On Bullshit

Some thoughts from Harry Frankfurt on the nature of bullshit, of which in contemporary life there seems to be an ever increasing quantity.



The comments about the relation to sincerity reminds me of the situation with Tony Blair over the Iraqi weapons of mass destruction. The public, and to some extent also the media, were primarily concerned with whether the claims being made were true or not, but to Blair what mattered was not the truth or falsity of the matter but instead whether he sincerely believed that the weapons existed. This highlights the tenuous predicament of the bullshit artist.

Of course blogging represents the very pinnacle of bullshitting, so I by no means exempt myself from membership of this category. But I think (however inexpertly) that BS transcends merely an agnostic relationship to the truth. Much of what seems to be going on could be ascribed to a sort of social persuasion or social bonding - a kind of glue which is maintaining some coherency between members of a group. Large swathes of politics, religion, media and marketing falls into this persuasiveness and bonding mode of communication. What's being said could be true or could be false, but in a sense the content of the message itself doesn't matter so long as it serves the function of binding the group in collective action. Without some degree of bullshit, goods might not be sold, relationships might never be formed, polite conversations about the weather never be engaged in and novels go unwritten.

The telerobotic bosses

An interesting article on the various uses of telerobots highlights, as might be expected, that the main hurdles involved are the human factors rather than technological ones. For mobile telepresence the technology required such as high speed wireless networks, cameras and small computers has been easily available at consumer price levels for at least the last few years.

The article suggests that a telerobot is viewed by its users as more than just a webcam on wheels, and that people soon associate the robot with the corporeal presence of the teleoperator rather than merely a piece of machinery.

In an office where a telerobot is operated by a single user it takes about 4 days before coworkers start to refer to the robot by the operator’s name. It’s not, “where’s the robot?” it’s “where’s Jeff?”

In previous work I have also come across the difficulties of deploying applications containing cameras into "secure" environments, so it's not too surprising that similar red tape also applies to telerobots. In some 2D barcode verification software which I wrote a couple of years ago I added a "paranoid mode" which permitted the system to visually verify markings without displaying any images or video on a screen at military sites. So banks and military facilities are probably not a good types of industry for early adoption of roaming video cameras, although once this becomes a more normal practice in the wider society I expect that attitudes and red tape will change out of necessity.

Another observation isn't what I'd expect.

About 30% of operating time is spent driving the robot around. Nearly 100% of users, however, didn’t want this time to be reduced. Driving around time was great for walking conversations, thinking, etc.

This might be the case if you're exploring a new remote site, or only visiting it via the telerobot quite infrequently. If you're teleoperating the robot on a more regular basis though at the same location I think the initial novelty is going to wear thin and that operators will want to keep the driving time to a minimum. This intuition comes partly from driving telerobots myself, such as the Surveyor SRV and GROK 1, and also from articles describing military use of Pacbots where there seems to be a demand for greater autonomy and reduced manual driving time. A compromise similar to that used on the rovers on Mars might be a simple waypoint based driving system, where the operator can designate a point on the floor that they wish to drive to and the moment-to-moment control of movement to the destination is then handled by the robot.

Also on the topic of human factors there's the issue of migration. Is a telerobotic worker "migrating" in some sense if the robot is located in another country? For pure telepresence this isn't an issue, but if the robot is going to be used to carry out something resembling a job, such as a doctor performing surgery, then red tape is definitely going to get involved. Many countries have quotas for migrant workers, and especially in hard times migrants generally become a focus of suspicion and hatred. Governments may view telerobotic working as a way for compaines to evade paying local employment taxes such as National Insurance.

I'll be very surprised if telerobotics doesn't become a growth area within the next five years, because all of the technology is here and there are social, environmental and economic forces pushing in this direction. No doubt the Great Recession has been holding things back, but that won't remain the case indefinitely. A decade from now I think it's entirely possible that telerobots could be so ubiquitous as to be thought of as unremarkable - like office chairs or domestic electrical appliances - and its quite foreseeable that teleoperation interfaces could become an integral part of social networking systems ("click here to visit me", etc).

How to avoid signing NDAs

Eric Raymond dispenses some advice on how to avoid signing the dreaded Non Disclosure Agreements (NDAs). I've signed a few in my time, but on occasion managed to avoid them by quibbling over the legal minutiae. For example, at a job interview a few months ago I was asked to sign an NDA before getting a tour of the factory. Amongst several objectionable paragraphs the most ridiculous was one requiring that the agreement not only by binding on myself but also on my descendents too. I don't have any descendents, but can't rule out the possibility that I might have some at a yet to be determined point in future. It seemed unreasonable that persons as yet unborn should be legally compelled to observe the terms of an agreement made by me, so I refused to sign. Just as Raymond suggests, they had little interest in defending the zany terms of the NDA, saying that it had been written "by some American lawyer", and I still got my factory tour regardless.

In general it's a good idea to be wary of signing anything of a legal nature, unless you're sure what you're signing yourself up to. Otherwise you could be signing all of your possessions, inventions and/or rights away or doing yourself other kinds of disservice. Next time another NDA situation turns up I may try following Raymond's advice. Talking about fiduciary responsibility is certainly likely to bamboozle most secretarial workers, unless they've previously studied law.

Howard Gordon, RIP

I was saddened to hear of the death of Howard Gordon of Surveyor Corporation with whom I worked quite extensively over the last year or two on the Surveyor SVS. A quick search shows up about a thousand email correspondences with him in the last 12 months. Howard was clearly very talented, knowledgeable on software and electronics, enthusiastic about robotics and always helpful in resolving bugs or suggesting changes.



I don't know how much I should divulge, because presumably Surveyor Corporation will continue, but the last I heard at the beginning of this year he was working on some interesting stuff relating to navigation of robots in a home or office environment. By the beginning of 2010 I'd developed the firmware stereo algorithms about as far as they were likely to go, given the CPU and memory constraints of the Blackfin DSPs, and had done some experiments with occupancy grid mapping and visual odometry using the SVS. These early mapping experiments were purely open loop, since the surveyor robot I was using didn't have closed loop speed control, and so suffered from the gradual drift which all such systems encounter. However, towards the end of 2009 Howard was developing a new system which would have an accelerometer for inertial sensing, encoders for closed loop motor control and GPS for use outdoors. I was optimistic that combining the vision system with this additional sensing could yield a pretty good navigation capability at a cost in the order of a few hundred dollars.

Towards public sector accountability

Open sourcing public spending in CSV format is an excellent idea, and I find myself in the rare position of agreeing with a policy of the current government. In the past the details of how government departments spent public money was largely secret, and widely suspected to be subject to corruption. Perhaps it may have been possible to uncover some aspects of public accounts by visiting a library or making a freedom of information request, but few people - even professional journalists - would have bothered to go out of their way to do that, since you would need to have a good idea of what you were looking for in advance, and the information would have been at a high aggregated level rather than on an item by item basis.

This helps to close the loop on public accountability, allowing interested citizens to raise concerns if they find anything which looks excessive or inappropriate, or perhaps even to suggest alternative spending strategies which might deliver better value. Closed loop systems generally perform far better than open loop ones, and it's much harder to slide towards tyranny or become seriously out of step with public expectations if a balance of power exists between governors and the electorate.

However, I would like to see the transparency go further down to items of lower cost than £25,000, otherwise corrupt bureaucrats could simply make repeated payments of slightly less than this amount in order to remain under the radar of scrutiny. I'd also like to see publishing of spending details enshrined as a legal requirement for all future governments, rather than merely being just the policy of one particular administration.

Robonaut

After more than a decade in development it's encouraging to see that Robonaut is going to get some real testing in space.



By now we're quite accustomed to seeing astronauts busying themselves around the international space station. They make it look easy, but this is difficult and dangerous work, with one of the biggest hazards being exposure to solar radiation, which a space suit doesn't provide very much protection against. If a robot had a similar degree of dexterity to a human astronaut it would be far wiser to carry out construction work in space via telerobot. Use of such robots would also make space based construction a far more scalable proposition, since the human operators could potentially be ground based - especially for geostationary construction sites.

As always with telerobots the ergonomic factors will be the decider as to whether Robonaut is a success or failure. Really it needs to be as easy and intuitive as possible to operate the robot, so that the user doesn't quickly become tired, frustrated or disorientated. There may also be Terrestrial applications of this sort of development in terms of things like telerobotic care work or working in hazardous environments (fire fighting or deep sea operations, for example).

Lane departures

In terms of image processing this fruit is about as low hanging as it gets. Detection of lanes on motorways is really a special case, with the markings on other types of road being not nearly so consistent in appearance.



The processing needed to do this is fairly minimal, so it could be implemented on an FPGA or DSP inside the camera housing itself, with a GPS receiver being used to calculate approximate speed so that it doesn't need to be wired into the car's other systems. Alternatively speed could be estimated from optical flow, although that would require more processing power since it involves finding correspondences between frames.

Some other vehicle image processing features can be seen here.



Presently I think that systems like this are only implemented on very high end luxury vehicles, but I expect that before long this will be a standard feature in all vehicles.

Pedestrian detection

If this system works as well as advertised then I can imagine car insurance being higher if you don't have one of these fitted.



Knowing how fallible vision systems can be, especially when you have distortions produced by rain or snow or low lighting conditions, false positives could really be a problem if they cause the brakes to be automatically activated. The vision sensor could be combined with GPS to ensure that it's only active in areas where pedestrians are likely to be present (i.e. not on motorways), and if you wanted to get even more clever the pedestrian detections could be stored and uploaded to a central server in order to calculate a "probability of pedestrians" at each location, which can then be used as a prior probability by the detection algorithm.

Also another way to help reduce road accidents might be for the vehicle to have access to a traffic accident statistics database somewhere in "the cloud". You could then maybe have a coloured indicator on the dashboard to provide some notification to the driver if the GPS location is within an area where accidents are more frequent than average.

Stereo Assist

A talk by Gideon Stein about using stereo vision for detecting vehicles and pedestrians. It's an expert system, so this only applies in the narrow context of driving, but still could be quite useful. A combination of stereo correspondence and monocular object detection is used to improve the ranging performance. The fact that they aren't using longer range correspondences with small disparities suggests that they're not using sensor models or doing structure from motion. Possibly I could add this sort of functionality to v4l2stereo if I can get some test images.

Doomster status

The C-Realm podcast tackles the slippery topic of doomsterism - something which will be familiar to artificial intelligence aficionados. The basic thesis seems to be that doomsters are primarily low status individuals, who are seeking to enjoy an elevated status in the post-doom world which they have so cannily predicted. I'm not sure that this entirely explains doomsterism, and indeed the very vague and inaccurate nature of virtually all doom predictions means that doomsayers never actually get to the stage of being able to look back and say "I told you so".

Tim Tyler provides a more multi-faceted summation of the motivating factors behind doomsterism, but there may also be simpler origins. When a blackbird sees a cat prowling in its neighbourhood it sounds the alarm, partly to scare or annoy the cat, triggering a reflex called acoustic startle, and partly to alert other birds to the danger.



There's an obvious adaptive advantage to this behavior, and prognostications of doom may be a human equivalent of the bird's alarm call, explaining why so many people pay attention to them.

The main brand of doom discussed in the podcast relates to peak oil. The peak oil idea itself seems pretty difficult to refute. We have a good idea of how oil naturally formed, and we know that new oil is being created much too slowly for our rate of consumption to be sustainable. Sooner or later the oil will begin to run out, and most of the "easy oil" (that is, where you don't need to do much more than drill a hole in the ground to extract it) may already have been produced, as indicated by the declining volume of new discoveries. The main problem for doomsters is that oil will not run out suddenly in a single catastrophic event - like turning off a tap - but instead will from the consumer point of view appear to be becoming ever more expensive over time. Also, there are alternatives to oil, and possibly also synthetic ways of manufacturing oil using bio-engineered organisms.

The existing manner in which markets operate - however imperfectly - would seem to ensure that there will be a transition away from oil and towards other energy and hydrocarbon sources. As oil becomes scarcer and more expensive, the alternatives start to become cost effective by comparison and there will be more financial incentive to invest in those alternatives. If the transition is poorly managed at a political level, with anti-competitive interests seeking to block the transition for as long as possible, then there could be more wars or disputes in oil producing areas of the world, but unless these escalate into nuclear wars this isn't a scenario which spells doom for the whole of humanity. Things might be rocky for a while, but it's not an unsolvable problem. Doomsters typically assume that the status quo will continue until it breaks - a catastrophic state transition - and underestimate the adaptability and ingenuity of humans when faced with a challenge.

Here in the UK I expect doomsterism and doom-meisters to be on the rise in the next couple of years. Cuts in public services and rises in taxation will probably cause some degree of unrest, strikes, protests and so on. But this is really just civilization-as-usual, and shouldn't be confused with a species extinction event (although it may cause the extinction of the coalition government).

Mandriva paywalls

In 2003/4 Mandriva, then called Mandrake, was my favourite Linux distro and was probably the easiest to use out of the various distros which I'd tried up to that point. I think that even back then they were having financial difficulties, and I don't think that much has changed about their financial predicament since.



As this video suggests, trying to get users to pay for things which in other distros "just work" is the sign of a bad business model. Technically they're perfectly within their rights to charge for Gstreamer or other software components, since it's kosher to charge for free software if that's what you want to do. However, they would be much better off if the paying element was concentrated on product differentiating applications or services. Canonical does this with Ubuntu One and their music store, so there are existing precedents. If basic multimedia functionality which is nothing new appears to be broken or have a paywall in front of it, then users are just going to move along to the next non-blocking distro - I know I would.

Open Source in Government

A presentation by Arjen Kamphuis about the introduction of open source software for government use in the Netherlands. Back in 2002 I think the Dutch rail service could perhaps be forgiven for having a web site which only worked with Internet Explorer. At that time very few people were using Linux as a desktop operating system, and the browser wars had been over for some years with Microsoft having won by dumping and/or bundling its browser at zero cost.

I expect that something similar is about to happen in the UK. Vague noises have been made in manifestos about the possibility of using open source in government services, but if they really need to cut out somewhere between 20 and 40% of overall spending then they'll simply have no other option than to pursue the FOSS route. Possibly free web based services could be used, such as Google Docs, and even though Google may be relatively benevolent thus far, dependence upon a single proprietary supplier to run an entire government is probably not a recipe for success, with the possibility of improper influence, back doors or withdrawal of services if international relations turn frosty.

Mind Monkeys