Apple Siri. The Butlers are coming

Siri

Web2.0 democritised e-publishing and data creation in a friendly way for the masses and low and behold there are 182 million websites available on the net in 2011. Some of these websites have the lions share of the content (Facebook, Flickr, Google, Amazon etc) but collectively it's a grand publish of human thoughts, artefacts, wishes and desires.  What a wonder!

Creating content is one thing but leveraging insights across the content is more difficult. In truth there is still too much information for humans to effectively use and we find ourselves to be a gear in the machine rather than the driver - connecting systems together, cutting, pasting and rekeying.

I want to ask simple questions of my computers and have powerful background processing bring me the answer. Questions like "Which famous guitarists endorse products but don't use them in their live shows?" A query like this would require text analysis of the question to understand the meaning, scouring the net for famous guitarists,  checking which brands they claim to use in endorsements, checking their live 'kit' on websites, picture recognition of what guitars they are using, comparison of statements versus reality and then provide a weighted response based on the volume of data processed. Not easy and lots of key tapping.

Voice Control on the Bat Computer

It won't always be this way.

Batmans computer has been serving him for years (in the fictional world of DC Comics) controlled by his voice helping him fight crime. He simply asks the computer a question while he is driving or smashing heads of supervillans together and his Batcomputer gets back to him with the summary. Questions like  "Cross reference the known toxins that the Joker uses with chemical factories in the vicinity of Posion Ivy's locations over the past three months" are answered with ease. If a clarification is needed then it asks Batman. All achieved using natural language as the interface.

Digital buddies, assistants and advisors are here already for consumers albeit in the form mostly of recommendations engines and advertising systems. Last.fm helps reduce the millions of bands down to something I might like based on my previous listening while Amazon  advises me of books and products I might enjoy based on my previous activity.

These systems help us save time and slash the options and possibilities down to something we can handle. The volume of data falls below our eye and we can concentrate on the richer questions and answers.

For me the biggest aspect of the new iPhone 4S release was Siri - the virtual assistant. I think that as innocuous as it might appear on the surface (fixing calendars, looking up the weather, setting reminders) it is one of the first believable assistants that interact with consumers in a rich way.

Over time this service will grow to understand your accent, tone of voice and mood. It might voluntarily ask you what's wrong or question your commands if it thinks you are acting irrationally. It will potentially develop it's own personality and it will be answering more and more complex queries. Multiple Siris may even communicate and negotiate with one another to save their 'owners' from corresponding back and forth needlessly. Young children that can't type and older people may begin interacting with computers in richer ways. Siri may begin to find it's way into robots and other household devices outside of mobiles.

It's exciting and this is only the beginning. Others have tried to provide this kind of service but none have had the design and user base that Apple have in order to make it 'stick'.

I'll be watching this one carefully.

Circles and Ladders with Google+ Contact Classification Paradigm

Grouping contacts is an impossible feat isn't it? We have to add Bob to Sport, Work, Musician, 'Allowed to Call after 10pm' and all those other groups we never keep up to date.

Like all user supplied up-front people classification systems Google+ Circles can quite quickly turn into hierarchy ladders when you manage your contacts using them, especially in a social context. 

Contact grouping, grading and intimacy-scoring questions arise like : "Why am I not in your Personal folder?" "Why am I only in Acquaintances?", "Why am I not in group X?"

It makes for unhappiness not to mention all the manual labour of managing those connections. 

The flat monism of classifying all your people as simply 'friends' and allowing the system (not you) to speculatively match between profiles managed by the identity owner is elegant. It causes less arguments over status and how other people classify you. 

Baboons would be relieved to have such a thing. 

Flat'ish, loosely coupled metadata overlaps between people such as : attended same school, favourite band is x, graduated in Kent, holiday in France, has photo of Mt Everest provide a more resilient model in the end for both programmatic and humanistic reasons. It also a more natural petri-dish for harmonious social groups when developing new services.

The degree to which this metadata is enhanced as you interact with 'your people' defines the living breathing classification of what they mean or meant to you. It allows for relationship management (manual and auto) between the people you already know and it also allows for the emergence of machine dialogue such as 'People You Should Know'

It's dynamic and weighted through use - it's no longer the leaden categories of 'Work', 'Home', and 'France'.

last.fm - The Big Biological Model gets a cold

last.fm is a service I have used and admired for a long while now and I became a paid subscriber for £3 a month (or thereabouts) a few months back which is something notable in a sea of free music services.

Their service really does manage to play music 'like the music I like' while deftly avoiding the stuff I don't like. I've trained it over the past few years like a puppy to respond to my commands of love and hate and now it does a great job in creating my personal radio station.

Their simple delta of being able to mark something as 'not liked' provided them with the extra dimension of customer modeling that the other recommendation and fuzzy logic engines sorely missed. Recommendation models without an 'unlike' are akin to physics environments where there is no 'reaction' to the 'action' - the map is too one dimensional.

I remember reading a list of the jobs that last.fm had available in 2009 and you would have been hard pressed to discern between their job descriptions for marketing and technical staff and job listings for a biotech scientist. The guys at last.fm understand nodal modelling and that the real social graph looks more like something you see in a petri dish in a microscope - constantly changing, fault tolerant, nodal, duplicated, overloaded and alive.

Some howcome last.fm are making some cardinal sins with their customer relationships at the moment?

1. Introducing subscriptions that can only be paid for by PayPal. This was moderately annoying when I was in Latin America recently and my PayPal was suspended after one too many transactions in Brazil making their fraud algorithm jumpy.

2. Removing the two key features that most subscribers pony up their cash for - streaming personal playlists and 'loved tracks'.

The on-demand streaming isn't something I use last.fm for as I use it as my personal auto-pilot radio station and secondly as my database of music likes in the cloud. The truth is I use Spotify for on-demand '' listening. The changes in service however do bother the legions that had moved from their iPods to iPhones and Android devices to have on-demand on the move.

What bothers me is the way that they are making the changes.

No push of email to subscribers to alert them of the situation and a terse statement on the site inevitably sees a raft of users angrily hitting the forums and web. The path for resolution is simply to cancel your subscription if you are unhappy about it.

Problems with licensing seems to be the battle cry with most upheavals in music services but did they really try an explore all the options they could have taken?

• Two tier subscriptions with a higher delta for on-demand streaming.
• Aborting the free model and charging a nominal fee for all users.
• Offer different services/rights to different countries

.....or even less radical - cancel the services that they were going to but run an 'ease the shock' campaign pre-warning users of the reasons why they need to do this.

We're listening to last.fm but they aren't listening back.

Hunting Robots and How to Survive a Robot Uprising

The Pentagon recently announced that they want suppliers to provide a "Multi-Robot Pursuit System" that will let packs of robots "search for and detect a non-cooperative human".


I recommend boning up on robot pursuit avoidance now. How to Survive a Robot Uprising is the book you need. You'll read it in an hour and it may save your life one day or at least give you suggestions on how to get your leg out of an annoyed hoover.

The Prediction Model

I've become increasingly interested in the overlap of biology and computing over the past few years. It began with the realisation that the web and stock market are really biologic in nature with their fault tolerance, nodal shape, replication of information and distributed locus of control and was further prompted by work I undertook on AI systems for Advertising and Social Computing solutions.

I’m enjoying this overlap developing into a moderate obsession and I am trying to steer my thinking on all things computing into a more ‘biologic fashion’. I’ve always been a strong believer that people involved in one discipline can offer fresh insights on other sciences and that a good set of ‘first principals’ can work well cross domain. This cross pollination was the grease that helped the machine of the Industrial Revolution into being and obliquely it’s also the reason I give for sporting sideburns like some Manchester factory owner.

This post is inspired by Jeff Hawkins who is doing work into models of the brain and attempting to derive an overarching theory of the brain which is something that, despite the reams of data we have on the brain, we are as yet unable to articulate. His talk was on the use of a Prediction Model as the primary approach to developing a theory of the brain and he got my mind racing.

After graduating from Cornell in June 1979 he read a special issue of Scientific American on the brain. In it Francis Crick lamented the lack of a grand theory explaining how the brain functions.^[3] Initially, he attempted to start a new department on the subject at his employer Intel, but was refused. He also unsuccessfully attempted to join the MIT AI Lab. He eventually decided he would try to find success in the computer industry and then try to use it to support his serious work on brains, as described in his book On Intelligence

Jeff thinks that the reason we still haven’t managed to define intelligence well is that we don’t have this overarching theory of the brain or more accurately – intelligence. Jeff postulates that the brain isn't like a powerful computer processor and that instead it’s more like a memory system that records everything we experience and helps us predict, intelligently, what will happen next.

Things like these stop me sleeping at night and last Sunday I leaned over to my girlfriend at 2am and whispered to her “I have to write some stuff”.

I slipped out of bed and knocked up the notes below. They are presented here un-edited and what you see is the first pass brain dump of some of my thoughts and concepts surrounding a Prediction Model (It's probably best to click on one and open up the set in Flickr and view from there).

If you are involved in this area at all I would love to hear from you as I intend to delve deeper. Physics has alot to add to this area with work in quantum theory and calculations surrounding boundaries of event horizons for black holes all being of relevance to the model of the brain and prediction.

Photosynth and how the 'collective image memory' is harversted

We're building a collective digital memory with all those:

• votes and ratings
• comments and blogs
• tags and bookmarks

We can put this data on google maps, and provide strong links between place and time as well as invent applications that use this data to create new environments. We don't even need to use the common map metaphor to see our data with IBM's wonderful tool 'Many Eyes' which allows us to analyse data in interactive graphs and visualistions. Data can be processed by simple XML allowing for automated feeds of information and graphic representation such as the example below:




And then there is some next level image-onomy or whatever new paradigm term we need to invent that Photosynth ushers in. A technology acquired by Microsoft and originally developed by Blaise Aguera y Arcas.

It allows a feed of photos to build up a map of the earth and places not just using flyover images by aeroplanes or satellite data but by using our own photographs and even illustrations. Photosynth uses public images and it doesn't matter whether these photos are taken by a £10 disposable camera or a posh SLR - it can stitch them together and produce a never ending tapestry that allows you to move around geographic areas and locations with ease.

With Photosynth you can:

• Walk or fly through a scene to see photos from any angle.
• Seamlessly zoom in or out of a photo whether it's megapixels or gigapixels in size.
• See where pictures were taken in relation to one another.
• Find similar photos to the one you're currently viewing.
• Send a collection - or a particular view of one - to a friend.

Zooming in might have you moving through 10 photos using your own as a starting point. Your landscape shot of the fair ex-mining town of Cowdenbeath on your digital camera might be part of a family of 1000 photos of Cowdenbeath. Using this pool of images like stones in the middle of a pond you can step and zoom in deeper and deeper to find the Forth Road Bridge in detail when it was just a red spec on your own photo.

Photosynth takes data from everyone - from the collective memory of what the world looks like. A model emerges of the entire earth as our own photos get tagged with other peoples metadata and the mesh of linking becomes tighter and stronger. The network effect continually enriches the space and easily provides cross user and cross model experiences and information.

This is the real semantic web or 'Web3.0' along with the Social Graph developing through the use of people networks. These inferences are taking a life of their own and one can only wonder at what Web5.0 might be.

There's a great demo hosted by TED where Blaise runs through the application with jaw dropping effect.

Photosynth modestly state "Our software takes a large collection of photos of a place or an object, analyzes them for similarities, and displays them in a reconstructed three-dimensional space."

This experience is on the web to try right now but be warned Mac fans - this web experience is PC only for now.

Cheap Ideas for advancing Biologic Computing

One of the biggest problems in Biologic Computing today is the predictability of bacterium's movements...

How about this simple, relatively cheap project a cross functional team might be able to do at a University...

- Get feeds from the IBM public visualisation tool - specifically pictures of datasets. They can be Social Network activity feeds (or hub/colony'esque data). Here's one of many examples
- Grab the visualisations of them as well as the raw data
- Run some visual pattern matching software to compare these images against bacterium imagery - at varying scales of magnification
- Do the same pattern matching on the numbers
- See if anything interesting pops up in the pattern matching

Developing continually running real world feeds (API's/RSS or otherwise) from these types of public systems to visual biology computing resources would be potentially useful.

Their benefit is that they are
- continually updated, for free, and have simple XML descriptions of data.
- a constant public feed allow large elements of automation in such a project (bar the human analysis of 'matches' by the system)
- non-proprietary in nature and will 'out', in the end, for generating useful patterns vs custom expensive data capture

If you are up to stuff like this then I'd love to know about it. Mail me.

Recommended Reading: "Genesis Machines" by Martyn Amos

---