Apes and bacteria

In a H+ article Tom Barbalet proposes a new (or maybe an old) kind of intelligence metric - survival intelligence. As he defines it, it's a crude estimate of how many humans it would take to stop a given system from working.

"The road system of a given city may require between ten to a hundred human obstructions to shut it down. This is what it takes to cut off major arterial roadways and possibly other minor linking streets. Actually, it may take significantly more human obstructions to fully shut the system down, but these numbers give a good baseline metric of a survival intelligence of between one and two."

That's on a logarithmic scale. I'm not sure that using numbers of humans is a terribly good way of measuring intelligence, but intelligence is such a vaguely defined concept anyhow that you might as well throw it into the menagerie with the rest. Amongst AI aficionados a great deal of time and energy is expended (and wasted) trying to pin down a precise definition for intelligence - a term which covers a wide variety of competences.

Ultimately, most of a creature's skills - at least those which are not merely an accidental byproduct - are in some way directly or indirectly related to its survival. In an Alife simulation you could define intelligence in terms similar to those used by Steve Grand, as the tendency of a system to persist through space and time, whilst resisting perturbations from elsewhere and also adding its own self reinforcing perturbations into the mix.

So, a rock may persist through time for quite a while, but if I smash it in half it doesn't make any attempt to resist this perturbartion by healing itself. Nor do rocks usually make any changes to their immediate surroundings which tend to protect them from externally introduced changes, or cause other similar rocks to be synthesised. A successful system can actively resist change and seek to maintain a homeostasis by adapting itself appropriately.



The trouble with using survival as an intelligence metric is that it challenges the traditional notions of what an intelligent system should look like. So under Barbalet's definition bacteria would have a very high survival intelligence, equivalent to millions of humans. If you imagine the most doom-laden or collapsitarian scenario that you can think of - short of the complete destruction of planet Earth - almost certainly bacteria would survive. Even if every human decided that they were going to eradicate as many bacteria as possible it's unlikely that they would be very successful, and indeed most animals and plants fundamentally rely upon bacteria for their own survival. The goal would be a self defeating one. Most people though would not regard bacteria as very intelligent. In Moravecian or Kurzwilian intelligence scale diagrams bacteria wallow unimpressively somewhere to the bottom left of the graph, but I think this is because the term "intelligence" is usually used somewhat chauvinistically to mean "things which appear to behave or are constructed similarly to myself".

The picture above is taken from The Ape and the Child: A comparative study, by Winthrop Niles Kellogg.

GROK2 composite images test

A test of producing composite images from the pan and tilt mechanism on the GROK2 robot.

Companies which hate your freedom

Nokia and Siemens - helping to disconnect people (then throw them in jail).

Fighting robots repurposed to fight fires

A rather lazy piece of journalism by the BBC highlights the possible uses of robots for fire fighting duties. At first glance this seems like little more than a marketing exercise by QinetiQ, and the reporter doesn't give even a nominal amount of background context.

"Talon is a small and highly manoeuvrable robot that runs on tracks...It has already proven its worth considerably in Iraq, where it has been used for bomb disposal"

Actually it looks as if this robot can do more than inspect roadside bombs.



So here's the background context. The Iraq war is over, at least as far as Britain is concerned, so the number of "bomb disposal" robots which a military contractor like QinetiQ can ship to the UK MoD is going to be reduced. Some of the unemployed robots will no doubt be redeployed to Afghanistan, but I suspect that the rhetoric about substantially increasing the military deployment there isn't going to pan out, because after many years of rather pointless fighting, combined with the gloomy economic scenario, the war chest is empty and the government is likely to be looking for cuts rather than spending increases.

So if you can't sell robots to the military, who can you sell them to?

Fire fighters!

The sales department must be patting themselves on the back for executing such a stroke of marketing genius. In truth, I think there are possible applications in fire fighting, and eventually also in policing. A robot which is of lower mass than a human fire fighter carrying an oxygen cylinder, and equipped with a camera, could move up stairways damaged by fire quickly to look for people trapped in upper floors of a building. That type of application would be something which was genuinely an improvement on the way things happen currently.

Disposing of hot gas cylinders is also a good use for a robot, although this is going to be a very niche type of application probably mainly restricted to industrial sites or rare types of road accident.

Along similar lines - and this might be happening already - I predict that military UAVs will be marketed to police services as a replacement for the traditional (but expensive to run) helicopters. The UAVs will be easy for lapcops to operate remotely from the comfort of their offices, and possibly could spot unusual vehicle behavior autonomously, such as speeding or dangerous driving.

LiveAndroid 0.2

Out of curiosity I tried the Live CD for Google's Android operating system. It's early days I suppose, but this doesn't look anywhere near ready for laptop or netbook usage. An immediate irritation is that there is no mouse cursor, so you're forced to use keys for everything. The interface is extremely simplistic, as is to be expected on a system intended for small mobile devices. There's a large search entry box at the top of the screen, a few icons and a tab to the right side which shows a control panel of settings.

I was unable to get Wifi working - it simply wouldn't let me enter any wifi settings - so this was really a show stopper for a web oriented OS. I guess we'll need to wait for Chrome OS next year to see anything more substantial.

Whilst distro hopping I also tried Fedora 11. Unlike Android, Fedora is a completely usable desktop OS, and due to the Gnome window manager if I wasn't paying attention it could easily be mistaken for Ubuntu. Booting from the Live CD you get Abiword rather than the usual OpenOffice, but asside from occasionally plotting graphs in spreadsheets I'm not a big office software user anyway. Getting wifi working was a single click operation and I was web surfing within seconds. The package manager is different from Ubuntu, but most of the rest looks the same, including the Rhythmbox music player. Is it sufficiently tempting to use as my main OS? Probably not, but this is only because the margin of difference is negligible.

Primordial soup

Tom Barbalet mixes up a simulated chemical cocktail.



This might turn out to be something like Wolfram's automata, except with more spatial dimensions and state transition rules. Assuming that the system is materially closed, most concoctions will reach a static or dynamic equilibrium after a short time, but there will probably also be a few soups which have a more interesting behavior, like rule 30.

Short sighted

Sometimes a technology which looks really good in the short term can turn out to be a liability in the long term. Synthesising oil using bioengineered algae seems like a great idea, if it can be done both technically and in an economically viable manner. The immediate problem is that within a relatively short space of time fossil oil supplies will begin to decline. Since much of human society fundamentally runs on oil this could be a catastrophe if alternative sources of energy aren't used. Producing oil from algae might fix the problem.

Well, suppose that this technology project is entirely successful. What does this mean for the future? It means that you need large expanses of water, exposed to sunlight, which probably means natural or artificial lakes, or areas of the sea. Anyone who has had an algae problem with the pond in their garden might know what comes next, but assume for a moment that proliferation can be controlled in some way and that the process of extracting the oil from the algae does not take a significant amount of energy in itself so that there is a substantial net gain. If you can use oil produced from algae within conventional engines then this means that the same highly inefficient and polluting technology can continue to dominate for decades or centuries more. Essentially the status quo rules supreme, and the environmental consequences continue to mount up.

In a way the problems of unsustainable fossil fuel consumption are a blessing in the long run, since they will eventually mean that we need to radically change our technology, rather than just tinkering about with early 20th century combustion engines.

Sentience on blackfins

An initial test of the stereo correspondence running embedded on the Surveyor SVS. The cameras aren't calibrated, so there is probably room for improvement. Segmentation of the floor might also be useful here.

Unsustainable bubbles

What will the next economic bubble be based upon? The joint winners of the poll are:

Housing (here we go again)
Since houses are typically the most expensive things which people buy during their lifetime this does make sense. Clearly absolutely nothing was learned from the previous housing bubble of the late 1980s, so the pattern seems likely to repeat once memories of the current troubles fade sufficiently. In areas of the world where population density is fairly high, such as the UK, there is a real (not merely hyped up) demand for housing and if you couple this with changes in lifestyle which mean that more people tend to live alone rather than in larger families housing still seems like a safe bet, provided that you buy when prices are reasonably low.

Energy generation
Has traditionally been an unglamourous topic, but a combination of environmental concern and rising costs as fossil fuels become less plentiful could mean that there will be an investment boom in new types of energy generation. It looks like once solar technology falls below a certain price point there might be a sudden rush of demand for roof mounted panels or heaters. If the payback period for a typical solar installation on the roof of an average sized house is less than five years, and the total cost of installation to cover a substantial fraction of roof area is less than $3000 then I think a sales boom could occur. Instead of "a PC on every desktop" the mantra could be "solar on every rooftop".

Gold/Oil
There have been rushes on both of these things in the past. There's no doubt that oil will become more expensive over time, and panics over supply can lead to classic bubble behavior, as occurred in 2008. Gold is only likely to be bubbleworthy either if there is a big discovery of a new source, or if currencies collapse (as frequently prophesied by the collapsitarians) and there's a rush to something which is seen as more likely to retain its value.

The Cloud
It could be argued that the cloud computing bubble is already underway - it's just that the revolution has been put on ice by trouble in other areas of the economy. The cloud seems to be mainly about convenience, since local disk storage capacity - even on small hand held devices - still looks likely to continue rising in the coming years. Having your data in the cloud saves you from needing to manually transfer it from one device to another, and facilitates collaborative activities. Cloud based services looks like a growth area, but whether the growth will be strong enough to result in a bubble I'm not so sure.


The picture is by h.koppdelaney under Creative Commons license.

Pan and tilt

A new simplified procedure for calibrating the pan and tilt mechanism on the GROK2 robot. A coloured object - in this case the green top of a container - is placed on the floor a known distance away from the robot. The robot moves its head whilst observing the position of the target, and knowing the field of view of the cameras allows servo position commands to be mapped to head pose in degrees.



The result for the tilt axes of the forward and backward looking stereo cameras are shown below. Negative numbers on the vertical axis indicate looking towards the ground (-90 would be directly down). There is some difference between the two, and this is because the camera mounts are connected to opposite ends of the control horn on the same servo. You can also see occasional outliers, where something other than the calibration target has been detected within the field of view.



Why bother doing this at all? In an ideal world all the mechanical components of the robot would be built to very fine tolerances using precision machining equipment. In industrial applications this usually is the case, but a hobby robot built by hand using only primitive tools is quite different. The exact behavior of the R/C servos may not be known, and there are other unknowns due to slop in the connecting rods, gearboxes and hinges. On the head of GROK2 each stereo camera tilts on a door hinge, which contains some small amount of slop in its motion. Even with the most careful construction there's no way to avoid these uncertainties. Instead we can characterise the uncertainty as part of the calibration procedure and explicitly include this within the motion model for the robot.

The yellow lines on the graphs indicate the amount of variation around the best fit line, so that the mechanical variation can be quantified. For the tilt axes it's about plus or minus 1.5 degrees, and for the pan axis it's less than half a degree because the head is directly connected to the servo with no interconnecting rods or hinges.

The pan axis looks like this:

Enough helicopters?

How many times can Gordon Brown say the word "helicopter" within two and a half minutes? Let's find out...



Personally I'm less interested in the low level details of the prosecution of the Afghan war, as measured in helicopters, and more interested in why after eight years we're still there and what the whole purpose of the mission is.

The only purpose which I've heard declared by government ministers (a long time ago) is to capture and/or kill Bin Laden and his followers, and to shut down the "training camps" used for terrorism. Surely if these training camps still exist after all this time they can be destroyed from the air, by drones if necessary, and there's no need for year after year of close quarters fighting with local militias. A lot of lives seem to be being wasted, with really very little tangible benefit to show for it.

More tests

Testing the new cameras and servo controller on the GROK2 robot. The wider 12cm stereo baseline results in bigger disparities and a longer effective ranging distance. Although the disparities are bigger this isn't really a problem for the stereo correspondence algorithm, which still works well with negligible noise (bad matches) with disparities up to 40% of the image width. The current stereo correspondence algorithm being used is also significantly faster than all previous versions, using exclusively integer maths and bitwise feature descriptor matching.



Views from both cameras during calibration can be seen here. Things in the far distance should not appear to move, whereas nearer objects move more.

Heart of the sun

The negative effects of memes

As far as I know this experiment hasn't been done. My guess would be that the result would be the opposite, in that there would be no significant neurological difference between the devout cultist and other people. Most people when they're in their late teens or early 20s, if they're smart, are looking for something to believe in or something worthwhile to do with the rest of their lives (if they're not so smart they just don't care). I think that whether you become a cultist or not depends upon your immediate social environment, particularly during an early adulthood "sensitive period", which can be governed to a large extent by chance interactions with strangers.

I expect that from a neurological standpoint there is little difference between a cultist or a high ranking member of a mainstream political party or an enthusiastic fan of a pop group or a singularitarian "believer". To me these things look qualitatively like the same sort of phenomena.

Those terror threat levels in full

In my view these terror threat levels are actually pointless, and simply used in a divisive way when politicians want to pass what would otherwise be unpopular legislation. They need to reduce the threat level in order to be able to increase it again some time in the near future. If it simply stayed at the top all the time it would be politically useless.

For convenience the official threat levels are listed below.

• unnewsworthy
• moderate
• politically expedient
• real and serious
• unreal and laughably implausible
• highly likely
• likely high
• higher! higher!
• a safe bet
• substantial
  
• severe
• severely substantial
  
• ZOMG! INCOMING!

What to expect during a (real) pandemic

This article caused a bit of grim humour. I don't believe that the current swine flu flap poses a very serious threat to public health, since the figures on mortality rates appear to indicate that it's not significantly more deadly than ordinary human flu. But if there were to be a much more harmful pandemic similar to avian flu, with a 50% or above mortality rate, the BBC article gives an indication of the sort of help that you can expect from the government.

A helpline, or "FluLine".

Of course, in the event of a real national emergency you would never be able to get through to the help line, and likely would be held indefinitely in a queue at premium rate listening to Vivaldi or Queen. Even if you could get through the advice you'd receive would likely be very unhelpful, and probably be a menu driven pre-recorded message so that there's no possibility of dialogue.

People like to believe that the government is all-powerful and that it would save them from peril in their hour of need - like the cavalry in old western films. I very much doubt that this would be the case in reality. The best that you could possibly hope for I think would be a vaccine issued maybe 6 months or more after the outbreak begins. In the intervening time people would simply be left to their own devices, which in a serious pandemic would be not much more than the law of the jungle.

Small wind turbines not so powerful

I've been reading stuff about wind turbines, since I might try making one as a fun project. These things are extremely simple - essentially 19th century technology - and can be built from scratch at negligible cost using scrap or low cost components. If the collapsitarians are right and the end of the world as we know it is nigh (I don't expect that it is), knowing how to build a wind turbine might be quite handy.

However, the amount of energy which even the commercial small wind turbines can generate in a typical urban area may be severely over-hyped, according to this assesment.

"Bottom line is about £10 worth of energy per year on this site."

So even if the cost of electricity continues to inflate, as it has done in the last couple of years, it looks as if many of the commercial roof mounted wind turbines are never going to even manage to pay for themselves on any sensible time scale, and likely will suffer mechanical failures before achieving payback.

In the long run it looks as if solar energy is going to be more practical for most locations, provided that the cost of manufacture per watt continues to reduce.

As far as wind energy goes, this site seems to be one of the best with lots of useful information.

Robots that walk

An interview with Jaimie Mantzel. This guy is like a one man industry, and he's built various walking machines in the past.

His current Giant Robot Project is basically a bigger version of the small hexapod shown below, but with enough room on top to carry a passenger/driver.

More reprapping

A talk about the Makerbot, which is a commercial version of the RepRap. I liked the idea of making a similar machine for 3D scanning, using a laser pointer/line and a webcam. Even cooler would be to add the scanning ability to the RepRap or Makerbot, so that scanning and printing functionality are combined into a single machine. It probably wouldn't be very difficult to do, and the camera/laser could be mounted alongside the plastruder head, perhaps with a removable cover over the camera lens to protect it from contamination when it's not being used.

I can think of some situations where a machine like this would be quite useful. So for example the camera mountings for the GROK2 robot are just hacked together from wood, but with a RepRap type machine I could make a much more accurate and professional-looking camera mount, and maybe do the same for the pan and tilt mechanism too.



I think what these guys are doing is just the very early stages of what will probably be a bigger industry in future. Machines of this kind won't be able to compete with existing mass production methods, but they will be able to compete in markets were low production volume but high end user customisation are required. As mentioned at the end of the talk, frames for spectacles are a perfect example because each wearer is different. It doesn't take a big leap of imagination to envisage setting up a business where you have a web site which allows users to design their own spectacle frames, then have a RepRap machine print them out and ship them by post. The same could be done for chess pieces or any similar small objects.

The integrated circuit

ICs are something that we take completely for granted. In fact, it's hard to imagine a time before they were invented...

Germany tries to ban computer games

Bizarre, but apparently true. Germany wants to ban video games which contain violence. Doesn't that mean just about all computer games, apart from maybe Pong and iboobs?

There has always been talk about the harmful effects of computer games ever since I was a kid, and most of the arguments seem to me to be nonsense. Certainly some of the early games, such as Space Invaders, Apple Panic or Skate or Die were pretty violent, and even in the 8 bit era there were calls from a prudish minority for games to be banned.

Does this mean that the playing of Chuckie Egg should be banned in Germany? There are no laser guns or explosions, but the player's rotund hat wearing character can be instantly killed by large blue anatidae, or chased mercilessly on the final level by a gigantic menacing yellow duck.

New cameras ready for robotic excursions

All four new cameras are now fitted to the pan and tilt head of GROK2, and the robot is once again ready to run on Linux. The new cameras are Quickcam Pro 9000's, with all the casing and microphones removed.

On the arbitrary nature of memorials

The BBC web site has an article on a memorial to the people who were killed in the 7/7 bombings of 2005. It doesn't look like a very imaginative sculpture - just some poles stuck in the ground. It doesn't even seem to represent "52 unique individuals" as claimed, since all the poles are the same.

The way that we choose to memorialise certain things and not others seems very arbitrary to me. So for example a similar number of people are killed in traffic accidents every day, or die whilst waiting to receive hostpital treatment, but these people don't usually have statues - artistically interesting, or otherwise - erected in their memory.



So what seems to matter as far as memorials are concerned is not just the number of people killed, nor whether they were "innocent" or "non-political", but whether the event in which they died was itself of a highly political nature.

Needless to say, if I were unfortunate enough to be killed in an event resembling 7/7 I would't wish to be represented as a featureless grey monolith.

Four years on from 7/7 there still remain some inadequetely explained components to the story, and the lack of a public enquiry into exactly what happened that day has meant that various possible alternative explanations, questions about what the security services knew about the bombers and miscellaneous conspirarcy theories have arrisen. In hindsight I think that not holding a public enquiry was a mistake, and that probably the best policy when something like this happens would be to open source the evidence as soon as it becomes available, so that the wilder theories can be clearly ruled out.

GROK2 gets new cameras

I've been searching for possible replacement cameras for GROK2 for the last couple of years, and it's been quite tricky because the image quality of the Creative Webcam NX Ultra is quite hard to beat. Had it not been for recent kernel changes which meant that the existing cameras no longer worked on Linux I probably would have just carried on using them.

The new cameras are Quickcam Pro 9000's. These have an unusually elongated circuit board, so one of the cameras is mounted upside down with the image being inverted in software. The mounting holes on the circuit boards are asymmetrical, hence the rather skewed appearance of the wooden bar that they're fixed onto when one board is upside down.

The new appearance is nerdier and more Cog-like, and the protective hood will be reattached once I've replaced both sets of cameras.



Until recently I'd been thinking of using the Minoru stereo webcams as replacements, but after experimenting with these, including changing the lenses for wide angle versions, it seems that the 6cm baseline is just too short to get very good depth perception beyond a metre or so. The Minoru might still be used though on the Rodney robot.

I've increased the separation between the new cameras to 12cm, in order to get a little more ranging performance. For this robot close up recognition isn't the main research goal.



The new cameras are considerably smaller, and can be tilted almost vertically down towards the floor. They also have a wider exposure range, which means that in theory the robot could be used outdoors, although at present I'm not intending to do that.

Battle of the browsers

Interesting to see that in different countries there's a huge difference in the search volume for IE and firefox. So in Ireland, South Korea and to a lesser extent in China there's a larger number of searchers for IE, but in India, Russia, Europe and the far east more people are searching for Firefox. Notice that the pattern of searches in Hong Kong are opposite to those in the rest of China.

Sign of the times

Three quarters of your pay is belonging to us

This kind of thing seems to be becoming quite common. I expect that for most employees on ordinary wage levels a pay cut of 75% would simply be impractical, because the cost of living is not falling at a similarly rapid rate - in fact retail prices seem to be continuing to rise (aka stagflation).

A month or so ago, when I was still full time employed myself, there were many confident-sounding predictions that the economy would begin to turn around in the second half of 2009. Whether or not that will actually happen remains to be seen, but recessions don't last indefinitely so sooner or later things will begin to pick back up again.

Wiggle it

More progress on the Willow Garage robot. Plugging yourself in is actually quite a hard manipulation task, which is sometimes even challenging for humans. I notice that in the video the wall socket cover is cracked, so maybe the robot bashed it too hard during development.



There would be easier ways to do this, such as having a special recharging station, as used by Roombas and other mobile robots. However, being able to use regular mains sockets would be a significant advantage in terms of maximising the ease of deployment, since such sockets usually exist in every room and require no special setup.

Since I have far more free time at present than would usually be the case I may investigate the possibility of using their Robot Operating System. I have my own system, written in C#, which does substantially the same sort of message passing between processes, but if it looks like there's something out there which is reasonably easy to use and has an existing development community, it might be a better long term strategy to combine efforts.

Maybe there ought to be a scientific term for "wiggling". You could call it "off-axis rotational re-alignment" or "closed loop manipulator pose optimisation".

Swine flu: avoid sensationalism and check the figures

It seems that there's still a lot of media hype going on with regard to swine flu, with claims being made that it's "unstoppable" and out of control. As I've mentioned previously the main thing that you need to focus on isn't how many people are contracting the illness but how many actually die as a direct consequence.

The Wikipedia entry for 2009 flu pandemic gives a set of figures which I havn't seen reported in any of the mainstream media articles (which seem to be mostly only succeeding in promoting fear and hysteria). Assuming that these figures are reasonably accurate the mortality rate worldwide is around 0.5%, or 1 in 200, with the numbers varying from one country to the next (in the UK it's only 0.05%). This is higher than so-called "seasonal" flu, but not by a margin great enough to be convincingly beyond measurement error, since there will of course be many people who become sick and either don't know that it's swine flu, are never tested, or choose not to report it and subsequently recover. Realistically if I were to get flu-like symptoms and am not specifically tested for H1N1, how am I going to know the difference? I'm probably not.

Compare this with avian flu, and there's a gigantic difference in mortality. In the case of H5N1 over half of those people who contracted it died within a few weeks.

So I'd say that swine flu is only something worth worrying about if it mutates into a form with a significantly higher mortality rate. Thus far those people who have been victims seem to have largely fallen into the traditional categories of those at the extreme ends of the age range, or who have preexisting medical conditions which compromise their immunity.

Although it's claimed that 40 people per day could die from swine flu by the end of August, to put this in context - something which mainstream media reporters never do - from the latest figures I could find (2002) on the National Statistics website about 100 people are killed or seriously injured every day in road accidents in the UK - almost equivalent to two 7/7 terrorist attacks (which really makes me wonder about the cost/benefit of all the anti terror measures taken in the last five years).

A seriously hard object tracking task

As you can see the human brain is far better at detecting and tracking separate objects than computers currently are.

Right next to the Beluga

More singularitarians spotted, right next to the Beluga.