Monday, April 24, 2017

Snapchat’s smart pivot into an AR company but is AR ready for learning?

Augmentation, in ‘augmented’ reality, comes in all shapes, layers and forms, from bulky headsets and glasses to smartphones. At present the market has been characterised by a mix of expensive solutions (Hololens), failures (Google Glass, Snap Spectacles) and spectacular successes (Pokemon Go, Snapchat Filters). So where is all of this going?
SnapChat has pivoted, cleverly, into being not just another messenger service, but the world’s largest Augmented Reality company. Its ‘filters’, that change every day, use face recognition (AI) and layered graphics to deliver some fun stuff and more importantly, advertising. It is a clever ploy, as it plays to the personal. You can use fun filters, create your own filter with a piece of dynamic art or buy one. It’s here that they’re building an advertising and corporate business on designed filters around events and products. That’s smart and explains why their valuation is stratospheric. Once you play around with Snapchat, you get why it’s such a big deal. As usual, it’s simple, useful, personal and compelling. With over 150 million users and advertising revenue model, that works on straight ads, sponsored filters and sponsored lenses (interactive filters), it has tapped into a market that simply never existed.
Snap Spectacles
Snap Spectacles was their interesting foray into the glasses augmented market – but more of a gimmick than realistic consumer product. Targeted only at Snapchat users, you can’t really wear them with regular glasses and all they do is record video – but, to be fair, they do that well. However, as with Google Glass, you feel like a bit of a twat. Not really a big impact product.
With its AI driven interfaces – point head, gesture or voice recognition, it is neat but at $3000 a pop – not really a commercial proposition for Microsoft. As for the ‘experience’, the limited rectangle, that is the field of view, is disappointing, and ‘killer’ applications absent. There have been games, Skype applications, 2D & 3D visualisations but nothing yet that really blows the mind – forget the idea of Sci-fi holograms, it’s still all a bit ‘Pepper’s Ghost’ in feel, still tethered and has a long way to go before being a viable product.
Magic Leap
Bit of a mystery still as they are a secretive lot. Despite having raised more than $1.4 billion from Google, Alibaba and Andreessen Horowitz, it has still to deliver whatever it is that they want to deliver. Mired in technical problems, they may still pull something out of the bag with their glasses release this year – but it seems you have to wear a belt with some kit attached. Watch this space, as they say, as it is nothing but an empty space for now.
Pokemon Go
We saw the way this market was really going with Pokemon Go, layers of reality on a smartphone. Photographic Camera layer, idealised graphic map layer, graphic Pokemon layer, graphic Pokestops layer, GPS layer, internet layer – all layered on to one screen into a compelling social and games experience. Your mind simply brings them altogether into one conscious, beautiful, blended reality – more importantly, it was fun. This may be where augmented reality will remain until the minaturisation of headsets and glasses get down to the level of contact lenses.
I still prefer the full punch VR immersive experience. AR, in its current form is a halfway house experience. The headset and glasses seem like a stretch for all sorts of reasons. You simply have to ask yourself, do I need all of this cost and equipment, to see a solar system float in space, when I can see it in 3D on my computer screen? There are clearly many situations in which one would want to ‘layer’ on to reality but in many learning situations, there may be simpler solutions.
So let’s look at specific learning outcomes that could be delivered and enhanced by Augmented Reality.
1. Explanations
Explanations, causes, rules, processes… delivered as text, audio, 2D, 3D images in physics, chemistry, biology, hydraulics, pneumatics, maths and so on. The superimposition of explanatory diagrams, arrows, flows and explanations, have obvious theoretical and practical applications, delivering explanations in the context of the real world. Performance support is another option with ‘contextual’ learning to increases retention & recall. The delivery of explanations, determined by your own personal needs and identified context is appealing in training.
2. Problem solving
Explore real places: museum, art gallery, virtual excursion, virtual experience, real factory and solve real problems in maths, science, language, historical, architectural and natural environment. This problem solving can be task driven in induction/on-boarding, fault finding, maintenance tasks, language learning and so on.
3. Learn by doing
We largely learn by doing but are largely taught while doing nothing. With a hands-free device you can return to more appropriate forms of learning by doing. Motion sensing & GPS helps enormously and you can’t fool it easily, which is useful in assessment. Do experiments/tasks in science, practical tasks and learn skills, cheap devices in AR could revolutionise vocational learning.
4. Social learning
Groups (Pokemobs) out in real world, searched & completed tasks showing that the social dimension in learning can be enhanced. AR, such as Hololens, does give you contact, via Skype, with others, so that they can draw and you see it appear on your display. So there are social possibilities.
5. Tutor-led
There are signs that Magic Leap have a tiny assistant that sits in your hand, then there’s Skype on AR, which can offer tutoring at a distance for groups of learners.
Tutor-led/assisted, with a real or created tutor (AI-driven bot or avatar) can make the learning more personal & adaptive.
6. Deliberate/spaced practice
Learning can be enhanced by deliberate practice. AR gives us the opportunity to practice, again and again, repeat a skill in different contexts, offer adaptive and tutor-led deliberate practice.
7. Simulations
Critical training for the police, fire & emergency with realistic augmentation of bombs, fires, damage and casualties is all possible. Control layers can be used to test & train simultaneously to deliver lessons about optimal tactics. Things can appear and happen in certain timeframes. Already used by NASA, closed, limited or open-world simulations are all possible.
8. Assessment
For vocational training one could test learners (uniquely identified) in real time as assessment would not be separate from performance. Assessment at a distance is also possible.
9. M-learning
As we saw with Pokemon Go and now with Snapchat filters, AR gives you a compelling reason to use your phone, a powerful, personal and portable AR device. It is AR that may open the floodgates to new and fascinating forms of m-learning.
10. Habitual learning
Mobile behaviour is highly habitual and AR could mean frictionless and more habitual learning. It opens up possibilities in informal learning, making blended learning and 70:20:10 realisable.
RR – Real Reality
But before you start, do the RR test – that’s Real Reality. It may be better to stick to the real physical world. Consciousness is, after all a form of augmented reality – it is reality reconstructed by the brain. A text or podcast allows us to layer in the imagination, a form of augmentation that can be more useful in learning than trite imagery. AR can be delivered via screens. You need to think carefully before letting this technology, especially in its immature form, lead you towards expensive projects that may be better delivered by more conventional technology.

Augmented reality is not one thing – it’s best seen as a way of layering, altering or interacting with reality. At present all the action is on smartphones. In a sense Google Maps and GPS-like applications are augmentations. Pokemon Go showed the potential, albeit with a flash in the pan application but it is Snapchat, with its filters that has had the most sustainable success. Their move towards augmentation has been clever and you can expect a lot more from them. I’m less convinced by Hololens and Google seems, once again, to have failed in product development with Google Glass, as there have been no further releases. As usual consumers are attracted by fun not functionality.

 Subscribe to RSS

Friday, April 21, 2017

AI fail that will make you gag with disgust……

One of the first consumer robots were vacuum cleaners that bumped around your floors sucking up dirt and dust. The first models simply sensed when they hit something, a wall or piece of furniture, turned the wheels and headed off in a different direction.
The latest vacuum robots actually map out your room and mathematically calculate the optimum route to evenly vacuum the floor. They have a memory, build a mathematical model of the room, with laser mapping, 360 degree cameras, can detect objects in real time and have clever corner cleaning capability.  They move room to room, can be operated from a mobile app - scheduling and so on. They will even automatically recharge when their batteries get low and resume from the point they left. Very impressive.
That’s not to say they’re perfect. Take this example, that happened to a friend of mine. He has a pert dog and sure enough, the vacuum cleaner would bump into the dog on the carpet, turn and move on. The dog was initially puzzled, sniffed it a bit, but learned to ignore this new pet, as something beneath his contempt as top-dog. Cats even like to sit on them and take rides around the house.
Then, one day, the owner came back, opened his front door and was hit by a horrific wall of smell.  The dog had taken a dump and the robot cleaner had smeared the shit evenly across the entire carpet, even into the corners, room by room, with a mathematical exactitude that was superior to that of any human cleaner. The smell was overwhelming and the clean up a Herculean task on hands and knees, accompanied by regular gagging.

The lesson here is that AI is smart, can replace humans in all sorts of tasks but doesn’t have the checks and balances of normal human intelligence. In fact the attribution of the word intelligence, I'd argue (and have here), is an anthropomorphic category error, taking one category and applying it in a separate and compeltely different domain. It’s good at one thing, or a few things, such as moving, mapping and sucking, but it doesn’t know when the shit hits the fan.

 Subscribe to RSS

Thursday, March 30, 2017

Do bears shit in the woods? 7 reasons why data analytics a misleading, myopic use of AI in HE

I’m increasingly convinced that HE is being pulled in the wrong with its obsession with data analytics, at the expense of more fruitful uses of AI in learning. Sure it has some efficacy but the money being spent at present, may be mostly wasted.
1. Bears in woods
Much of what is being paid for here is what I’d say was answers to the question, ‘Do bears shit in the woods?’ What insights are being uncovered here? That drop-out is being caused by poor teaching and poor student support? That students with English as a second language struggle? Ask yourself whether these insights really are insights or whether they’re something everyone knew in the first place.
2. You call that data?
The problem here is the paucity of data. Most Universities don’t even know how many students attend lectures (few record attendance), as they’re scared of the results. I can tell you that the actual data, when collected, paints a picture of catastrophic absence. That’s the first problem – poor data. Other data sources are similarly flawed, as there's little in the way of fine-grained feedback. It's small data sets, often messy, poorly structured and not understood.
3. Easier ways
Much of this so-called use of AI is like going over top of head with your right hand to scratch your left ear. Complex algorithmic approaches are likely to be more expensive and far less reliable and verifiable than simple measures like using a spreadsheet or making what little data you have available, in a digestible form, to faculty.
4. Better uses of resources
The problem with spending all of your money on diagnosis, especially when the diagnosis is an obvious limited set of possible causes, that were probably already known, is that the money is usually better spent on treatment. Look at improving student support, teaching and learning, not dodgy diagnosis.
5. Action not analytics
In practice, when those amazing insights come through, what do institutions actually do? Do they record lectures because students with English as a foreign language find some lecturers difficult and the psychology of learning screams at us to let students have repeated access to resources? Do they tackle the issue of poor teaching by specific lecturers? Do they question the use of lectures? (Easily the most important intervention, as the research shows is the shift to active learning. Do they increase response times on feedback to students? Do they drop the essay as a lazy and monolithic form of assessment? Or do they waffle on about improving the ‘student experience’ where nothing much changes?
6. Evaluation
I see a lot of presentations about why one should do data analytics  - mostly around preventing drop-out. I don’t see much in the way of verifiable analysis that data analytics has been the actual causal factor in preventing future drop-out. I mean a cost-effectiveness analysis. This is not easy but it would convince me,
7.  Myopic view of AI
AI is many things and a far better use of AI in HE, is, in my opinion, to improve teaching through personalised, adaptive learning, better feedback, student support, active learning, content creation and and assessment. All of these are available right now. They address the REAL problem – teaching and learning.
To be fair I applaud efforts from the likes of JISC to offer a data locker, so that institutions can store, share and use bigger data sets. This solves some legal problems but looks at addressing the issue of small data. But this is, as yet, a wholly unproven approach.

I work in AI in learning, have an AI learning company, invest in AI EdTech companies, am on the board of an AI learning company, speak on the subject all over the world, write constantly on the subject . You’d expect me to be a big fan of data analytics in HE – I’m not. Not yet. I’d never say never but so much of this seems like playing around with the problem, rather than facing up to solving the problem.

 Subscribe to RSS

Sunday, February 26, 2017

AI is the new UI: 7 ways AI shapes your online experience

HAL stands for ‘Heuristically programmed ALgorithmic computer’. Turns out that HAL has become a reality. Indeed we deal with thousands of useful HALs every time we go online. Whenever you are online, you are using AI. As the online revolution has accelerated, the often invisible application of AI and algorithms has crept into a vast range of our online activities. A brief history of algorithms includes the Sumerians, Euclid, the origins of the term (Al Khwarismi), Fibonacci, Leibniz, Gauss, Laplace, Boole and Bayes but in the 21st century ubiquitous computing and the internet has taken algorithms into the homes and minds of everyone who uses the web.
You’re reading this from a network, using software, on a device, all of which rely fundamentally on algorithms and AI. The vast portion of the software iceberg that lies beneath the surface, doing its clever but invisible thing, the real building blocks of contemporary computing – are algorithms and AI. Whenever you search, get online recommendations, engage with social media, buy, do online banking, online dating, see online ads; algorithms are doing their devilishly clever work.
BCG’s ten most innovative companies 2016
Boston Consulting Group publish this list every year:
  1. Apple
  2. Google
  3. Tesla
  4. Microsoft
  5. Amazon
  6. Netflix
  7. Samsung
  8. Toyota
  9. Facebook
  10. IBM
Note how it is dominated by companies that deliver access and services online. Note that all, apart perhaps from Toyota, are turning themselves into AI companies. Some, such as IBM, Google and Microsoft have been explicit on this strategy. Others, such as Apple, Samsung, Netflix and Facebook have been acquiring skills and have huge research resources in AI. Note also that Tesla, albeit a car company, is really an AI company. Their cars are always on, learning robots. We are seeing a shift in technology towards ubiquitous AI.
1. Search
We have all been immersed in AI since we first started using Google. Google is AI. Google exemplifies the success of AI in having created one of the most successful companies even on the back of AI. Beyond simple search, they also enable more specific AI-driven search through Google Scholar, Google Maps and other services. Whether it is documents, videos, images, audio or maps, search has become the ubiquitous mode of access. AI is the real enabler when it comes to access. Search Engine Indexing finds needles in the world’s biggest haystack. Search for something on the web and you’re ‘indexing’ billions of documents and images. Not a trivial task and it needs smart algorithms to do it at all, never mind in a tiny fraction of a second. PageRank was the technology that made Google one of the biggest companies in the world. Google has moved on, nevertheless, the multiple algorithms that rank results when you search are very smart. We all have, at our fingertips, the ability to research and find the things that only a tiny elite had access to only 20 years ago.
2. Recommendations
Amazon has built the world’s largest retail company with a raw focus on the user experience, presented by their recommendation engine. Their AI platform, Alexa, now delivers a range of services but it was made famous by its recommendations on first books, now other goods. But recommendation engines are now everywhere on the web. You are more often than not presented with choices that are pre-selected, rather than the result of a search. Netflix is a good example, where the tiling is tailored to your needs. Most social media feeds are now AI-driven, as are many online services, where, what you (and others) do, determines what you see.
3. Communications
Siri, VIV, Cortana, Alexa… voice recognition, enabled by advanced in AI through Natural Language Programming, has changed the way we communicate with technology. As speech is our natural form of communication, it is a more natural interface, giving significant advantages in some contexts. We are now in a position of seeing speech recognition move from being a topic of research to real commercial application as AI, in many forms but particularly deep learning and large data sets, have allowed some of the world’s largest tech companies to use it with hundreds of millions of customers; Apple, Amazon, Google, Microsoft, Samsung and others.
4. Translation
In translation, the recent shift in approach from large scale pattern matching to more focused AI techniques gave Google a gear change in efficacy. Deep learning translation is so powerful that it now works with any new languages, without the need for huge data sets. Ready translation on social media, real time translation on Skype are here now. Language hurdles can be overcome with realtime online translation, available for voice calls and instant messaging. Skype Translator uses AI, machine learning, so the more you use it, the better it gets.
5. Social
This is the age of algorithms. We open a file, it is decompressed, we save a file, it is compressed, we send a file, it is managed across a global network. We Skype, WebX, SnapChat, WhatsApp, Facetime – all of this is enabled by smart AI in terms of compression, networks and decompression. The underlying technology is fundamentally algorithmic. When we zip files, compress for transmission, decompress for use. Lossless and lossy compression and decompression magically squeeze big files into little files for transfer. On top of this are error correcting codes, mistakes that fix themselves, so that sound, pictures and videos can be saved, stored and retrieved without loss, especially across networks, where these clever algorithms maintain quality. Beyond this is the work of AI in determining news feeds and ads in social media.
6. Databases
The advent of big data means that the balance, in some contexts, has swung away from algorithms, towards the power of massive data sets. Nevertheless, when you use a database you use some clever algorithms. Databases are used for many forms of content storage and, although you may not know it most of the time, whenever you access a learning management system, VLE or learning content, you will have been using algorithm-driven databases. In other words, algorithms already lie at the heart of learning, albeit in an almost invisible and indirect way. We now see the emergence of blockchain, a distributed, hackproof, database structure, that may enable finance and learning applications of a different order.
7. Commerce
Public key cryptography is how encryption works and keeps your credit card details safe when buying stuff. Amazon, Ebay, PayPal, credit cards and the entire world of online retail would not exist without this algorithm. Spam filters, phishing, even higher order cyber-threats, are all handled by AI.
Going back to our top ten list of innovative companies. They all see software that learns, as an integral part of their products and services. Machine ‘learning’, products and services that, the more you use them, the better they get, places ‘learning’ at the core of their businesses. Yet there is another sense in which AI can deliver ‘learning.
As most learning is informal, not through formal online learning courses, most online learning, through search, social media, communications and other online services, can be said to be AI-driven and mediated. AI has enabled informal online learning. AI now also delivers AI-driven content creation, curation, chat and consolidation through tools such as WildFire. Adaptive learning is also being delivered in large formal courses. Adaptive assessment, automated essay marking, face recognition, typing recognition are also AI-driven. Even plagiarism checking is now AI driven. AI is the new UI. AI is also the new UI for learning.
Bright young AI mathematicians and coders, no longer yearn to work on Wall Street or in banks but in start-ups, incubators and business creation. This has been a long time coming but at last human talent is being directed, not towards the mere management of money, but the creation of new ways of creating jobs and shaping the future. The question remains, that the Age of the Algorithm may destroy more jobs than it creates. Nevertheless, for the moment, it holds the promise of getting us out of boom-bust cycles where maths was forever blowing financial bubbles, into maths that make things work. As we have revealed the potency of algorithms, one can’t fail to admire the elegance of these carefully constructed, magic, mathematical spells. They are stunningly clever.

 Subscribe to RSS