I've photographed and uploaded a first world war plane spotters handbook to archive.org. This is one of the more unusual and presumably rare books in my personal library. It was used to assist pilots in identifying friendly and enemy aircraft and airships. Considering its age and wartime usage it's in quite good condition, and shows all the main aircraft which were in service in 1916.
It includes useful advice, such as:
"Troops should on no account fire at an Aeroplane unless the German black cross is distinctly visible."
Passed on to F/Sgt K.H.W. Carroll D.F.M. 454 Squadron B Flight Egypt, Palestine, Transjordan, Syria, Iraq, Libya, North Africa, Cyprus, Sicily, Italy
It would be interesting, if possible, to find out who these men were, and whether they survived the war. They belonged to the Royal Flying Corps, which towards the end of the war was amalgamated into the RAF. The Distinguished Flying Medal acronym after the second fellow's name suggests that he was a decorated war hero. According to Wikipedia only 105 of these medals were awarded during the first world war.
Another article about use of telerobots in contemporary warfare makes an interesting point:
"would the enemy be forced to look for ways to hit back at the operators located in their home countries? Considering that the operators are running their robots from within buildings and facilities around populated areas, wouldn't that classify them as hiding behind human shields and subjecting their own civilian populations to retaliation for a war they are remotely fighting on the other side of the globe?"
It's safe to assume that future military telesupervisors will be buried deep underground in bunkers which are inaccessible even to the so-called "bunker busting" bombs. If minimisation of civilian casualties is the main aim then you might imagine that these bunkers will be constructed in the middle of unpopulated areas. But there's a tension here between administrative convenience and protection of civilians. Telesupervisors need to be able to commute to their bunker and back again on a daily, or otherwise quite frequent basis, so the bunkers probably can't be located too far away from cities. The opposite strategy, as the article suggests, would be to use cities as a sort of human shield, with telesupervisory bunkers situated directly beneath them. Presumably attacking cities with semi-autonomous or autonomous robots will be somewhat politically unacceptable - at least initially - since this would resemble something close to a real terminator scenario even if negligible artificial intelligence is involved and the machines are just acting like agents within a real time strategy game.
Having robots just fight other robots does seem fairly pointless. You might as well have the leaders of rival nations play a game of chess in order to decide who is the winner. So the primary objectives in future wars will probably be not only to neutralise incoming robots but also to cut off or jam the line of communications between the bunkered telesupervisors and the robotic armies which they administrate, with the civilian population sandwiched somewhere in the middle.
Jaron Lanier is on a book tour, promoting his newly completed manifesto. Lanier seems to fall into a similar category as Joseph Weizenbaum, as a kind of digital refusenick. Like Weizenbaum, he's a smart fellow schooled in the ways of all things computational who came to reject some of the most sacred offerings of the modern computer driven age.
A few of Lanier's views I'm sympathetic towards. For example, he thinks of singularitarianism as a kind of absurd pseudo-religious cult. But most of what I've read of his views I'd disagree with. Perhaps he is being deliberately contrarian. After all, controversial views are a far better basis for shifting copies of books and appearing in news articles than ones which are closer to the mainstream.
In this Guardian article he focuses most of his criticism on free culture, which he describes as a form of "digital maoism" which forces knowledge workers to become peasant labourers. The quandary over how people producing creative works should receive financial compensation is genuinely a difficult one, which has yet to be fully resolved.
"In You Are Not A Gadget I propose five different approaches to a solution. The one I am the most hopeful about is to return to the very first vision of the web: a universal micro-payment system. For practical purposes, that would mean that there's only one copy of a creative thing, and you pay a half penny every time you access it."
On the face of it a micropayment system might be quite a good thing. People such as Richard Stallman have also promoted similar ideas as a way of supporting artists. But of course there is a dark side to the above comment. If you follow through the logic what it implies is a world of lockdown and social exclusion. Imagine a world in which all computers are iPads, controlled by a single corporation with restrictions to prevent the duplication of any information without official approval by a central authority. To people in rich western countries making tiny payments might be no big deal, provided that the technical implementation was sufficiently slick. But even a micropayment system of this sort means that a large segment of the world's population would be restrained behind a giant digital paywall, which stands as an economic barrier to information, entertainment and news. You may have solved the problem of how artists get paid, but at the price of creating artificial scarcity or information poverty, and a high degree of centralized control. I'm old enough to remember what the world of information poverty was like, and it was not some grand panacea, although a few rock stars did manage to live a life of affluence.
"I'd much rather see a world where, when you make some quirky comment on a blog or news story or you upload a video clip, instead of just a moment of fame for your pseudonym, you'll get 50 bucks. The first time that happens, you'll realise that you're a full-class citizen."
Unfortunately there's a kind of invisible psychological line in the sand which gets crossed when money becomes involved. You can see the effects of this in Second Life, where everything is commercialized to an almost ridiculously high degree. What you end up with is a proliferation of low quality content which is being produced purely with the motivation of obtaining the 50 bucks per hit. Rewind 25 years and the world of pop music and television was exactly like this, and arguably not much has changed. In that kind of world media only needs to overcome the purchasing threshold of the lowest cultural denominator, and tends to produce very much the sort of homogeneous "mush" which Lanier is so dismissive of.
"having everything freely accessible to everyone else actually just creates a mediocre mush. The wisdom of crowds works when the crowd is choosing the price of an ox, when there's a single numeric average. But if it's a design or something that matters, the decision is made by committee, and that's crap. You want people and groups who are able to think thoughts before they share."
I'd agree that design by committee is usually bad, but this isn't really how free culture operates in practice. The works of free culture tend to come from individuals or small groups, rather than marauding hordes of maoists. If you look at how open source software is built and used for example it's really impossible to make the case that this results in bland homogeneous mushiness which noone cares about. Do either the creators or users or free software, or creative commons licensed media suffer from loss of identity amongst the crowd? I hardly think so.
There are plenty of advantages to using social networking services, like Facebook, provided that you choose your friends wisely. I've never believed in having thousands of friends - the human mind just doesn't stretch that far - and try to limit my social networking contacts to a minimal set. I'd rather have a small number of high quality friends than a large number of connections to people I have little knowledge of or interest in.
Ever since social networking sites began to become popular it seemed like a no-brainer that sooner or later this data would become of high interest to insurance companies trying to assess risk, and this article is the first that I've read which backs up this suspicion. It doesn't explicitly mention insurance companies, but since risk assessment is their core business it would seem very likely that they are also in the same game as the money lenders.
According to this article there are ways that you can try to reduce the chances of your social networking activity being leaked to companies who might use the information against your interests, but since sites like Facebook are privately owned they have both opportunity and motive to sell your data to whomever they choose, and terms of service can be altered at any time and without negotiation.
If this seems harmless, remember that insurance companies include medical insurance, and that giving these organisations more reasons to raise their premiums may make the difference between being able to pay them or not for people in some countries.
So if you think social networking is just about you and your friends, and maybe a few advertisers too, this may not be a very realistic perception. Overall though I think the advantages outweigh the perils, provided that you take sensible precautions and are careful not to post information which could be used against you. Keep in mind that you are your own public relations department online.
It's almost time once again to ask the question as to whether Google can predict the result of an election. My intuition, without looking at Google is that the coming UK general election will be a landslide conservative win. The labour party have been in power for a long time, made many unpopular decisions, and are typically held to be at least partly responsible for the current economic stagnation. But my predictions about election outcomes are frequently wrong, and Google suggests that this is going to be the case once more.
Looking at the trends for the three main parties over the last year it looks like the conservatives are consistently getting more news coverage, but that in search volume - arguably a better measurement of what the populous are really interested in - labour still ranks more highly than conservative by a ratio of 2:1. Looking back to 2009 and 2008 we can see that the search gap between the two parties has been narrowing, but not by enough to cause the lines to cross.
Going back to the previous general election in 2005 it can be seen that Google successfully anticipates the outcome based upon search volume, even though the news volume was slightly higher for the conservatives prior to election day. Unless electioneering dramatically alters the views of floating voters it seems possible that the same pattern could be repeated, contradicting my prior expectations - which are mainly derived from reading the news.
With the exit of Ben Goertzel as maintainer the AGI mailing list now appears to have entered its final entropic death spiral. The various lists and fora which I've read in relation to AI have enjoyed similar fates. Those which survived for some appreciable duration tended to reach a point of peak intelligence early in their history, after which a gradual decline ensued.
What usually happens is that trolls drive away those members who really had anything interesting to say, and replace it with noise and voluminous philosophical diatribes which few have either the time or inclination to read. There's also a phenomena which may pertain especially to AI, in that long term lack of progress leads to frustration and infighting between groups whose research has stalled after what they had thought were preliminary successes.
Yet more telerobots being built by Willow Garage. There have been a few false starts, but it does seem that the age when telerobots become commonplace might not be too far away. The main goal with something like this should really be to demonstrate that it can be used on a daily basis in a way which is helpful and requires minimal user intervention. If you have to manually recharge the thing every couple of hours then that would probably be a show stopper as far as mass market adoption is concerned. Likewise I think a lot of attention will need to be paid to the teleoperator user interface to ensure that it minimises cognitive workload. If it's too much fuss and bother to operate the robot then users will just stick with the conventional forms of teleconferencing on laptops, desktops, netbooks or mobile phones.
If it's possible to demonstrate that the robot is really a useful thing which people might want - and ergonomic factors permitting that shouldn't be too hard, because there are multiple societal and environmental forces pushing in this direction - then it's not going to be very difficult to persuade the big electrical appliance manufacturers to start getting into the robot business, because the hardware required to build a telerobot has been available for some years and is continually falling in cost.
The enabling factors are:
Ubiquitous wireless networking. Wifi and broadband internet services arrived in the early to mid 2000s
Small and cheap computers Small versions of PC motherboards have been around since the 1990s. We now have netbooks, and even some mobile phones are sufficiently powerful to run something resembling a telerobot application.
Flat screen monitors These have been around since the early 2000s. Even small screens such as those on iPhones I think would be sufficient to display the teleoperator's face.
Digital cameras Webcam type devices have been around since the mid 1990s. With the advent of USB Video Class (UVC) they're now far more standardised than they once were.
Audio-visual streaming software Has been consistently improving, and the current generation of such software - like Flash, H.264 and Gstreamer - are easily sufficient for telepresence. There are also web services such as uStream which conceivably could be adapted to fit a telerobot mode of operation.
I think augmented reality of going to become a major entertainment and information medium within the next ten years, and you can see an early glimpse of what this kind of world might be like with this demo of a mapping system from Microsoft. The integration with Photosynth is a really nice idea. What they also don't mention is that this sort of geometric data will also be very useful for automatic navigation of robots and vehicles.
Say "Bing!"
With an EyeTap-like viewing device you could be standing at a particular location and see a lot of meta data (and probably also Ads) superimposed into your field of view. You could also become a time traveller, moving forwards and backwards in time to see how things have changed. And no doubt there will also be new types of games which can be played in an augmented version of the physical world, together with whimsical applications such as highlighting all people who have similar interests to you so that it becomes easier to make new real world friends.
I've modified the head of the GROK2 robot with a new experimental camera configuration, codename "Marge". There is one forward pointing stereo camera, with pan and tilt ability, and another fixed upward pointing camera which looks at a set of five hemispherical mirrors. This reduces the total camera count to three.
If I was only researching navigation then I could remove the forward camera and reduce the number of cameras to one. The main reason for retaining the forward camera is for high resolution visual inspection - for example face or object recognition. After planar projection, the high amount of image distortion from the mirrors means that there just isn't enough resolving power to be able to see things like faces very clearly.
A key advantage is that the robot now has all round vision, which makes detection of obstacles or people easier than was previously the case. The mathematics for handling the spherical mirrors is more complex than for conventional stereo vision, but still quite manageable. Also with the mirror system there are no camera synchronisation or colour correction issues.
An earlier test of the combined omnidirectional and stereo vision system looks like this. This is effectively the same as having five cameras with wide angle fisheye lenses. You can see how each mirror views the world from a different perspective, so stereo triangulation can be applied. Having multiple mirrors also helps to reduce matching ambiguity. The mirrors are just Christmas tree decorations sawn in half and stuck onto a sheet aluminium backing, so the whole arrangement is very lightweight and most importantly cheap to produce for recessionista roboticists.
Although the omnidirectional arrangement looks rather unwieldy, if smaller mirrors were used the whole system could also be made more compact, and for an industrialised system probably you would need to encase it within a clear plastic or glass cylinder to protect the camera and mirrors from contamination. A cylindrical encasement would also eliminate the need for the metal struts, maximising visibility.
An amusing BBC story about internet chat users being tricked into giving out personal details. This is an example of a real Turing Test being administered out there in the wild, and I imagine that if it's even mildly successful then crooks may start moving into the chatbot business. After all, just like the spam business model they only need a small percentage of users to be fooled for the enterprise to become financially viable.
It would be ironic if the first chatbot to consistently pass a Turing Test were to be developed in secret by some criminal organisation, rather than via the conventional AI research routes. There's already some history of this in terms of CAPTCHA cracking, but being able to fool users via conversation for a short period of time would be a more impressive achievement.
Kevin Warwick comes out of the closet as a singularitarian, in this video. I've read his earlier books, but I don't recall him mentioning anything like a "singularity" in those, although it's obvious that his message of doom fits quite neatly with the notion of a singularitarian crescendo.
It sounds as if UAVs may be adopted by the police force more or less in line with my previous expectations, according to a Wired article. If you consider the economics of the case, it's really difficult to argue against using drones.
Increased surveillance using drones is a relatively small outlay of cash for government agencies, especially compared to an equivalent service using manned helicopters, and in these austere times surveillance can be used as a mechanism to raise additional local taxes by levying fines upon anyone caught on camera who can plausibly be described as violating some law or code of conduct. Even if the surveillance data turns out to be questionable, or the wrong individuals are identified, often - as is the case for traffic offences - it's easier just to pay the fine rather than face potentially much larger legal costs and have to take time off work to attend court proceedings.
One thing that I think drones will be used for, apart from the predictable traffic monitoring, is enforcement of planning regulations. If you've built a garage, shed or greenhouse in your back yard, which can't be seen from the street and which you think nobody except yourself will care about, without going through the official channels and paying bribes (aka "application fees") to the requisite government departments, then drones could do a pretty effective job of detecting changes like this and issuing automatic fines. Automatic policing is attractive to government agencies, because it's potentially a highly profitable business. For this sort of task few or no human operators would be required, since all the drones need to do is compare previous maps with current ones and spot differences.
It's suggested that non-lethal weapons might be used on policing drones. This probably will occur, but it's implementation will need to be very carefully planned. Using these things in a military situation is one thing, but using them for policing poses a quite different set of liability issues. What if someone's hearing gets damaged by a drone LRAD blast that had been incorrectly calibrated? What if a strobe weapon happens to hit bystanders who have epilepsy? In a civilian situation all these people can potentially sue for damages, and suddenly your drone operation goes from being a nice little earner to looking not so profitable after all. So I think that if these weapons are fitted, they will only be used quite selectively in a limited range of circumstances where the legal ramifications are fairly well circumscribed.
But of course under less liberal political regimes, where people have no rights to sue and complaining can result in imprisonment (or worse) then sonic, strobe or even more harmful weapons usage could become very commonplace as an instrument of state terror keeping the population in check.
You can see an example sonic device being used on a UAV helicopter in the video below.
In this case it looks as if the LRAD has just been crudely bolted to the side of the craft for the purposes of carrying out a quick test. It would be better to mount it on a pan and tilt mechanism slung underneath, in a manner similar to the sensors on predator drones. That way the weapon can remain fixed on the target whilst the drone moves from one location to another.
Non-conventional uses
An then there's the small issue of non-conventional uses - something which I've mentioned on numerous previous occasions. Building a UAV with a weapon attached probably isn't a highly difficult or especially expensive task. I've seen large model aircraft which were mechanically far more complex than the helicopter in the video above, with ducted fans or multiple engines.
So I think it's entirely plausible that some lone lunatic with a grudge, or a terrorist group, or other "non-state actor", could build such a machine and launch it over a city. How would you stop a UAV in flight? It's probably too small and slow for military jets to shoot down, and won't have much of a radar signature. Chances are that before it can be intercepted it will have expended all of its ammunition.
These are amongst the tricky issues that increased use of telerobots will bring. Perhaps anti-UAV UAVs are the answer.
I've created a text to speech script for use with the X-Chat IRC client. This is inspired by the speech bot used by Jeri Ellsworth. My version isn't quite up to the same level of sophistication as Ellsworths, which uses multiple computers and the better quality synthesised voices available on the Mac, and I don't have a parrot, but it's still quite functional.
I'm not a big IRC user, but the main advantage of using something like this is that enables you to be doing something else other than looking at a screen, but still know what's being said in the chat channel.
The script is my first attempt to write anything in Python. Normally I wouldn't bother with interpreted or scripting languages, since most of my projects are related to computer vision where the speed of execution is important and lots of calculations are occurring, but for general internet related stuff speed is not so critical. An interesting feature of Python is its lack of curly brackets, which was a bit of a culture shock. Instead it just relies upon the indentation level to recognise scope. This is actually quite a good idea, which avoids the sometimes confusing sea of brackets which can occur using other languages, and it also forces the author to use a sensible indentation policy.
To install on an Ubuntu/Debian based system:
sudo apt-get install python espeak xchat-gnome
Download and extract the script from here. The current version is 1.02 codenamed Chatoholic Chinchilla.
Copy the xctts.py script into the ~/.xchat2 directory. Then open X-Chat and type:
/talk on
to enable speech. Each user is assigned random speech parameters, but they can customise their voices using a few commands (see the README for details).
This is also my first significant use of Launchpad. Since Canonical are based in the Isle of Man this means that the terms of service can't be restricted by the absurd and discriminatory US export regulations. I'm aware that Sourceforge have recently changed their policy with regard to US export, but the situation for them - and presumably other US based project hosting sites - remains highly unsatisfactory and incompatible with the terms of the GPL license and section 5 of the Open Source Definition.
An amusing dialogue between Ben Goertzel and the Siri iPhone personal assistant can be found here. This is in the same kind of style as Aaron Sloman's conversation with an Ikea chatbot some years ago. What this demonstrates is that even after decades of exponential improvement in computing hardware, the software needed for natural language understanding has barely improved at all since the Weisenbaum's original ELIZA chatbot. This seems to be an ELIZA with hyperlinks and maps added.
It's tempting to assume that after a few more years of refinement and accumulation of corpora something like the Siri assistant will improve in its understanding of common English usage, but the historical precedents are not good. This consistent lack of progress over a long period of time suggests to me that computational linguists are framing the problem in the wrong sort of way, and perhaps making mistaken assumptions about what language is and how it's commonly used and understood. On the other hand it might really just be that we don't have enough computational power for this task yet, but I'm becoming rather doubtful about that since an application like this could have access to virtually limitless "cloud" (aka server) based computing resources. Yet another possibility is that the training corpora for this sort of application are not sufficiently large, and that after several years of accumulation enough data may exist to be able to train a much more competent language understanding system.
A phenomena which I previously had no idea about, but which I stumbled across during pseudo-random internet browsing is that of number stations. The most popular interpretation of these unattributed numerical broadcasts by synthetic voices, which seems to be supported by a certain amount of evidence, is that they're coded secret messages being sent out to spies around the world.
That this system has been going since the first world war - practically as long as radio broadcasting technology has been around - indicates that the one-time pad is a simple yet effective cypher technology. Hypothetically, the secret agent - and here I'm imagining a Clouseau-like character - is issued with a pad containing numbers as he departs from the spy headquarters. There's maybe a set of numbers for each day or each week, depending upon how often he needs to receive orders, and in internet terminology terms these are his private decryption keys. Later on he can use a fairly crude radio gadget, pre-tuned to the relevant frequency to listen for his unique call number at a specified time of day. Even if he loses his gadget, he can still improvise by using an ordinary analogue portable radio, which are sufficiently commonplace in most areas of the world not to raise suspicion. He can then write down the numbers and decode them according to some particular algorithm using his private key. Provided that each key is used only once, and he destroys the message and the key after receiving it, it's a pretty secure system. The only potential weakness is retaining physical security of the one-time pad.
You can listen to an example of a number station broadcast here.
This isn't the first time that there have been warnings about possible interruptions to energy supplies in the UK within the next five or ten years, so I would treat this as a serious possibility. Before you rush out to buy an ebook reader, iPad or other electrical gadgets it may be worth considering whether you would still be able to use these if electricity supplies were to become significantly interrupted, or if the energy prices were to rise by 20% or more. It might also be worth considering whether you would still be able to light your home at night without a regular mains electricity supply, which is pretty much a minimal safety requirement.
"Our evidence shows that Britain has a window of opportunity to put in place far reaching reforms to meet the potential security of supply challenges we may face beyond the middle of this decade. We do not advocate change lightly, but all the facts point to the need for reforms now to provide resilient supply security."
It looks as if privatised energy utilities have maintained the status quo, but not invested sufficiently to ensure future supply. I see this as just a short term problem though. In the longer term energy supply - at least in terms of electricity - isn't going to be an issue. As mentioned in a previous blog post, a certain amount of heavy engineering in northern Africa, together with appropriate international agreements should mean that Europe can be self sufficient in electricity.
So after the next general election if it appears that the next administration isn't taking energy issues seriously, or is only paying lip service to them without the accompanying investment in engineering projects and changes to regulations, then it might be wise to adopt a slightly more survivalist mindset in the next five years in order to make sure that you can still live reasonably well if power interruptions do occur.
Some Americans are bound to hate the changes at NASA, but this actually does look like a progressive space policy, rather than continuing to languish in low Earth orbit or attempt to recapture past glories on the moon using near identical hardware to that of the 1960s.
Greater involvement of commercial entities in space rocketry will help to reduce costs and increase competition and innovation. In the previous paradigm where space vehicles were built and owned by the government, as epitomized by the space shuttle, there's really very little incentive indeed to reduce costs and plenty of incentive to add more expense and use the whole exercise as a bureaucratic job creation scheme.
If the cost of getting stuff out of the Earth's atmosphere can be reduced then that opens up other possibilities for the future, such as manufacturing, mining, energy generation and even leisure. Some commentators seem to be worried about safety of astronauts if costs are cut, but I doubt that anyone becomes an astronaut believing that it's a safe and easy occupation. In any genuinely pioneering effort occasional accidents can be expected, although you can try to put procedures and checks in place to minimize their frequency and severity.
