had Imagined” in the The Chronicle of Higher Education left me flabbergasted. The article reflects institutional inertia at its worst. Change is coming. The questions are when? and for whom?

To understand what the word “disruptive” really means, consider this statistical table U.S. Railroad Employment Statistics (1947-Present) and this PDF of Motor Vehicle-related employment (1990-2013), and Employment of Teachers and Software Developers, May 2012.

Railroads: 85% drop in employment

Motor Vehicles: 53% drop in employment peak to trough

Millions of teachers. Under a million software developers. For now.

Whether you think traditional educational systems are like railroads or motor vehicle manufacturing, new MOOC-like systems are going to “transport” some learners to their learning destinations in new ways. That scenario becomes disruptive when teachers transition from “driving trains” to “building cars” and students transition from “riding trains” to “driving cars”. Looked at from this perspective, some form of disruption appears highly likely in coming years — very possibly involving a lot more personal “travel”.

Who will go on these new, digitally-enabled learning journeys? The “disruptive” changes may take place far from traditional educational settings. If Twitter “disrupted” politics in the Arab Spring, imagine what MOOCs might do in other countries.

Anecdotally, I just saw a woman wearing a full, black burqa — her face completely covered — typing away intently on her white iPhone …

Kiwi PyCon 2013: A SlideSpeech-ified Conference?

At Kiwi PyCon 2009, I gave a presentation titled Voice Interaction in Python to the breakout room in Christchurch. As I recall, 2 people were in the room. I put the talk on Slideshare, however, painstakingly adding an audio voice over recording and syncing it to the slides. Four years later, that talk has been viewed over 7,000 times!

At Kiwi PyCon 2011, SlideSpeech was still called Wiki-to-Speech, but the intrepid Ben Healey managed to add speaker notes to his slides to make a talking version of his talk, Document Classification using the Natural Language Toolkit, using a computer voice. Here's the video version with over 1,000 views.

From Audrey Roy, we had Python and the Web: Can we keep up?.

Jeff Rush contributed The Magic of Metaprogramming.

And Glenn Ramsey gave us Design Patterns in Python.

Those three presentations took their scripts from notes I "transcribed" during the talks.

A few months later, in February 2012, SlideSpeech Limited was founded and the open source Python project served as the prototype for the Java-based presentation converter service available today at http://slidespeech.com. You can learn more about the SlideSpeech system from the FAQ.

In the last two weeks, I've learned enough about two languages I had never worked with before, Go and Dart, that I was able to write and deploy a Go-based web application called Assignment Sheet Builder, running on Google AppEngine, and use it to begin to create an online "course" about Dart.

I would like to persuade you that the combination of SlideSpeech + Assignment Sheet Builder, or something similar, could be useful for sharing the talks and workshops of Kiwi PyCon 2013 for the online audience.

There are several ways to put voice over audio together with slides using SlideSpeech.

The simplest and best is to type the script of the voice over in the speaker notes of your slides. This is simple because it involves word processing instead of public speaking + audio recording/editing. In my time trials, it cuts the time of making a talking presentation in half relative to recording and editing audio or video. The text-based approach is best because the resulting script can be updated and improved at any time like a wiki page and is searchable (like a Wikipedia article).

If you want more control over the voice, you can use the SlideSpeech plugin for LibreOffice to drive the voice on your PC or Mac while making a screen cast of your talk.

If you want a human voice recording, Audacity allows chopping an audio recording into pieces based on labels (the feature is under File > Export Multiple ...) and SlideSpeech allows uploading these, one by one, onto your slides. The results sound like this (46 slides, 1 hour long).

It would be great if Kiwi PyCon -- which contributed so much to my learning over the past 4 years at AUT -- could be the first SlideSpeech-ified conference, with a fully searchable set of conference presentations.

What do you say?

Learn about the research behind Google's 20% policy and how it can be applied in K-12 education in this MOOC on Schoology. Access code: ZXQ2B-8CWMV

Making Learning Socially Responsible

Connecting the grassroots Mozilla Maker Party with the Richard Branson's top down B Team approach, Socially Responsible Learning offers a way forward. Let's learn to share knowledge collaboratively.

Connecting the Dots

Rory O'Brien presented on Rapid eLearning in deMOOC from New South Wales, Australia. Lesley Gardner presented "Where has all the noise gone" from University of Auckland in New Zealand. Students in Mr. Salmon's 3rd and 4th grade classes presented on Plants from Edmonton, Alberta, Canada. But wait! These presentations were made around the world, yet I watched them all from my mobile phone at home!

The vision of a planet united in learning, any time, any where, is real. We needed the ability to freely share our knowledge. Now we have it. The next step is relatively simple: linking different pieces of learning material together. As I suggested to Mr. Salmon, if his students had each worked on one aspect of plants, instead of each repeating the same (or similar) presentation, the collection of presentations could become a whole unit on plants for other students to study. This is how Wikipedia collaboratively built up into the largest reference work on the planet. We can now achieve the same scale with learning materials. I call this Socially Responsible Learning, where the output of learning becomes the input for other learners.

If you are as excited as I am about this prospect, please join the Google+ community called Using Google Apps as a Free Learning Management System which you can read more about here.

Socially Responsible Learning

In March, 2010, David Wiley gave a TEDxNYED talk, titled Open Education and the Future, explaining the significance of sharing teaching materials. Despite the growing popularity of massively open online courses (MOOCs) and other video-based resources such as Khan Academy, learners experience a high level of variability when they try to learn from current systems (see the report Measures of Effective Teaching).

Fragmentation of learning opportunities makes this variability systematic. Learners are not tapping into shared learning materials the way they tap into shared reference materials at Wikipedia. We lack a focal point for such a collaborative effort.

Video-based systems have the advantage of reaching a large broadcast audience, but they are built at significant cost in an inherently static format which inhibits collaboration and continual improvement. So, instead of fostering a collaboration around personalized learning, MOOCs and other forms of delivered instruction are following the same top-down model which crushes creativity, as Ken Robinson explained.

These problems call for solutions where learners publicly create and share learning opportunities for one another, so they are learning in the system, rather than from it. Michael Idinopulos called this working "In-the-Flow" rather then "Above-the-Flow". Call it Socially Responsible Learning.

Given the economics of Wikipedia, publicly sharing knowledge as it is acquired in an open, collaborative system has the potential to reduce the cost of learning 1000 fold. For everyone.

SlideSpeech and Science

Steven Levy's book, In the Plex, quotes Tim Armstrong, Google's top sales executive in New York as saying, "[Our job was] bringing science to the art of advertising and being able to scale the art of advertising through science." SlideSpeech aims to take the same concepts and apply them to learning: bringing science to the art of learning and being able to scale the art of learning through science.

In search, Google tapped into our fundamental need for answers. When we have some idea what we want, Google can help us to find it. The focus at Google is getting that match right, fast. Unfortunately, learning has two characteristics which prevent search answers from being learning solutions: one is the meta-problem of knowing what to ask and the other is the importance of learning from mistakes.

The traditional educational system, like the traditional system of advertising before Google, actually depends on ignorance about what works and what doesn't work to justify the prices charged. A "good" school does not necessarily offer measurably more learning; quality is gauged by reputation. In advertising, Google found very effective ways to measure and price the actual desired results of an ad: sales of the advertised product. In learning, we need to build equally sophisticated systems to measure and price the desired results from a learning experience: increased knowledge and skills.

The keys to success here are scale and data. Learners own their knowledge. Learning should be an investment with a future payoff. Yet most people lack data about the extent and value of the things they know. The extent or amount of knowledge is currently measured using the extraordinarily coarse-grained metric of the diploma or degree. Meanwhile, the cost of a degree has gone up significantly relative to the return from having a degree in terms of earnings. Student loan debt is a huge problem.

Google determines the value of an ad using an auction. Advertisers bid for the opportunity to attract customers to their product. Simultaneously, consumers vote (via click-through) for the advertisements and products they find most attractive. Thus Google's system helps make connections between what consumers want to buy and what advertisers have to sell. The effectiveness of each ad is measurable at the end of the process when there is a conversion, or sale.

The traditional educational system, like the traditional system of advertising before Google, involves pre-payments. This puts all the risk on the buyer. If the knowledge acquired turns out to be worthless or the advertisement fails to attract customers, the schools or media channels keep the money while the students or the advertisers take the loss. SlideSpeech aims to put payment where it belongs to align everyone's incentives toward maximizing learning effectiveness: at the end.

Learners need meta-data about their learning: what they know, what they need to learn next and the value of having specific knowledge. This data can be organized in a system which includes fine-grained progress tracking, visualized connections from the current state of knowledge to possible future states, and the cumulative price paid by learners who previously reached each future state. Knowledge has time value. Given the speed with which knowledge advances, the value in learning new knowledge comes from learning it sooner and faster. Thus, the first learners of something new should pay the most, while everyone else who comes later should pay less. This aligns with the reality of internet distribution: once content is created, it can be distributed globally at scale.

Teachers in the SlideSpeech system are paid for completions. Students must pay for their completions to have their progress tracked. This is the restaurant model, where you pay for what you eat after you eat it. In a restaurant, you might go in the back and wash dishes if you can't pay your bill. SlideSpeech offers a similar option: create content for other learners and earn out the amount owed to get the completion credit. Thus SlideSpeech becomes a collaborative, crowd sourced platform for learning materials like Wikipedia, but monetized at the point of progress tracking.

Some SlideSpeech Basics

Making computers talk is cool.

If you think about it, writing was invented so we could record our talking. Now writing can play talking. The playback process is called text-to-speech. While most people don't like the sound of a computer voice, the technology has recently advanced to the point where it is hard to tell what is real and what is not. For example, I have no trouble imagining "Veena" is a real person talking.

Google shows how valuable it can be to search text. As we write more text-to-speech scripts, they will be searchable as well, making it possible to look up and watch (or, more precisely, watch, listen and interact). So as I see it text currently beats video in three ways: we can write/edit text, we can use full text search and we can script interactions using text. Meanwhile, video beats text by recording body language and intonation. However, simulated bodies are becoming more and more expressive, as shown in the movie Avatar (2009) and a computer voice has been developed to sing opera, so you can imagine what may soon be possible with more expressive voices.

Disruptive new technologies have a well understood life cycle. They start small and don't work very well initially. Once they become established, however, they begin a process of continual improvement and refinement until they surpass the capabilities of the incumbent technology.

I think we are on the verge of developing a system of collaboratively produced, computer-delivered learning materials. Unlike the current video-based approach where some person needs to deliver the presentation, the fact that we can now work together to get the computer to deliver presentations offers the opportunity for many hands to make linked, interactive learning materials. At Wikipedia scale.

Thus, the real revolution in on-line learning isn't the Massively Open Online Course (MOOC). That approach is essentially open in only one direction: the content flows from the source to the crowd. With Wikipedia, the crowd is the source.

The cost/benefit of Wikipedia is astounding. With an annual budget for 2012-13 of only US$42.1 million, Wikipedia serves over 500 million people every month. Compare that with the US$68.4 billion budget of the US Department of Education which serves far fewer people; the population of the entire United States is only 316 million. What if learning cost under 10 cents instead of over $215 per capita?

Imagine the simplicity and impact of being able to search for anything and get an interactive, verbal explanation. Or if the explanation doesn't exist, being able to make a searchable verbal explanation using only a web browser. No video required. Just slides and text.

This is an idea to change the world and dramatically accelerate the distribution and development of knowledge.

Please join SlideSpeech on your preferred social media and share your thoughts.

Balloon Club is a Google+ community, described as "a community for schools and other organizations and individuals that are interested in transforming learning, with or without technology, and have adopted or are looking at Google Apps for Education."

#teachtheweb

Taking the first steps to learn HTML in Thimble.

#teachtheweb

ocTEL webinar summary (27 slides)

#ocTEL

Comparison of two MOOCs with the Learning Map idea

#octel

Trending Blogs

Recently Viewed Blogs

SlideSpeech