Can Cameras know what they are looking at?

It seems that a world where cameras will know what they are looking at has almost arrived. This is exciting stuff because it’s happened fast and it’s powerful.

Some 6 years ago I spoke with Roberto Chipola of Cam Uni Eng – who had created a demo of a camera pointing at Cambridge (UK) Regent Street and by “un-distorting” the street’s apparent perspective – could enable said camera to know precisely “where it was”. The thought of mapping the planet, so that this could be scaled everywhere seemed immense. Then within two years, Google’s street view arrived.

A flirtation 4 years ago, with Nick Kingsbury’s Cam Uni Signal Processing group’s complex wavelet transforms, also enabled a camera to pick out what shop in Burleigh Street it was looking at – but the scale up to global ambition was going to take a combination device software skills, communications skills and back end architecture skills – all in the same group. Now you’d have though that Cambridge would relish such a challenge but it stalled. Gameware Ltd were up for the challenge if truth be told and their time may have come.

Now we start to find a host of smart companies nibbling at the edges of this tasty dish… Xsight in Israel, Cortexica in London,

So now look at this Ted Video and others and see where the state of the art has moved to in just a few short years. You’ll see a map with 3d rendering and an in-built knowledge of its own 3d geometry – into which stills and video can be overlaid. This is huge. Other large software vendors have also spoken of pulling such frameworks together so that smartly mined information can be overlaid with location – in an easily publish/easy consume manor.

What would be great is to now see a slew of real/virtual world blends – with augmented reality character overlays – that make Layer look like child’s-play.

Lets see TBL & Nigel Shadbolt‘s open data initiatives intertwined with this new name space/framework and much much more.

The world, its location and it’s geometry have started to become an open framework for enormous amounts of application and fun. Lets use it.

A question to leave with – should this all be wrapped up in an application (or many) or should this be available through device primatives, and if the latter, how?

Amadeus’ Seed Fund – seeks to fund such early stage ventures.

Comments Off on Can Cameras know what they are looking at?

Filed under Ideas

Thought, “Compressed Sensing” and the maths behind Occam’s razor

“Praise be to heros neptune, the titanic sails at dawn, everybody’s shouting.. which side are you on.” BD

So I’ve been wowed once again by Wired magazine, which seems to have summed up some recent conversations with quite delicious serendipity.

This post relates to both thinking styles and to a new branch of exciting mathematics called “compressed sensing”.

Firstly let me ask you; did you ever do any advanced mathematics. If so (and I did a bit), you’ll remember the stage where you could always work through the problems from first principles. Then, BAM! it all stops working… you find that the only way to solve problems is to “somehow guess” the form of the answer, and then show that by working backwards, this answer fits the problem. (complex differentiation/integration).

This may seem geeky, so let’s come back down to earth. Some people have the skill to solve problems by thinking through situations from first principles; with the rules and the facts available. Some do it quickly and some slowly – an analytical approach – lets call it bottom up or “end state first”.  All very sensible. Others (and I confess I’m one) find this hard and take an approach which is best described as “guessing” the end state or answer, and then seeing if it fits all the facts available. If there’s no fit, throw the guess away and guess again (guesses are cheap after-all) and see if this fits all the facts. Lets call this top down or perhaps “end-state first”.

Both approaches are quite fine and just a matter of style. However, in a situations where there are too few facts or rules to allow us to use a “start to finish” approach, i.e situations where one has a sparsely populated data set – the “end-state first” approach is helpful. This occurs very frequently – indeed as seed investors, we’re faced with this situation very frequently.

Now, lets take Wired magazine’s article on Compressed Sensing. Here, a bunch of researchers have managed to deal with sparse data in a quite magical way. They have managed to reduce MRI scanner times by factors of 100x, and increase apparent photo resolution by 100x similarly etc etc. The break-through logic goes something like this. Imagine that instead of a nice high resolution photo, I have a lower resolution set of randomly spaced pixels. Now take some clever maths “l1 minimization” that “guesses” “end-state first” at a set of coloured blocks that fit amongst the random pixels, in such a way as to “match” the random pixels and make them fit into a pattern. Green dots must be in green blocks and yellow in yellow of course. Magic and maths then ensue – and it turns out that if you have enough random pixels – it’s mathematically very unlikely that the “block guess answer” is going to be incorrect – and just as magically, researchers are finding that they can “decompress” sparse data into amazing results. “It’s as though you can give me the first 3 didgits of a 10 digit bank account number and I can use these to guess the next seven!” quotes Emmanuel Candes. This is quite simply Occam’s razor instantiated in mathematics.

This stuff is for real – already the military are using it to use sparse enemy signal capture to decompress into full data….

What’s going on? How cab this be true? We all know that we cannot “guess the last seven digits” of an account code if given the first three unless there is a pattern. Yet we often spot patterns around us based on very limited data – Blink (Malcolm Gladwell) is surely about this if about anything. Blink’s premise is that humans have not only developed an analytical side to their brains but also an intuitive fast responding judgment pattern recognising side. It’s this that is at issue here. Thinking without thinking, working with seemingly too little data.

So what’s the point. Well, machines are being systematically taught to pull patterns from limited data in ever more magical ways. My fear is that in climates of uncertainty, and in a world that is changing so rapidly, where old data is wrong data; humans are tempted to reach for ever increasing amounts of data and proof, and are thereby increasingly going to fail. All to often, large quantities of data and proof is just not present until “it’s too late”. The moment will have passed by the time we have collected it – so I’m with Gladwell – Humans need to practice and harness their “compressed sensing” skills in line with the machines.

A serious point – anyone in the UK working on such maths and its application – or anyone who knows of anyone – please get in touch – I’d love to talk, I’d love to find applications and I’d love to fund.

Comments Off on Thought, “Compressed Sensing” and the maths behind Occam’s razor

Filed under Ideas

Place based Alumni Groups

“.. the only sound that’s left, after the ambulances go, is Cinderella sweeping up on desolation row.” B.D.

I have just spent three hours in deepest darkest south London, Tulse Hill no less. The trip there was akin to a trip through the favela’ with a full wallet – even the taxi driver was nervous. What’s happened down here – south London was on the up when I lived here 15 years ago but it’s all evaporated. Drugs, crime and real poverty. An amazing mix of properly fancy but housing association run building stock surrounded by some gruesome 1960’s concrete apartment blocks. I left this place after 8 years residence when I got mugged and have only been back to visit a dentist! Scary stuff. However, the people are real!

Now lets turn to Iris Lapinski, and her mobile apps initiative for social inclusion. Iris has brought together a (Dell Sponsored) programme whereby youths in south London are taught about mobile phone application design with a view to helping them get more digitally included in their community. The initiative is called CDI or Centre for Digital Inclusion and it’s excellent. Twitter #appslaunch. Before you click away ~ it’s excellent, one of the freshest initiatives I’ve seen for years, please read on.

I didn’t know what to expect from the South London launch and late on a Wed evening, I confess that I didn’t really want to go. However, I’m so very glad I did. After getting lost in the high trees estate looking for the venue, I snuck in the back.. late. What I saw and heard was  a lively, engaged local community having a discussion about how to get “digitally included”. “Who writes our apps” asked one, “I have lots of ideas ~ who owns the IP” asked another… “hey Mr. biggish cheese Dell ~ will you give us a job after the course” said a third. These were guys who mugged me?!@? (well of course not actually) asking about how they could write community-useful apps…. egged on by their mothers, local press and city councillors of course. Mr Alan Moore, this was engagement marketing, you’d be proud.

Post the discussion and over some excellent food.. I met the “yoofs”. Nice guys actually, wanting to believe and needing some hope. They loved the idea of writing a game app ~ but actually had never seen code. “Writin’ code.. wa’s vat?” In fact these guys didn’t know what software was, had never seen any (teachers teachers what are you missing). Quite shocking really ~ a quick view of some HTML source code on the PC next to the chicken drumsticks was quite a revelation to them. “Oh so that’s software”. Quite a hill to climb but well worth it.

I met some lovely cuddly lady councillors & trustees one of whom introduced Daniel her son (2:22mins into vid below) as soon as she heard that I “give away money for a living”. Daniel has a need for money for his work which sees him mentor “the fastest 17 year old in the country and keep the yoofs off the bad stuff”. A brief discussion with him sparked an idea and question:

Why isn’t there a “Tulse Hill alumni group”? i.e there may well be no-one still living there with money to donate, because the monied people who reached “escape velocity wealth” have moved on and moved out. However, they must have gone somewhere and if they were made aware of these initiatives Iris’ and Daniel’, they couldn’t help but GIVE.

So, HELP ~ does anyone out there know of any PLACE BASED ALUMNI initiatives, tools etc etc… it seems an obvious engagement mechanism.

Lastly, full marks Iris.. full marks Daniel.. you can be proud.

Proofing credit to CDP – thanks.

Comments Off on Place based Alumni Groups

Filed under Ideas

Internet-Scale Bus; how we do apps and data in future?

“… I was so much older then, I’m younger than that now!”

The older I get, the more excited I am by watching waves of innovation. OK I’m not a computer scientist; I’m a VC, but it’s becoming increasingly apparent to me that the world needs an internet scale bus. In much the same way as Tibco became huge on the basis of the “enterprise bus”, we need a new and bigger solution. Why do I think that? because this stuff is going on:
1. Applications are proliferating radically.
2. Integration is commercially hard, and technically brittle.
3. Data sources are increasingly open, unstructured and live.
4. The requirements for applications need to stay agile.
5. There is no clear integrator or hierarchy and so open build access is important.

Just take a look at the at the sheer variety of web and iPhone applications, now being followed by Android of course – project that forward to the iPad and beyond, and it becomes clear that we have an app bonanza. Perhaps modest wealth creation but massive activity. Now consider requiring all these applications to move at the speed of a usual “enterprise data integration project” and you’ll stop. Add a massive and exploding measure of “link Data Initiative – RDF data publishing”. Stand back and take a breath before cupping your head, and realise that time to market and application/business agility is crucial to success and lastly realise that this may well not be organised by W3C.

Consider Twitter – is this actually the closest and largest current version of an internet-scale bus that we’ve seen to date? It’s pretty light and crude but we have applications being built on it – there is a classic #tag publish and subscribe architecture and it has massive volume… not bad folks but there’s much much more to do.

This is the merchandising bus of the future where #products and #pricing come together with #persona and #situation -An example of this working in a real world app would go something like this; “Hi we’re the Audi app, we just noticed your call to Jaguar and happen to know that your Audi is 3 yrs old”… this is an example of a valuable merchandising opportunity for us. The conversation continues “there’s a restaurant 1.5 km away, it’s Sat lunchtime – have lunch with/on us and we’ll have the audi of your choice in the carpark for a test drive afterwards.” All of this can manifest as its latently available in the application on your iPad- should you wish it to be as its potentially hidden under a “Surprise me” button.

How on earth could this ever happen unless we have a tight/loose coupling of #products, #pricing, #persona, #situation, #offers, #analytics, #location etc etc…. This world is already emerging and will create a whole new set of industries and services over the coming 6 years.

Pre-post note: if you have a resource, technology or company in this area – we’re looking to invest.

Comments Off on Internet-Scale Bus; how we do apps and data in future?

Filed under Ideas

Analogy Engines – semantic concept description

Some London folks attended a very cool event hosted by BLN and TaylorWessing on 3rd Dec 2009 to network and discuss tech venture in Europe. There I had a fascinating interchange with David Sinclair of Imense and Camtology about “Concept Search”. I had a simple request – “please could you build me a technology where I could type into a search engine a Concept or Idea – and it would return matches, similarities, analogies and perhaps opposites.” Now David is a smart chap and said – yes! for £2m he could. This was cool enough but actually it was the discussion that was more interesting.

If a concept could be described appropriately!! and that’s the challenge, then concept search, mapping, brianstorming, resource finding and virtual team building would be fabulously enhanced. Somehow a “Tarte Tartin” and an “apple pie” would be linked and perhaps even a relationship that says that a one it the opposite of the other but with a lid! Wow! now that would be useful. I can now search for “Tony Blair – Clause 4 – great guy” and it would know to associate links to “Tony Blair – started Iraq War – bad guy” articles for example. This is true knowledge mapping.

So, how do we achieve this? It feels sure that progress is going to be made over the coming four years and I would like to fund startups in this space. I don’t think that this is the Cyc common sense engine that is aiming to “boil the ocean” – nor is it the True knowledge “question answering service”.

This is described most interestingly by David as an analogy engine. Now we will start the process of working out whether baisian stats or NLP are the right tool for building this gem.

NOTED: after discussing this with a number of companies and start-ups, I find that people are getting closer to implementing such ideas in arenas such as IP_hubs, the Document_connections, Expert_tracking through social networks etc.. this field is alive and well if early. More connections to follow.

RESOURCES: – Similar

Comments Off on Analogy Engines – semantic concept description

Filed under Ideas

Paul Smith should design the EU’s flag

Has anyone seen the barcoding proposal for the EU flag… well for those of you who admire Paul Smith.. spot the difference.

here's the EU's effort - quite nice

here's the EU's effort - quite nice

and here's Paul Smith's

and here's Paul Smith's

Comments Off on Paul Smith should design the EU’s flag

Filed under Uncategorized

Models or stats, is science really dead

I emerged.. but scathed….. from an ignorance session where Ned Wakeman and I tried to discuss web3-4 “fractal semantic data integration, auto-ontology generation” with too few facts to share between us. it came to this:

To reason over what action to take next, for example which “merchandising offer” shall I place before this customer NOW, do I need a model of the world’s knowledge or do I just need to spot patterns ad-hoc. Today, it was put to me that models are for the birds and that they will never be fresh, complete general enough to be useful. This is not a new view! MIT robotics labs have abandoned some time ago, the attempts to give their robot bugs models of the world, and have instead allowed them to devise, from simple environmentally learned rules, how they should function. Google itself does remarkably well at answering questions without a “model of knowledge” and speech recognition improved remarkably when all the NLP/linguists were fired! If we look at the brain, there seems to be no place where a world model is stored and as we explore our daily lives, we don’t seem to search or querie a central “world model data-set” in our heads. So no models required it seems – just agile stats.

However, True knowledge has made remarkable progress in modeling knowledge and making it useful. We should expect to see that by “blending data” ceverly, we can gain a power law efficacy where 2 facts plus another two facts can be added up to 6 facts. Ubisense is a stunning company selling ultra-fine-grain location tracking systems, with in-building models and fixed UWBand infrastructure, they can build unparalleled systems that would just not be practical with ad-hoc systems. Furthermore, we all know that in signal processing, it’s far far easier to pull a signal from the noise if you have an idea of what you are looking for WCDMA etc etc. Synature has used this to optimise advertising copy in a way that Google could never do by just crunching data raw.

So back to it – as the number of “semantic deals” rises – do we need knowledge structures for the semantic web, ontologies and models? or does the phrase auto-ontology mean something and/or will baysian statistics render our models irrelevant.  Where would you put your efforts and me my money?

As commented on my previous post – it would be nice to be able to see the “geography of this question and knowledge set” but for now… “for/Wired” – “against

Comments Off on Models or stats, is science really dead

Filed under Ideas

Semantic vectors and bounded subjects

I have a request!

Does anyone out there know if it’s possible to capture the bounds of a subject domain, out there on the inter-tubes I mean. For example, if a surveyor came to my house, he could describe it pretty succinctly and appropriately comprehensively but if I do a search on blood coagulation measurement – I get nothing of the kind. “Coagulation measurement devices“… I get 436,000 articles from uncle Google but not a “Read these articles and you have it covered”, not “here are the dimensions/vectors of the subject.. click on these and you’ll have it covered”. Can it be so hard?

With semantic web technologies, ontologies and knowledge bases – can it be so hard to use services like Twine, True Knowledge etc etc to build subject maps.

You could even take me on a tour of the web, showing me where to click and learn! Or if I click on a “Hitler’s a nice chap” article.. perhaps you could show me that “counter view points” exist. Well show them to me!

Shouldn’t creationist articles be a widget-click away from Darwinist articles etc etc.. this seems a rich theme. Anyone wanna discuss…. there’s tech in here, there’s value, interesting UI’s etc. anyone start something in this field?

Comments Off on Semantic vectors and bounded subjects

Filed under Uncategorized

kiki booba


You see it take shape… So guys, which shape is called “Kiki” and which shape is called “Booba”?

Synesthesia is a very cool phenomenon, it’s where people get blended senses, i.e. “see #8 as Green or taste chocolate as red”. This is much more common in artists and “creatives” but is thought to be exhibited in all of us. One piece of the brain – implicated in this phenomena is the angular gyrus which is 8x bigger in higher primates than other species. It’s the cross roads of vision audio and touch and it’s here that very cool things like metaphor are thought to stem from… ps: 98% of people agree that the right shape is called “Kiki” and it’s thought to be due to the audio/visual similarities “synesthesiastic links” that we experience.

Vilayanur Ramachandran: A journey to the center of your mind is a classic Ted Video and sets some this out… but my interest is in its links to our “seeing of patterns” in seemingly unconnected areas.

One theory about entrepreneurs is that they see deep patterns in the world, but perhaps more importantly, they see deep dislocations. They can carry these around with them, seemingly unable to express them to others and in no way clutching at trivial (anyone could do that) solutions. It is then their profound fixing of these profound dislocations that become the great breakthroughs…. We like to hang around with and back these guys. Tough people to deal with but the impact they can have on markets is profound.

It’s also said that creative types see all the problems in the world, and that creativity is the urge to make the world work differently and perhaps better.

For my part – seed investment is SUCH a gloriously risky pass time that I would really like to know how the world would be different and better should a proposed venture succeed. I’d like some deep and profound thought, layered on top of much ambition and a layer of getting it done.

So, to seed investment models – I ask the open question, if the bar is lowered, will more or less profound change and subsequent value be created. Darwin allowed massive parallel experimentation that resulted in wonderful “us”… however, the survival of the fittest is a brutally high bar. Do we expand Y-combinator models or do we demand more and keep the bar higher – and if the latter – how much higher.

Comments Off on kiki booba

Filed under Uncategorized, VC and stuff

Changes to VC #1

Nice post on why VC must change to support a new type of company… by True Ventures, Oleg – thanks for the tip

“….the emergence of extreme capital efficiency from new web infrastructures and evolving software architectures has also been critical to this shift. The bottom line is that the venture industry is out of alignment with the needs of early stage entrepreneurs, and it is early stage entrepreneurs who are vital to pulling the US (and rest of world) out of the current economic situation.”

What would be exciting is to see a bubbling up of more such “extreme capital efficiency from new web ventures”

Our Amadeus Seed Fund has 3 or 4 such ventures all currently in stealth mode, but we’d love to see & do more… they just scale so well.

Comments Off on Changes to VC #1

Filed under Ideas