"The age of television is ending"

Even the BBC now admits what I've been saying probably for the last decade. The age of television is drawing to a close. This doesn't mean that TV will disappear entirely but, like radio before it, it will merely become less important as a source of entertainment and news.

In my opinion the BBC has no particularly "unique status", and the way it's funded is archaic and makes little sense in the modern era of subscription services. Certainly the type of "public service broadcasting" which existed when I was growing up, where you had very large audiences watching one or two TV channels, is now very much dead and not coming back under any foreseeable circumstances.

Whilst one media format declines others are ascendant. Video games and general internet usage have greatly increased in popularity, but the real phase shift is going to come once augmented reality becomes a consumer commodity. Augmented reality entertainment, delivered by wearing a device similar to the Eyetap with a built in gyro/accelerometer so that it knows your head orientation relative to gravity, I think is going to be a gigantic industry producing experiences which are highly compelling and where the style of interaction is very natural.

Lenses

Here are the M12 lenses which I've tried using so far with the Minoru webcam. The 3.6mm lens fits, but gives a field of view smaller than the original lens. The 2.1mm lens does give a significantly wider field of view, but also seems to have a focus longer than the height of the lens holder above the circuit board. To be able to use the 2.1mm lens I would need to either use a different lens mounting, or some sort of extension. The trouble with trying to make an extension is that without precision machining equipment, armed primarily only as a pre-industrial craftsman with a drill and junior hacksaw, it would be hard to make something with enough accuracy to get good alignment - and good alignment is very important for stereo vision in order to ensure that the epipolar constraint applies.

I've ordered yet another set of wide angle lenses with only 5mm "back focus". Presumably "back focus length" means the distance between the back of the lens and the CCD/CMOS chip. Unfortunately without detailed drawings, which are almost never available, it's hard to know in advance whether a given lens type will be suitable, so it's mostly just down to trial and error. Fortunately though plastic camera lenses intended for webcams or security cameras are quite cheap, so if they turn out not to be appropriate it's no great financial loss.

In this picture you can see the infrared filter, which is normally a square piece of plastic superglued to the back of the lens.

Dear carbon mass...

This made me laugh.

World builder

World Builder from Bruce Branit on Vimeo.

Image processing mistakes as an artform

I've entitled this one The Struggle for Agrarian Self-Actualization.




And this one is entitled Noospheric Disenfranchisement Al Dente.

Blair was aware of torture policy

A story which seems to be all over the blogosphere, but doesn't appear to have been covered in the mainstream news outside of the original Guardian article, is that evidence has emerged that Tony Blair was aware that people were being tortured as part of his counter-terrorism policy. Allegedly the torture victims even included British citizens.

I expect that the new Iraq inquiry won't cover this, or if it does all the relevant documents will be withheld from public scrutiny in the name of "national security", or wotnot.

Probably the rebuttal that Blair did not "authorise" torture is (we hope) true, but being aware that it was going on under circumstances which could reasonably be described as being within his influence or control, and yet choosing not to intervene, would put him into territory of questionable legality. So it's not surprising that he feared a "show trial" if the Iraq inquiry were to be held in public.

Monodevelop goes full circle

It now seems that there is a version of Monodevelop for Windows. This is ironic considering that MonoDevelop originally came from Sharpdevelop - an open source Windows based alternative to Visual Studio.

I suppose it's nice that there is a Windows version, since this encourages people away from the proprietary software. Whether there is really a significant demand for GTK/Mono applications on Windows though I think is questionable.

I've been using Mono and Monodevelop for over two years now, and when I started out it was far less controversial than it is today. From a technology point of view I think C# is a pretty good language, comparable to C++ or Java, but I do agree that there might be patent issues further down the line, especially since the implementation of Mono seems to significantly exceed the ECMA C# specification with extensions such as Windows.Forms. Fortunately though I don't expect my own projects to have a very large user base, and even if Microsoft were to do their worst, porting the code to another language would certainly be feasible. What's important as far as I'm concerned are the algorithms, not the specific language that they're implemented in.

Replacing webcam lenses

My initial attempt at replacing the lenses on the Minoru webcam appears to be a complete FAIL. I've tried using a 90 degree FOV 3.6mm "camera board" lens, which fits perfectly well, but the results look like the following.



On the left is the original lens, on the right is the 90 degree FOV lens. As you can see the field of view on the right actually looks narrower than the original. The difference in colour is due to the new lens having no infrared filter attached. At this point I'm not really very concerned about not having an IR filter - in fact it may be an advantage in low lighting conditions.

I think the problem here may be the chip size. The "90 degree" figure may be assuming a larger CCD/CMOS sensor size than exists on the webcam.

Stereo vision with the Minoru webcam

I now have a version of software suitable for use with the Minoru webcam. It's similar to earlier versions of sparse stereo correspondence, except that I've tried to optimize it for speed using binary descriptors for each detected image feature which can be compared very rapidly. Also the matching routine does not assume that it has access to both images at the same time - only sets of features and their descriptors. This means that it can be run in a relatively decentralised way which I can implement embedded onto the Surveyor SVS, where there is some limited amount of bandwidth between two otherwise independent camera boards.

Testing using the Minoru at 320x240 resolution this stereo correspondence algorithm is completely real time, running at frame rate speeds.

There are a couple of improvements upon previous sparse stereo implementations. I'm using an eigenimage-like method to compare the feature descriptors, which gives better discrimination between features which have many bits in common, and I'm also using both correlation and anti-correlation, as described by Read and Cumming. Good matches tend to correspond to local minima in correlation space and local maxima in anti-correlation space. If the maxima and minima don't occur together, it's probably a bad match.

Fun with electronics

Although I do have a certain amount of electronics piled up in the vicinity I'm mostly a software engineer. However, when I was at school I had a friend who was just like the guys in this video and was always coming up with wacky electrical and/or computer projects. In 1983 he actually taught me some aspects of BBC BASIC during a series of boring history lessons. I think some time in the mid to late 1980s I sold my Sinclair ZX-80 to him and he used it to control some fancy disco-style LED lighting.

The Legend of Neil Robot Part 1 from Jeri Ellsworth on Vimeo.

Owen Rubin Shows "Kermit" Ataris Prototype Robot Pet from Jeri Ellsworth on Vimeo.

By the standards of the late 1970s the "Kermit" robot would have been ultra sophisticated, and pre-dates the MIT mobots of Brooks & co. You could argue that not much progress occurred in mobile robotics for at least another decade.

Decoy

There has been such a huge media flap over the last month or so regarding politician's expense claims that this makes me suspicious that perhaps other more interesting stories might be being "buried" (to use the New Labour terminology).

One thing which has been underway is "quantitative easing", which as far as I can discern involves stealing massive amounts of cash from tax payers.

"The Bank of England started printing money in March and has now pumped an extra £75 billion into the economy to encourage lending, boost spending and stimulate growth."

Presumably eventually this spending splurge is going to be underwritten with tax money, and we can expect taxes to rise significantly over the next few years. Now I'm not trying to defend the politicians. What they did with their expenses was certainly dishonest and unethical, and in some cases possibly also illegal, but the amounts of money involved are dwarfed by the utter waste of quantitative easing and other questionable policies such as the national identity card. You may also wish to peruse some interesting economic statistics.

"The Labour government has thrown all of the fiscal rules out of the window and has embarked on a debt busting inflationary programme of money printing. The Bank of England having first been paralysed into inaction despite overwhelming evidence that interest rates should be cut in an orderly manner towards a target of 3%. Having acted too late have now slashed interest rates from 5% to 2% in less than 3 months. With expectation of further cuts to take interest rates to the target rate of 1% and highly likely lower still. All of these actions reek of nothing short of panic to prevent an 1930's style deflationary depression. The consequences of which will be at the very least stagflation and if the level of incompetence continues then highly likely hyper inflation along the lines of the 1970's starting during 2010 and continuing for many years thereafter."

Strange resurgence of internet utopianism

It's as if the last 15 years of digital history never happened, as Gordo preaches a strange form of internet utopianism, saying that "You cannot have Rwanda again".

Due to what? Twitter?

As the internet and the world wide web were emerging into the collective public consciousness in the early 1990s there was a lot of talk about it having a radical democratising effect upon society. Some people thought that it would render nation states irrelevant, and that we would all become part of one big, happy, digitised global village. Others envisaged a sort of digital anarcho-capitalism in which conventional politics and politicians would be obsolete and where citizens could vote electronically in real time using computers or mobile phones to make collective decisions without the need for elected representatives or houses of parliament.

Unfortunately it didn't quite turn out that way, although it does appear that the internet has fostered greater communication between people in diverse areas of the world who would otherwise have never been aware of each other's existence.

I would like to believe that Twitter could prevent genocide, but sadly I don't think it can. What Gordo doesn't mention is that technology cuts both ways, and that were a genocide like the one which occurred in Rwanda to occur again internet communication could make it all the more rapid, efficient and coordinated (from the murderer's point of view). The assumption which is being made is that lack of information allows these things to happen. To some extent that's true, but I remember that at the time the Rwandan genocide was actually covered by the media and that people around the world were to a significant extent aware of what was going on. What was lacking was any action at the political level to intervene and prevent people from being slaughtered.

The question of how to prevent future genocides is a difficult one, which I think doesn't have an easy technological fix because it originates from deep seated weaknesses within human psychology - it's more of a firmware bug than a software one.

There are some reasons to be optimistic though. I think it's almost certain that new genocides will occur in future, and that they will take place within the context of the modern digital communications infrastructure. This opens up possibility of doing data mining - a sort of memetic archeology - to discover exactly how certain ideas are transmitted between people and characterise the trajectory of an idea as it undergoes positive feedback spreading across multiple minds. As memetics becomes less of a hand-waving exercise for philosophers and more of a science it may be possible to identify the spread of harmful ideologies and either stop them before they manage to take hold or apply a perturbation which causes the meme to mutate into a less harmful form. From the vantage point of the present it's very unclear how this would be practically achieved, but in principle it might be something which becomes possible.

Robots for small and medium sized businesses

It's to be expected that as conventional industrial robotics becomes cheaper - mainly due to falling costs of the servo drives and software - that the sorts of automation previously only used in large factories will also become available to smaller businesses.

Security related vision system applications: a policy statement

Whilst looking for a potential new job in the last week or so I've been contacted by several organizations (in one case repeatedly by the same company) in relation to security related applications of computer vision. The salaries quoted seem unusually high for a software engineer (software engineers are usually somewhere marginally above cleaners and receptionists in the salary food chain), but upon examining the web sites of the companies involved it seems extremely unclear as to what they do or what products they actually sell.

After some consideration, and putting the tin foil hat to one side for a moment, it appears to me that these companies are part of what's known as the "black economy" or "security/industrial complex" - that is, a secret largely government funded part of the economy which is neither known about nor regulated by the general public.

If I knew more about the products being developed I might be more sympathetic to the idea of joining such an organisation. If for example a company wants to use computer vision to check which cars are in their carpark I'd have no particular objections to doing that, but I'm very keen not to get involved with anything which might bring me into disrepute, or be regarded by myself or by my peers as being an unethical use of technology. Also, I'd regard any restrictions upon my ability to talk freely about technical matters - above and beyond the normal level of business confidentiality which is to be expected - as completely unacceptable.

So I wouldn't be well suited to doing secret work for secret organisations, since at heart I'm an open source developer and I'm really useless at keeping secrets. If someone asks me a question - especially a technical question - I generally tend to try to answer it as honestly as possible. Also I have very little nationalist instinct, so appealing to notions of national identity absolutely doesn't work with me, and indeed most of the development which I've done in the past has been very international in nature.

Another sham inquiry

With the Iraq war now more or less over, at least as far as Britain is concerned, it looks like there will be yet another sham inquiry to examine the whole affair - or at least go through the motions of pretending to examine it.

"It will start next month and take at least a year, Mr Brown said. It will not aim to 'apportion blame', he added."

Almost certainly this will follow the same pattern as the earlier Hutton and Butler inquiries, where the role of politicians will not be permitted to be scrutinized in any depth and nobody will really be held to account for what must probably rank as the most cynical war fought within my lifetime.

My predictions for this inquiry are as follows:

• None of the critical information will be examined, such as the role of Tony Blair and his closest advisors circa 2002/3 in the decision to go to war.

• The objections of the intelligence services during the preparation of the 2002 dossier "Iraq's weapons of Mass Destruction" will likely be either ignored or omitted due to "national security" reasons. The role of David Kelly, who tried to alert the media (via a BBC reporter) to the problem of evidence being distorted or fabricated, I think will be skipped over "because this matter was addressed in a previous inquiry".

• There will either be no investigation, or only a cursory mention, of the "dodgy dossier" because this would make certain civil service departments look either foolish or dishonest.

• There will be no examination of any correspondence between Tony Blair and George Bush in the months preceding the outbreak of war.

• The inquiry will focus primarily on relatively low level details of the prosecution of the war, such as the issuing of soldier's equipment (bullet proof vests, boots, etc) and whether or not a sufficient number of helicopters were available to evacuate casualties. These aspects tend to obsess the popular press, because they involve stories of individual tragedy or heroism, and provide a substantial distraction from the higher level decision making at the political level.

Straightforward tyranny

It's hard to believe that something like this is happening in Britain in the 21st century, whereby someone can be held under indefinite house arrest with no proper legal process. If it's believed that this man is, or was, a terrorist he should be charged and have his day in court with the evidence presented against him. There are already quite adequate laws in place for dealing with people plotting terrorist acts, such as "conspiracy to murder" or "belonging to a banned organization".

Pleo: the rise and fall of a companion robot

I quite enjoyed watching a video in which John Sosoka - Ugobe's former chief technology officer - talks about the kinds of problems which were encountered in the making and selling of Pleo. I'm interested in knowing what it takes to build a robot startup company, and how and why these kinds of businesses fail. In the not too distant future I expect robotics startups to become a more common phenomena, both in the toy industry and in other kinds of industry.

The biggest and most obvious reason for failure is just that the economy tanked. In an economic climate where even ordinarily very profitable "blue chip" companies (like the one I was working for until last week) are struggling those making luxury items which depend upon consumers having disposable income are going to be amongst the first to get into difficulty.

Lack of focus is also criticised. That is, not having any specific target market in mind then optimising the product for that particular market. But in the case of something which is really intended to be a "companion robot" I think that lack of focus was probably exactly the right thing to do.

One lesson seems to be that trying to scale up sales too quickly can have negative effects. A better strategy might be to initially sell limited quantities to a small group of hardcore enthusiasts, then let those customers do some of the marketing for you virally, through blogs and videos. The viral marketing route would help to reduce costs. Slow but steady growth appears to be a better business model than trying to scale exponentially.

Another thing which I think Ugobe did right was to try to make the robot into an open platform for developers, because as I well know acquiring any moderately sophisticated robot suitable to use as a software development platform for only a few hundred dollars remains really hard - although perhaps some of the small humanoids would fall into this category.

I expect that a decade from now toys with Pleo's level of electromechanical sophistication will be common and cheap to buy. Probably Ugobe just suffered from unfortunate timing - releasing a luxury product into a market where people are cutting back on, or cutting out, luxury purchases.

World oil production peaks, maybe

It's probably still too early to say whether this is really the peak, or merely a temporary plateau, but some commentators are suggesting that world oil production peaked somewhere between 2005 and 2008.

"The US Energy Information Administration (EIA) and the International Energy Agency (IEA) should make official statements about declining world oil production now to renew the focus on oil conservation and alternative renewable energy sources."

Even if this turns out to be a false alarm it would still be wise to invest now in infrastructure and technology projects preparing for a post-petroleum economy. I think the official statement from the government, given a while back in response to a petition asking for greater preparation measures to be taken in anticipation of peak oil, is that production will continue to meet rising demand for another 20 or 30 years.

The only thing that we do know for sure is that oil is a finite non-renewable resource which will begin to run out sooner or later, and that our present society is highly dependent upon this substance. Just like an addict hooked on a narcotic, we either need to lose our dependency, or lose our current standards of living. Unfortunately the transition does depend to some extent upon enlightened political leadership. Individual citizens can go some way towards reducing their usage, but to really move away from an oil-based economy reorganization needs to take place at a high level. What does this mean in practical, non hand-wavy terms:

• The big automotive manufacturers need to reduce their output of conventional internal combustion engines and substantially increase their output of electric and hydrogen based vehicles. This means doing far more than simply parading a few concept cars at trade shows.

• Governments need to work together with the automotive industry to ensure that there is an infrastructure in place to support electric vehicles - regular recharging locations in carparks and streets. Also making public transport more convenient and more fashionable will help.
  

• There needs to be continued and increasing levels of investment in solar and battery technology research and development. Potentially solar could provide a large percentage of our energy needs, provided that the cost per watt is right. Look at all the unused rooftops in a typical city, then do the math. Once cheap solar is available consumerism and the normal operation of market forces will do the rest. In the last few years there have been quite a number of encouraging sounding developments, but these need to be transferred from the lab bench to commercial production if they're to have any impact, and governments can assist in this regard with subsidies or more lenient regulations and tax rates for certain kinds of enterprise.
  

• As a fallback measure probably more nuclear fission based power stations need to be built. This is already happening to some extent. Nuclear fission is dirty and dangerous - it's a high risk and environmentally unsound strategy - but we may be able to use it temporarily to buy a few more decades in which to develop replacement technologies, such as fusion reactors or better renewables (e.g. tidal power).

Tim Tyler critiques "the singularity"

Tim Tyler suggests that a technological singularity is poorly defined, inconsistent with existing terminology and that our prediction horizon will advance at a pace commensurate with improvements in computational performance.



I'm inclined to agree with the notion of the "singularity folk" being salespeople, and you can see this clearly manifested within the recently founded Singularity University which seems aimed primarily at relieving wealthy technocrats of their burdensome cash.

However, I find Tyler's arguments about the ill defined nature of "human level" intelligence not entirely convincing. It is true that there exist variations in cognitive performance between people - at least as measured by certain types of test - but the standard deviation in human intelligence could turn out to be small compared to the rate at which machine intelligence advances, such that the rising tide analogy used by Moravec and others could turn out to be more of a rapid and overwhelming tsunami. Just because the waters have been rising slowly thus far doesn't imply that this will remain the case in future.

Also, whether the term "singularity", "revolution" or "takeover" are used seems like semantic quibbling, although a word such as "revolution" would be more consistent with descriptions of other historical periods of significant qualitative change.

Remembering Tiananmen 20 years on

Everyone of my generation remembers "the tank man". I was a teenager at the time of the Tiananmen Square massacre, and for a while afterwards I kept a newspaper photo of the lone individual standing in front of the tank pinned up on my bedroom door.



In the late 1980s tyrannical communist regimes were in what seemed like a spiral dive towards oblivion, and the terrible crimes which they perpetrated against their own citizenries were being exposed for the rest of the world to see (for example in Romania and east Germany). At the time when students began protesting in Tiananmen Square it seemed as if China would be the next in the series of authoritarian dominoes to be toppled.

The tank man has a kind of iconic status. He represents competing priorities at the individual and collective levels of human organisation, but he also shows that people do not necessarily always follow orders given to them if these violate common standards of decency. The tank driver could have simply continued his mission and easily driven straight over the protester, but he chose not to.

The lesson of 20 years ago is there is nothing inevitable about the process of democratisation, and that tyrants can very effectively push back and crush those who oppose them.

Redundant!

Today I've been officially declared redundant, after a week or so of perfunctory and ultimately futile negotiations. By next week I will have been assimilated into the borg collective that is the two million or more unemployed. Given the current economic situation I think it's quite possible that I could be unemployed for a significant length of time.

So by next week I'll transition from being cash rich and time poor to time rich and cash poor. What to do with all this time surplus?

• Get the robot into a condition where it can be easily demonstrated

• Take some videos of the robot in operation and make a web site for it

• Create some stereo vision/odometry data sets using the robot. Standard data sets taken under realistic observing conditions are something which are really needed to turn robotics into more of a science than it currently is
  

• I'm a big fan of Green Power Science. I might try some low cost energy generation projects

• Possibly I might consider contributing to OpenCog - particularly connecting it to robotics simulators.

Another dubious industry report

I agree with this report that robotics, biotech and nanotechnology will be growth areas, but the idea that in a "knowledge economy" people need to cluster within "supercities" seems like nonsense to me. With high speed internet human knowledge can be piped from anywhere, so there is really no need to physically aggregate, unless the industries involved are primarily producing physical products. Improvements in robotics will also mean that telerobotic working becomes practical, further reducing the need for human workers to be grouped at the same location.

The stuff about Londoners moving North to avoid high living costs is an old chestnut which I've heard many times before. It sounds plausible, but I'll believe it when I actually see it happening.