Maybe the new eugenics in the 21st century is going to be quite benign after all. This documentary highlights the uses and abuses of Darwinian ideas in society, and there's a depiction of a sort of eugenics where children have DNA or blood samples taken which are then stored, and if two adults later want to have children they can then consult the database to decide whether they are "genetically compatible" and determine the probabilities of having offspring with certain genetic diseases. Assuming that the genetics information is reliable, that seems like a much more reasonable way of doing things, rather than forced sterilisations or genocide, and it's just a private decision where the state doesn't get involved.
There will no doubt be a lot of other dilemmas which arise from a better understanding of genetics. For example I can imagine a court case in which an employee sues their employer because a work situation in which they were placed caused a particular deleterious set of genes to be expressed, affecting their health. Situations like this might mean that certain types of jobs or insurance policies are only available to people with certain types of genetics or epigenetics, and this is all bound to be highly controversial.
Monday, February 28, 2011
Sunday, February 27, 2011
Robonaut: Scaling space manufacturing
Hysterical criticism of humanoid robots in space is probably misguided. I think that there is a place for humanoids, provided that tasks involving the combination of dextrous manipulation and sophisticated cognition, such as repair and assembly, remain difficult to automate entirely with AI. Humanoid telerobots would be a much safer way of providing exterior maintenance and assembly of space stations, avoiding exposing astronauts to radiation risks and also permitting greater scalability via ground based teleoperation.
The "project M" scenario, as depicted in the following video, is probably not the best use of robotics though, unless you wish to begin doing industrial manufacturing on the lunar surface. For exploration or science purposes a more conventional rover type of design would be likely to be more successful.
Those who are now criticizing Robonaut havn't thought carefully enough about what it would take to do things in space other than science in a cost effective manner - like manufacturing, mining and eventually building habitats for humans - under a conservative expectation that critical AI problems may remain difficult for at least the next few decades.
If after some testing it turns out that Robonaut can perform some basic teleoperated tasks, such as using a spanner, saw and drill (the same tools which astronauts use on EVAs), in a microgravity environment and in a way which isn't too demanding or confusing for the human operator, then the possibility exists to scale this up and have a workforce permanently stationed in orbit ready to receive modular components lifted by commercial rockets and assemble them on demand and without years of prior planning into whatever vehicles or products are required using terrestrially based tele-construction workers. Teleworkers on Earth can be hired from the construction industry, at pretty much regular wage rates, and if a humanoid design is used then the amount of re-training which they'll require can be minimized.
With that sort of manufacturing ability more ambitious projects become possible, such as the construction of large ships which can mine asteroids for materials, solar power factories or sending humans on extended deep space missions. A point worth noting is that humanoid telerobots do not necessarily need to be the same scale as a human - they could be much larger or much smaller, depending upon the tasks being undertaken. From an ergonomic point of view all that matters is that the body proportions are sufficiently similar to a human to make the operation job as natural and cognitively lightweight as possible.
The "project M" scenario, as depicted in the following video, is probably not the best use of robotics though, unless you wish to begin doing industrial manufacturing on the lunar surface. For exploration or science purposes a more conventional rover type of design would be likely to be more successful.
Those who are now criticizing Robonaut havn't thought carefully enough about what it would take to do things in space other than science in a cost effective manner - like manufacturing, mining and eventually building habitats for humans - under a conservative expectation that critical AI problems may remain difficult for at least the next few decades.
If after some testing it turns out that Robonaut can perform some basic teleoperated tasks, such as using a spanner, saw and drill (the same tools which astronauts use on EVAs), in a microgravity environment and in a way which isn't too demanding or confusing for the human operator, then the possibility exists to scale this up and have a workforce permanently stationed in orbit ready to receive modular components lifted by commercial rockets and assemble them on demand and without years of prior planning into whatever vehicles or products are required using terrestrially based tele-construction workers. Teleworkers on Earth can be hired from the construction industry, at pretty much regular wage rates, and if a humanoid design is used then the amount of re-training which they'll require can be minimized.
With that sort of manufacturing ability more ambitious projects become possible, such as the construction of large ships which can mine asteroids for materials, solar power factories or sending humans on extended deep space missions. A point worth noting is that humanoid telerobots do not necessarily need to be the same scale as a human - they could be much larger or much smaller, depending upon the tasks being undertaken. From an ergonomic point of view all that matters is that the body proportions are sufficiently similar to a human to make the operation job as natural and cognitively lightweight as possible.
Thursday, February 24, 2011
Control theory
Quite a good series of lectures on control theory. These principles apply not just to industrial engineering, but to any self-regulating system, such as a thermostat, a living cell, your heart rate and blood pressure, companies, cities, governments, economies and the biosphere as a whole.
Saturday, February 19, 2011
Friday, February 18, 2011
AV will make no difference
There is supposed to be an upcoming referendum in May on whether or not to change the voting system in the UK from first past the post to an Alternative Voting (AV) system in which voters rank their choices (first preference, second preference, etc).
I think what's needed isn't a change in the type of voting, although a move to more secure and verifiable voting would be a benefit. What's needed instead is a deeper democracy, where people get to vote on decisions which actually impact upon their lives, such as for example local council spending, combined with what's called "open government" where voters can freely get access to the data generated by government agencies.
There is progress being made on the open government aspect, but what voters can do once they have more access to information and become better informed still remains extremely limited, to the extent that it violates the law of requisite variety as an effective method of governance. As a citizen I can vote every four or five years for a new government, but that's both the beginning and the end of my participation, and those who are being voted for frequently don't even make a good faith attempt to abide by their manifesto pledges anyway. In the absence of enough variety this vacuum is occupied by other interest groups - big companies and wealthy individuals - who generate variety through a complex system of lobbying and social networking.
Fundamentally off-topic
The pro/anti AV debate is an interesting phenomena in itself, because I expect that a great deal of fervent argumentation is going to take place in the Westminster village and in newspapers/TV over the next couple of months, whereas in reality the particular type of voting system used could hardly be a topic more remotely distanced from the list of high priority issues currently concerning UK citizens, which are typically connected to unemployment, inflation, education, taxes and public services. If there are any debates held in public I'm expecting to hear cries of "we don't care about this issue, we want referendum on X", where X could be university fees or NHS privatisation or one of many other hot topics.
Exasperation with the status quo
On a personal note, my view of politics as currently practiced has not improved one iota, and is unlikely to change even if AV is adopted. In the 2010 elections I tried to do what voters are theoretically supposed to do, by examining the various manifestos and trying to determine which one most closely matched my views. Luckily the party I (belatedly) voted for actually did get into office, but already we can see the usual practice of manifesto pledges being jettisoned or ignored which leaves me wondering whether voting really has any meaning at all, other than perhaps as a woefully inadequate gesture of participation.
I think what's needed isn't a change in the type of voting, although a move to more secure and verifiable voting would be a benefit. What's needed instead is a deeper democracy, where people get to vote on decisions which actually impact upon their lives, such as for example local council spending, combined with what's called "open government" where voters can freely get access to the data generated by government agencies.
There is progress being made on the open government aspect, but what voters can do once they have more access to information and become better informed still remains extremely limited, to the extent that it violates the law of requisite variety as an effective method of governance. As a citizen I can vote every four or five years for a new government, but that's both the beginning and the end of my participation, and those who are being voted for frequently don't even make a good faith attempt to abide by their manifesto pledges anyway. In the absence of enough variety this vacuum is occupied by other interest groups - big companies and wealthy individuals - who generate variety through a complex system of lobbying and social networking.
Fundamentally off-topic
The pro/anti AV debate is an interesting phenomena in itself, because I expect that a great deal of fervent argumentation is going to take place in the Westminster village and in newspapers/TV over the next couple of months, whereas in reality the particular type of voting system used could hardly be a topic more remotely distanced from the list of high priority issues currently concerning UK citizens, which are typically connected to unemployment, inflation, education, taxes and public services. If there are any debates held in public I'm expecting to hear cries of "we don't care about this issue, we want referendum on X", where X could be university fees or NHS privatisation or one of many other hot topics.
Exasperation with the status quo
On a personal note, my view of politics as currently practiced has not improved one iota, and is unlikely to change even if AV is adopted. In the 2010 elections I tried to do what voters are theoretically supposed to do, by examining the various manifestos and trying to determine which one most closely matched my views. Luckily the party I (belatedly) voted for actually did get into office, but already we can see the usual practice of manifesto pledges being jettisoned or ignored which leaves me wondering whether voting really has any meaning at all, other than perhaps as a woefully inadequate gesture of participation.
Thursday, February 17, 2011
Deconstructing the Zeitgeist Movement
I've seen the various Zeitgeist movies, including most of the first one which was pretty dubious and has been widely criticized. Here is some commentary on the latest such movie, called Zeitgeist: Moving Forward.
If you've seen any of the other Zeitgeist movies beyond the first one then you can save yourself quite a lot of time because the latest version doesn't contain much in terms of new concepts, but rather it presents the existing ideas in a slightly more entertaining way. If not, then here is the latest one.
In my estimation, I'm sympathetic towards the resource based economy idea, and taking a systems approach to running things. In the long term actually I don't think there's going to be very much alternative to managing economies as integrated machine-managed systems, as opposed to the highly unsustainable and periodically dysfunctional way in which things work currently. The depiction of future cities is nice, but there's also a vast amount of hand-waving here (practically flapping) and not very much depth to the grand visions.
A world without money
One of the major themes of Zeitgeist is the possibility of a world without money, where nanotechnology - or some other futuristic wizardry, perhaps involving robots - can synthesise and recycle products on demand. For many sorts of consumer products this may be quite feasible, and it's not too difficult to imagine bigger and better RepRaps combined with more traditional forms of industrial automation. But there may still be competing zero sum interests, and how should these be arbitrated? Suppose citizen X wants a boat and citizen Y wants a car, but the computerised management system indicates that there is not enough resources in the environmental inventory to satisfy both demands simultaneously. There may still be a need for some forms of trade, and where there's trade there is inevitably a need for some kind of abstract representation of value. The non-homogeneous distribution of resources around the planet will surely mean that there will be surpluses of resources in some areas and deficits in others, and traditionally this has been addressed with some kind of trading system.
If money still exists in the distant future, perhaps cryptocurrencies are the way forward, avoiding the issues of financial fraud and centralized banking which currently bedevil our monetary system.
Scientism
My other main criticism is that the depiction of science in the Zeitgeist movies.
Firstly the treatment of the role of genetics and environment seems rather loaded towards a notion that genes don't matter very much. In human society there are numerous phenomena which have little or nothing to do with genes, but claiming that none of our behaviors or biases are in any way regulated by genetics I think is probably not well founded in contemporary biology. For instance, it's quite likely that it is to some degree "genetically determined" that most humans have two legs, and five fingers on each hand. If the morphology of the body is regulated by genes it seems at least plausible that the morphology of its organs, including the brain, are also influenced to some extent by genetics. Our understanding of genetics is still at quite an primitive stage, so there will be much more waiting to be discovered, but it's likely that life is a deviously complex mashup of genetic and environmental interactions, where genetics does have some role in setting up biases and propensities towards certain kinds of behavior or emotional reactions. An example would be that it's believed that some people possess tetrachromatic vision, which gives them slightly different visual sensing capability and maybe different behaviors or behavioral biases become possible as a consequence.
Also in Zeitgeist it's assumed that there is a single "correct" solution to resource management problems. In complex systems such as a society usually there are a range of possible solutions, all with their own accompanying costs and benefits. None of these may be "the best", and choosing one depends upon what direction you wish the society to move in. This may be something which requires decisions to be voted on by citizens, rather than made automatically, otherwise you're putting a computer system in charge of the long term overall evolution of society - which may be asking too much of a system whose main function is resource allocation.
Future Architecture
In the podcast there's some criticism made that the circular cities envisaged aren't easily scalable. This is a reasonable criticism, but not really a show stopper. I can imagine other designs which would scale in a modular way.
If we're thinking about the architecture of the distant future, say a century or more from now, then this is a time when resource shortages of raw materials is going to be ever more acute and an increasingly large fraction of resources will need to be either reusable, recyclable or decomposable (made from organics). Assuming more advanced genetic engineering capabilities it might be possible to have buildings which are partly or entirely organic - made from a wood-like material which can self repair by extracting carbon from the air using sunlight as an energy source to assemble it into a super-strong nanotube like material. Another idea would be to have the walls of buildings contain a soil layer within which the roots of plants could grow. This might mean that you can grow vegetables along the sides of high rise buildings (rather than only in the windows) and have robots periodically harvesting them as part of a localised intensive food production system capable of supporting the appetites of megacities. These green buildings could be fertilised by processed sewage produced by the occupants, reducing the strain on sewer systems which has been a perpetual issue for large conurbations, and I think there are existing "vertical farming" proposals of this kind already. Under this scenario buildings would from a distance have a fuzzy appearance and varying colouration, depending upon what type of vegetation was growing on them. Residents could also get creative with configurable robotic plantation schemes to create artwork, brand logos, slogans or badges of status/allegiance along the sides of their residences. Another, perhaps more likely, situation would be that all buildings are coated with a photovoltaic material which produces electricity.
As for monolithic "totalitarian" style of construction, I agree with the comments made in the podcast about this not reflecting the diversity of human culture. Cities in the future are likely to be at least as diverse as they are now, containing a combination of old and new. A good example of this is the city of York in which I lived for quite a while. It's a 2000 year old city, where you have architecture which is many centuries old standing next to modern office blocks, and a sewer system which was constructed by the Romans in the first few centuries of its existence. It's a rare occasion where cities are built entirely from scratch, and even then all cities evolve over time in a way which is difficult to plan for unless you have a crystal ball which can foresee future transport systems and types of economy (the original roads of York were the width of a chariot, i.e. two horses side by side). I've seen the totalitarian architectural style myself in various cities. Typically it was built in the 1960s and consists primarily of grey concrete blocks as far as the eye can see. It's pretty depressing, and no doubt is a contributory factor to the psychosocial stress of its inhabitants. But people can be more imaginative than this, especially when they're empowered by technology.
Zeitgeist as a meme
On the whole in spite of their failings and vagueness the Zeitgeist movies (at least the later ones) are a well crated attempt to bring ideas from cybernetics into the public arena. I'm not aware of anyone in the mainstream media who even remotely touches upon these sorts of issues. If you're not paying attention then it's easy to adopt a whiggish attitude towards the way that current societies and economies operate, but the only thing which is for certain is that the way we live now is definitely unsustainable, and that things will need to change in what by todays standards are fairly substantive ways if there is to be any longer term future for human civilization.
Zeitgeist also provides a much more optimistic counterpoint to the the frequently gloomy presentations about the future made by Singularitarians, who don't take a systems viewpoint (hence the runaway scenarios) and all too often rely upon scare stories to get their message across.
If you've seen any of the other Zeitgeist movies beyond the first one then you can save yourself quite a lot of time because the latest version doesn't contain much in terms of new concepts, but rather it presents the existing ideas in a slightly more entertaining way. If not, then here is the latest one.
In my estimation, I'm sympathetic towards the resource based economy idea, and taking a systems approach to running things. In the long term actually I don't think there's going to be very much alternative to managing economies as integrated machine-managed systems, as opposed to the highly unsustainable and periodically dysfunctional way in which things work currently. The depiction of future cities is nice, but there's also a vast amount of hand-waving here (practically flapping) and not very much depth to the grand visions.
A world without money
One of the major themes of Zeitgeist is the possibility of a world without money, where nanotechnology - or some other futuristic wizardry, perhaps involving robots - can synthesise and recycle products on demand. For many sorts of consumer products this may be quite feasible, and it's not too difficult to imagine bigger and better RepRaps combined with more traditional forms of industrial automation. But there may still be competing zero sum interests, and how should these be arbitrated? Suppose citizen X wants a boat and citizen Y wants a car, but the computerised management system indicates that there is not enough resources in the environmental inventory to satisfy both demands simultaneously. There may still be a need for some forms of trade, and where there's trade there is inevitably a need for some kind of abstract representation of value. The non-homogeneous distribution of resources around the planet will surely mean that there will be surpluses of resources in some areas and deficits in others, and traditionally this has been addressed with some kind of trading system.
If money still exists in the distant future, perhaps cryptocurrencies are the way forward, avoiding the issues of financial fraud and centralized banking which currently bedevil our monetary system.
Scientism
My other main criticism is that the depiction of science in the Zeitgeist movies.
Firstly the treatment of the role of genetics and environment seems rather loaded towards a notion that genes don't matter very much. In human society there are numerous phenomena which have little or nothing to do with genes, but claiming that none of our behaviors or biases are in any way regulated by genetics I think is probably not well founded in contemporary biology. For instance, it's quite likely that it is to some degree "genetically determined" that most humans have two legs, and five fingers on each hand. If the morphology of the body is regulated by genes it seems at least plausible that the morphology of its organs, including the brain, are also influenced to some extent by genetics. Our understanding of genetics is still at quite an primitive stage, so there will be much more waiting to be discovered, but it's likely that life is a deviously complex mashup of genetic and environmental interactions, where genetics does have some role in setting up biases and propensities towards certain kinds of behavior or emotional reactions. An example would be that it's believed that some people possess tetrachromatic vision, which gives them slightly different visual sensing capability and maybe different behaviors or behavioral biases become possible as a consequence.
Also in Zeitgeist it's assumed that there is a single "correct" solution to resource management problems. In complex systems such as a society usually there are a range of possible solutions, all with their own accompanying costs and benefits. None of these may be "the best", and choosing one depends upon what direction you wish the society to move in. This may be something which requires decisions to be voted on by citizens, rather than made automatically, otherwise you're putting a computer system in charge of the long term overall evolution of society - which may be asking too much of a system whose main function is resource allocation.
Future Architecture
In the podcast there's some criticism made that the circular cities envisaged aren't easily scalable. This is a reasonable criticism, but not really a show stopper. I can imagine other designs which would scale in a modular way.
If we're thinking about the architecture of the distant future, say a century or more from now, then this is a time when resource shortages of raw materials is going to be ever more acute and an increasingly large fraction of resources will need to be either reusable, recyclable or decomposable (made from organics). Assuming more advanced genetic engineering capabilities it might be possible to have buildings which are partly or entirely organic - made from a wood-like material which can self repair by extracting carbon from the air using sunlight as an energy source to assemble it into a super-strong nanotube like material. Another idea would be to have the walls of buildings contain a soil layer within which the roots of plants could grow. This might mean that you can grow vegetables along the sides of high rise buildings (rather than only in the windows) and have robots periodically harvesting them as part of a localised intensive food production system capable of supporting the appetites of megacities. These green buildings could be fertilised by processed sewage produced by the occupants, reducing the strain on sewer systems which has been a perpetual issue for large conurbations, and I think there are existing "vertical farming" proposals of this kind already. Under this scenario buildings would from a distance have a fuzzy appearance and varying colouration, depending upon what type of vegetation was growing on them. Residents could also get creative with configurable robotic plantation schemes to create artwork, brand logos, slogans or badges of status/allegiance along the sides of their residences. Another, perhaps more likely, situation would be that all buildings are coated with a photovoltaic material which produces electricity.
As for monolithic "totalitarian" style of construction, I agree with the comments made in the podcast about this not reflecting the diversity of human culture. Cities in the future are likely to be at least as diverse as they are now, containing a combination of old and new. A good example of this is the city of York in which I lived for quite a while. It's a 2000 year old city, where you have architecture which is many centuries old standing next to modern office blocks, and a sewer system which was constructed by the Romans in the first few centuries of its existence. It's a rare occasion where cities are built entirely from scratch, and even then all cities evolve over time in a way which is difficult to plan for unless you have a crystal ball which can foresee future transport systems and types of economy (the original roads of York were the width of a chariot, i.e. two horses side by side). I've seen the totalitarian architectural style myself in various cities. Typically it was built in the 1960s and consists primarily of grey concrete blocks as far as the eye can see. It's pretty depressing, and no doubt is a contributory factor to the psychosocial stress of its inhabitants. But people can be more imaginative than this, especially when they're empowered by technology.
Zeitgeist as a meme
On the whole in spite of their failings and vagueness the Zeitgeist movies (at least the later ones) are a well crated attempt to bring ideas from cybernetics into the public arena. I'm not aware of anyone in the mainstream media who even remotely touches upon these sorts of issues. If you're not paying attention then it's easy to adopt a whiggish attitude towards the way that current societies and economies operate, but the only thing which is for certain is that the way we live now is definitely unsustainable, and that things will need to change in what by todays standards are fairly substantive ways if there is to be any longer term future for human civilization.
Zeitgeist also provides a much more optimistic counterpoint to the the frequently gloomy presentations about the future made by Singularitarians, who don't take a systems viewpoint (hence the runaway scenarios) and all too often rely upon scare stories to get their message across.
Freedom Box: A more pragmatic view
"Social networking—that is, the ability to use free form methods of communication from many to many, now, in an instantaneous fashion—changes the balance of power in society away from highly organized vehicles of state control towards people in their own lives."
I generally agree with Eben Moglen about the possible benefits of plug computers, and I'm running such a system myself. In addition to the freedom aspect it should also be emphasized that there are also many pragmatic and economic benefits to running your own server, especially if it's on a low power device which costs a negligible amount of money in electricity to run.
If you expect to get a high amount of web traffic, like Wikipedia or YouTube, then plug computing isn't a good solution, but if you want to run a personal home site to host your own content and blog then it's a decent alternative to the current "cloud" based offerings. For example, if I wanted to host the same amount of data as I'm currently hosting on DropBox then I'd be paying top whack (about $240 per year, or roughly the cost of a netbook), whereas hosted on the plug there are no additional costs. The same savings apply for hosting a web site, wiki or blog - all of which is possible in "the cloud", but which you usually have to pay a never ending subscription free for if you want a non-trivial amount of storage space. It is also possible to pay for a fancy domain name, but if you don't particularly care about that you can obtain free ones from DynDns.
With the first generation of plug computers you definitely needed to be technologically knowledgeable in order to be able to install and administrate it, but the software is getting simpler and much more consumer oriented, as described in this podcast about the TonidoPlug, so within a few years this should be something which is practical for a mainstream user base. As Moglen mentions in his speech there is also the Freedom Box Foundation, which aims to produce a Debian based distro oriented towards plug computer systems.
The typical sort of application for a plug computer would be storing music files, family videos/photos, or hosting a blog or web site. Some of this data you might want to be accessible only to a few people, or for personal use only, whereas if you put it into "the cloud" in an unencrypted form then potentially it could be accessed or copied by others without your knowledge or permission.
In my case I was less concerned about the freedom aspect than with more pragmatic factors, such as maintaining a web presence whilst keeping costs to the lowest possible level and enabling some resilience from the possibility of free hosting services becoming not so free under recession conditions. If I were a student or a soon to be made redundant civil servant who was planning a protest though then the freedom aspect would certainly matter much more to me. In that context freedom means the freedom to not have your cloud hosted web site or blog shut down because someone from the Home Office has made a discreet phone call to someone at Google (or whatever company is hosting the site), or made a bogus copyright violation claim in order to have the site removed. In the case of a copyright violation dispute at least if you're hosting your own data then you have the chance to negotiate with the claimant in a reasonable manner, and make any necessary adjustments, rather than just having your web site instantly removed by the hosting company.
Tuesday, February 15, 2011
A bigger Steamroller
I'm sure that Kurzweil will be making a lot of mileage out of this in subsequent books, but as an AI achievement it's not all that impressive. Noam Chomsky describes it as "a bigger steamroller", and I think I agree.
My knowledge of TV game shows is pretty limited, so I'm unfamiliar with the details of how this game works. It seems to be a case of given an answer you have to guess the question - the reverse of a general knowledge quiz.
If anything this Jeopardy playing computer is a good illustration of what's wrong with what passes for AI, for quite a laundry list of reasons. The first and most obvious is the question of where the system's knowledge comes from. It's claimed that it "learns", and if it did this by watching old TV shows that would be much more exciting since this would mean that it had needed to solve many classically hard AI problems, but scratch the surface and I think you'll discover a whole bunch of men behind curtains manually entering text into databases in a highly selective way. And that's just the beginning of the grievences.
Given a big database of past Jeopardy examples, and some other data sources like Wikipedia or Wordnet which can be used for additional semantics it would be possible to devise a classifier which matches the answers to the questions, or vice versa. You can get as fancy as you like, employing boosted-neural-genetic-fuzzy-probabilistic classification methods of all conceivable types and flavours, and even run them in parallel with a winner takes all arbitration at the end which could itself be optimised, and I expect that's all that this system really is. It doesn't need to have any self model or theory of mind, and in spite of superficial appearances it probably doesn't understand much human language either outside of the realm of Jeopardy questions. Any situations involving construction or interpretation of a narrative, mental imagery, mental rotations, planning or perspective taking will also be off limits.
There's also the snake-oil salesman aspect, which isn't necessarily anything sinister but is a natural consequence of being presented with limited or selective information. If you make gross simplifications - which is completely to be expected in the mainstream media - and vague claims like "It knows the English language" or "It has read Wikipedia" or "It learns new things as it goes" then it's easy for the naive viewer to conclude that this is a much bigger advance in technology than it really is by misinterpreting its limited scope and applying some anthromorphisms (I also expect some Singularitarians to leverage these misunderstandings for their own purposes).
Marvin Minsky's position on the workings of the mind is well known, but there is also an opposing view to this which comes from neuroscience. People like Jeff Hawkins claim that much of the brain - specifically the cortex - consists of very homogeneous electro-mechanics, and that the labels assigned to functionally distinguishable areas are similar to the labels assigned to different muscles in an anatomical diagram. The muscles do different things in different places, but fundamentally the mechanism is the same everywhere. Who is right will eventually become much clearer once the connectome can be analysed.
There are some benefits to the above demonstration though. It could be regarded primarily as a sort of public outreach, which might help to get young people initially interested in computer science or AI, which if Google ngrams can be believed has seen declining public interest since the late 1980s. There is also the obviously self-serving benefit for IBM of getting some prime time TV advertising for their servers or software.
Here's some more video of the computer playing the game.
My knowledge of TV game shows is pretty limited, so I'm unfamiliar with the details of how this game works. It seems to be a case of given an answer you have to guess the question - the reverse of a general knowledge quiz.
If anything this Jeopardy playing computer is a good illustration of what's wrong with what passes for AI, for quite a laundry list of reasons. The first and most obvious is the question of where the system's knowledge comes from. It's claimed that it "learns", and if it did this by watching old TV shows that would be much more exciting since this would mean that it had needed to solve many classically hard AI problems, but scratch the surface and I think you'll discover a whole bunch of men behind curtains manually entering text into databases in a highly selective way. And that's just the beginning of the grievences.
Given a big database of past Jeopardy examples, and some other data sources like Wikipedia or Wordnet which can be used for additional semantics it would be possible to devise a classifier which matches the answers to the questions, or vice versa. You can get as fancy as you like, employing boosted-neural-genetic-fuzzy-probabilistic classification methods of all conceivable types and flavours, and even run them in parallel with a winner takes all arbitration at the end which could itself be optimised, and I expect that's all that this system really is. It doesn't need to have any self model or theory of mind, and in spite of superficial appearances it probably doesn't understand much human language either outside of the realm of Jeopardy questions. Any situations involving construction or interpretation of a narrative, mental imagery, mental rotations, planning or perspective taking will also be off limits.
There's also the snake-oil salesman aspect, which isn't necessarily anything sinister but is a natural consequence of being presented with limited or selective information. If you make gross simplifications - which is completely to be expected in the mainstream media - and vague claims like "It knows the English language" or "It has read Wikipedia" or "It learns new things as it goes" then it's easy for the naive viewer to conclude that this is a much bigger advance in technology than it really is by misinterpreting its limited scope and applying some anthromorphisms (I also expect some Singularitarians to leverage these misunderstandings for their own purposes).
Marvin Minsky's position on the workings of the mind is well known, but there is also an opposing view to this which comes from neuroscience. People like Jeff Hawkins claim that much of the brain - specifically the cortex - consists of very homogeneous electro-mechanics, and that the labels assigned to functionally distinguishable areas are similar to the labels assigned to different muscles in an anatomical diagram. The muscles do different things in different places, but fundamentally the mechanism is the same everywhere. Who is right will eventually become much clearer once the connectome can be analysed.
There are some benefits to the above demonstration though. It could be regarded primarily as a sort of public outreach, which might help to get young people initially interested in computer science or AI, which if Google ngrams can be believed has seen declining public interest since the late 1980s. There is also the obviously self-serving benefit for IBM of getting some prime time TV advertising for their servers or software.
Here's some more video of the computer playing the game.
Sunday, February 13, 2011
Questionable primate biographies
I had an amusing interaction with an employment agent recently, who had asked for a description of what I'd been doing over the last year or so. I replied in an email, giving what I thought was a fairly lucid account of my recent activities and projects. The agent replied saying that this description was "insufficiently business-like", and that I needed to inject certain fictional elements into my account "to make it more credible". The more credible account is one where I have a much higher degree of personal freedom and agency than is really the case - a version of life with the ego dial turned up to eleven.
This little episode highlighted to me the desire of people to create and dwell within a semi-fictitious narrative, and that often a good story is preferred to an accurate story. We all do it of course, but some autobiographies can be more fictitious - or more consciously fictitious - than others.
A recognition of this kind of phenomena goes back a long way, but a nice recent example of the desire to live within a particular narrative, and the various deceptions and absurdities which result from that, is given in the story of Cheeta the chimp. The Cheeta story is amusing on several levels. It highlights the plight of animals living in the human world, the desire of humans to both deceive others and be deceived, the concept of nostalgia as a saleable commodity, the extreme lengths which humans will go to in order to obtain and assemble a storyline and the anthropomorphisation of the lives of non-human creatures.
Also this narrative generating ability - even if it has only the most capricious relationship to the data - I think will turn out to be a central feature of the construction of an artificial intelligence. The ability to answer the question "what did you do yesterday?" in a coherent manner, and to have an ongoing personal history is something which AI systems to date don't seem to have addressed, and certainly I've seen nothing along those lines in robotics so far. It's about assembling a mind, and then being able to have a theory of other minds too which is roughly based upon yours.
This little episode highlighted to me the desire of people to create and dwell within a semi-fictitious narrative, and that often a good story is preferred to an accurate story. We all do it of course, but some autobiographies can be more fictitious - or more consciously fictitious - than others.
A recognition of this kind of phenomena goes back a long way, but a nice recent example of the desire to live within a particular narrative, and the various deceptions and absurdities which result from that, is given in the story of Cheeta the chimp. The Cheeta story is amusing on several levels. It highlights the plight of animals living in the human world, the desire of humans to both deceive others and be deceived, the concept of nostalgia as a saleable commodity, the extreme lengths which humans will go to in order to obtain and assemble a storyline and the anthropomorphisation of the lives of non-human creatures.
Also this narrative generating ability - even if it has only the most capricious relationship to the data - I think will turn out to be a central feature of the construction of an artificial intelligence. The ability to answer the question "what did you do yesterday?" in a coherent manner, and to have an ongoing personal history is something which AI systems to date don't seem to have addressed, and certainly I've seen nothing along those lines in robotics so far. It's about assembling a mind, and then being able to have a theory of other minds too which is roughly based upon yours.
Friday, February 11, 2011
Telerobotic shopping
I expect that sights like these are going to become much more commonplace over the next 5-10 years. This is an outrageously expensive telerobot, but fundamentally there's no reason why robots like this couldn't be manufactured far more cheaply, since really it's just some motors and a motor controller, a battery and something equivalent to a mobile phone. Any location where you can broadcast video from a mobile phone is a place where a telerobot could potentially be used.
Tuesday, February 08, 2011
Copying books
This is all fine in the world as it exists now, but I expect that there will be a time at which paper books become obsolete. There are lots of issues around DRM and the practicality or otherwise of ebook readers, but it seems likely that at some point in the future - and who knows when that will be - these issues will be resolved to some mutually acceptable extent. In that world of paperless offices the business model described above breaks down, because you're no longer shipping atoms and your literary wares consist only of pure information.
Another view would be that people enjoy having books as physical items, like ornaments, and that physical objects containing or representing information will always exist.
A possible future business model for authors in a bookless, or almost bookless, world might be a charitable one. So at the end of the book it might say something like:
If you enjoyed this book please don't forget to donate to the author.
or
Sponsor the sequel by donating to...
in the classic cliff-hanger style. If a sufficient number of people liked what they'd read, and wanted to read more from the same author, then it seems quite likely that this approach might work. Without the physical atoms, publishing middle men and negligible distribution cost popular authors might still be able to make a reasonable living.
Microsoft Trained Brain Syndrome
I've seen all too many sufferers of Microsoft Trained Brain Syndrome, but just like any other addiction there is hope that with effort rehabilitation is possible. I am living proof that MTBS is not a fatal condition. For almost the entirety of the 1990s, with the exception of using Amiga OS for the first few years of that decade, I was inside of the Microsoft MSDOS and Windows bubble. I'd met people who knew something about Unix, but that seemed very foreign, and to have had any experience of Unix meant that you had to have either been a computer scientist or worked for a large corporation. In contrast I'd grown up with 8 bit home computers, and just been familiar with whatever software happened to predominate in the consumer computing marketplace. Also in the UK Apple hardware was not common in the 1980s and 90s, since it was typically far more expensive than other kinds of home computer.
It was only in the early 2000s that I began to entertain the possibility that there might be a world outside of the Microsoft ecosphere. The main factors encouraging me to explore this terra incognita were things like:
It was only in the early 2000s that I began to entertain the possibility that there might be a world outside of the Microsoft ecosphere. The main factors encouraging me to explore this terra incognita were things like:
- Software upgrades becoming less about improved technology and more about Microsoft wanting you to buy the latest version with a certain amount of arm twisting.
- Trying to find ways to avoid having to waste time dealing with malware, and the increasingly unpleasant consequences of installing anti-virus and anti-spyware software.
- Looking for cheaper solutions to the same problems (like replacements for Outlook).
- Microsoft abandoning older versions of their software, which then became unusable due to security vulnerabilities.
- The cost of Microsoft software continuing to escalate, despite an increasingly large market and economies of scale which would normally push prices down. This did change for development tools in recent years, with the Express versions becoming free, but by then I was already outside of the Microsoft box.
- Irritation with the cognitive dissonance associated with ubiquitous pirated software and Windows Genuine Advantage. Having to tell people that they should buy a copy of Windows which cost more than the (sometimes second hand) computer they were using.
- Microsoft's refusal to do anything with Internet Explorer, having won the "browser wars".
- Annoyance that Microsoft wasn't dealing with fundamental issues, such as the OS slowing down over time and requiring re-installations.
Saturday, February 05, 2011
Friday, February 04, 2011
William Hague under cyber attack
William Hague alerts the nation to the presence of cyber-reds under the Foreign Office bed, but if this article is anything to go by it just sounds as if the government IT systems aren't very secure and that they suffered from pretty ordinary malware and phishing attacks.
Here is the Zeus malware which Hague is referring to.
It does look quite nasty, and I'm so glad that I left most of these security woes behind by moving to Linux. I'd suggest that the Foreign Office do the same, and they might also save some money in the process, as well as increasing the security of the realm.
I'm not Bruce Schneier, but my understanding is that the problem of botnets is primarily a problem of the prevalence of older versions of Microsoft Windows using the notoriously insecure Internet Explorer browser and where the user logs in with administrator access by default. Conduct an IT audit and upgrade or replace the relevant systems, and this is a problem which you can solve without the need for creating a whole new government or military department populated by iPad-tapping cyber generals.
Here is the Zeus malware which Hague is referring to.
It does look quite nasty, and I'm so glad that I left most of these security woes behind by moving to Linux. I'd suggest that the Foreign Office do the same, and they might also save some money in the process, as well as increasing the security of the realm.
"We're witnessing an exponential rise in the misuse of cyberspace. In a decade, this could be out of control, and we have to start to do the thinking now."Five or more years ago I might have agreed with this opinion, with the tide of spam and computer viruses being seemingly unstoppable, but now I think this is just alarmism possibly disguising an agenda to create yet another quango.
"General Sir David Richards, chief of the defence staff, last month said the UK needed its own Cyber Command, similar to that set by by the US defence department."Modern software and good IT support: yes. A "cyber command"? Probably just an excuse to employ a few more of your friends and relations on exorbitant consultancy fees at taxpayers expense, and give them smart looking quasi-military uniforms.
I'm not Bruce Schneier, but my understanding is that the problem of botnets is primarily a problem of the prevalence of older versions of Microsoft Windows using the notoriously insecure Internet Explorer browser and where the user logs in with administrator access by default. Conduct an IT audit and upgrade or replace the relevant systems, and this is a problem which you can solve without the need for creating a whole new government or military department populated by iPad-tapping cyber generals.
A Failure of Teleoperation
There are worse examples of robotic bomb disposal, where the robot actually gets blown up, but this snafu illustrates the limitations of robotic perception and manipulation as it currently exists.
It looks as if there's a camera on a pan and tilt mechanism on top of a pole, and a second camera on the arm, but no cameras which give a good view around the wheels/tracks or under the gripper. Also, non-rigid objects like paper bags are not so easy to handle with a gripper.
A solution to the visibility problem might be just to attach an omnidirectional camera on the front of the robot below the base of the arm, which would enable the operator to see under the arm and also on the ground around the front wheels. Some unwarping of the image might be necessary to provide the operator with a more intelligible display, but that could be done quite easily in software.
It looks as if there's a camera on a pan and tilt mechanism on top of a pole, and a second camera on the arm, but no cameras which give a good view around the wheels/tracks or under the gripper. Also, non-rigid objects like paper bags are not so easy to handle with a gripper.
A solution to the visibility problem might be just to attach an omnidirectional camera on the front of the robot below the base of the arm, which would enable the operator to see under the arm and also on the ground around the front wheels. Some unwarping of the image might be necessary to provide the operator with a more intelligible display, but that could be done quite easily in software.
Thursday, February 03, 2011
ROS for the rest of us
A talk by Patrick Goebel of Pi Robot about ROS. At this point ROS is sufficiently commonly used that I expect that it's the way that things are likely to go, at least within the academic and hobby domains, within the next few years. It's also possible that commercial products could be based upon ROS (with some caveats for the patented feature detectors within OpenCV), since it's under a very permissive BSD license.
So far I've written a few things for ROS. There are some drivers for Phidgets devices, a driver for a USB to I2C converter and also a driver for use with a stereo camera and v4l2stereo.
So far I've written a few things for ROS. There are some drivers for Phidgets devices, a driver for a USB to I2C converter and also a driver for use with a stereo camera and v4l2stereo.
Tuesday, February 01, 2011
An interview question
At a job interview earlier today I think I flunked one of the Human Resources questions, but it was at least a question with an interesting background to it. The question was as follows:
"At [company X] what do you think are three attributes which your line manager would describe you as being?"And I said something like:
"Quiet. Analytical. Easy-going."They then later asked:
"What are three attributes which you would use to describe yourself?"And I said the same. In hindsight that maybe wasn't the best answer, and instead I should have said something similar but non-identical. The combination of these two questions are probably about an assessment of perspective taking, or what is known in psychology as theory of mind. In simple terms, whether you can put yourself into someone else's shoes and imagine the world from a point of view which may be slightly different to your own, perhaps also with a different status relationship. I expect that it will turn out that many social animals (non-human primates, cats, dogs, elephants, perhaps even cetaceans) have some degree of TOM capability, even if it's of a quite rudimentary nature compared to our own.
Subscribe to:
Posts (Atom)
