there's a classic paper by kenneth colby (citation below), in which he programmed an eliza-style chatterbot that tried to detect positive and negative undertones (positive attitude, agreement, interest, disinterest, humiliation, malevolence, etc.) in the person's text. those characteristics would then change the state of the bot's simple 'emotion system' (really simple, with clear internal variables for things like mistrust, fear, anger, etc.) which in turn would modulate the output - so the bot would appear withdrawn, angry, happy, and so on.
he then did a sort-of turing test where psychologists were supposed to differentiate conversation logs with the bot from conversation logs with real paranoid patients. and the beautiful thing was - the bot was a more convincing paranoid patient than the paranoid patients themselves!
the problem, of course, is that it's easy to write a paranoid bot, because paranoia is a very convenient excuse for not always necessarily making sense.:)
reference: --- COLBY K.M. (1973) Simulations of Belief Systems. In: SCHANK R., COLBY K.M. (Eds.), Computer Models of Thought and Language. S. Francisco, Freeman. ---
re each small candle - looks like he'll have a song in something like an amnesty international compilation. pretty cool, though he's done that before.
on the other hand, as admitted before (on the web site and in the new york times), he's been working on an opera (!!!) called ca ira for quite a while. it's about the french revolution, and planned to be released somewhere around fall 2000.
not to mention of course the recent tour of the u.s. (anyone else seen the chicago show?), and the planned 2000 tour...
part two - critique.
re waters' lyrics - i think they've grown quite a bit. if that means the loss of quasi-intellectual vagueness, so be it. it's for the better.
just look at the final cut - musically it's a poor brother of the wall, but the lyrics carry so much more significance! it's no longer a bitter semi-autobiographical satire of the society, but an powerful political commentary. the lyrics have been given greater prominence than in previous albums, and while some people may bitch about it, i consider it a big plus.
(then of course there's radio k.a.o.s. yeah, i know. it sucked. the story was okay, but the delivery - just the fact that he had to write out the synopsis before the lyrics started should've been a warning sign. oh well. there's one in every phonography.:)
pros and cons of hitchhiking, on the other hand, is sheer brilliance. the leitmotif of the whole album - hitchhiking as a metaphore for relationships, and travel as a way of exploring the world and exploring yourself - works incredibly well with the lyrics. the examination of painful ways in which people treat each other is especially compelling. btw, this might be the most verbose album in the history of rock. but it works very well.
and finally, amused to death. it's definitely not a pop album. it stands against consummerism, against war, against militarism, and against unexamined patriotism (which can be a medicine especially difficult to swallow here in the u.s.). the problem is, it exposes the audience (and the society at large) as a mindless mob, following the path of convenience and thoughtlessness, unaware that it only leads towards self-destruction. but it's not a message that anyone would like. nobody wants to be told they're not as hot as they think they are. [1]
but i'll admit readily, these last two are not easy albums. almost anyone can pick up dark side and connect with the music - because the lyrics were designed to speak of simple facts of life using simple words. [2] but the same is not true of amused, pros and cons, or even the final cut. those require not just listening, but reading the lyrics, and working to understand them. less like space rock, more like literature.
---
[1] btw, in regard to this, the album goes very much in the spirit of the original neil postman book, but unlike the book, it's quite coherent.:) [2] see schaffer's a saucerful of secrets - the pink floyd odyssey for a great account of the creation of dark side (not to mention the remainder of the history of the band).
the plot twist, characters, and dialogue (especially dialogue) were excellent, the numerous pokes at the catholic church were jaded but quite funny, but i wish kevin smith just stayed with the ace he had up his sleeve. instead, when he was done tearing the dogma apart, for some reason he decided to replace it with his own vision of what god and spirituality are all about. and the problem is - his version is unbelievably butchered.
let's take god, for instance. one moment god is wrathful and proud, the next moment she is benevolent and curious about the world she created; one moment she is bound by rules of human logic, another moment she exercises omnipotence; one moment she is the god of the old testament, the next moment of the new testament, and yet at others a very pantheist omnibeing permeating all of existence. never mind further consistency problems with other members of his pantheon (e.g. loki was a nordic god, not a biblical entity).
i really wish he hadn't done that. ideas like god having to obey human church's laws about atonement of sins are really quite funny, but how can you present a self-consistent version of spirituality based on that? you just can't. not even if you're kevin smith.
it always amazes me why people declare berners-lee to be a 'genius' or otherwise brilliant person for creating the basis of world wide web.
the sad thing is, http is neither the only, nor even the first hypertext protocol (anyone remember gopher?) - it just happened to be the one that people started using, thanks as much (if not more!) to the availability of client software such as ncsa mosaic. the ingenuity of the web comes from the ways in which people use it, not from the protocols on which it's built.
there already are codes of ethics and professional conduct in this field. see, for example, the acm code of ethics for computing professionals (acm being the association for computing machinery). i'm sure that ieee has one as well.
the only problem is the industry at large ignores them, and it's unlikely that anything short of legislative intervention will change that...
Even this little robot cylinder thing doesn't do more than navigate a path through a virtual field. Nothing most slashdot readers couldn't code in under an hour. What ever happened to breathroughs in AI? Why are the serious researchers leaving it for other fields? Should we forget the dream?
cye is not intended as a general ai solution. i saw the cye base in action at this year's aaai conference, and my understanding that the base was a proof of solution to a traditionally difficult problem in robotics - that of dead-reckoning (estimating your position based only on measurement of previous movement).
dead-reckoning is difficult because carpeted floors, which are the default for most office environments, wreak havoc on traditional wheel-and-shaft-encoder combinations - that is because the fibers in the carpet introduce very tiny amounts of variation into robot's movement, but those variations add up really really fast. the insight in cye is the wheel design, which helps solve that problem (i don't remember the encoders or brains being anything out of the ordinary, though).
btw, regarding "nothing that most slashdotters couldn't code up in an hour" - for dead-reckoning code that may be true. but i wouldn't extrapolate to other problems in robotics. i've seen many highly competent programmers getting their egos crushed by the sheer difficulty of interfacing a machine to the real world. it's a very difficult, and sadly largely underconstrained problem.
r
ps. and regarding the sad state of ai in entertainment - the times, they are a-changin'. check the july issue of game developer magazine (treebased medium, unfortunately) for a cool article on the state of ai in games, which also includes a few notes on the work between the entertainment industry and the academia.
first of all, artificial intelligence is not a theoretical field, but a science. unlike in theory fields (math, theory of cs), you can't just expect everything to be neat and clear and derivable from first principles. just remember how many centuries (millenia?) it took to come up with a reasonable model of the atom. we can't expect a reasonable model of the mind to just pop up overnight. for every brilliant insight there is a dozen detours. such is the way of science.
and secondly, re your comment of ai being an ill-defined field - it's not the field that's ill-defined, it's the definition of intelligence that keeps changing on us! back in the 50s, when first ai systems were born, people actually considered intelligence to be equivalent to formal inference, spatial reasonoing, and so on. but as computers started getting good at those, the definition kept changing, as if to exclude what computers were doing - people started realizing: what about emotions, what about social skills, what about pragmatics? but this is a vicious epistemological circle - ai trying to model intelligence which is constantly being redefined because of ai's successes. to blame ai for this circle would be as foolish as blaming mathematics for people's fear of differential equations. the question should be how to break it.
on one hand i agree that robo soccer is somewhat of a toy domain, and that work done on that problem may not scale up to real-life situations. but on the other, i think your extrapolation of the state of ai from that one project is a bit short-sighted.
i can completely understand your frustration with the lego kits. when the robot's perception consists of only touch sensors, IRs, and maybe a sonar or two, it's going to hit a performance wall really early. the simple fact is, there's only so much you can do with such limited perception. in order to do anything of research quality, you will need much more than that - real-time vision and a ring of sonars, for example. not to mention you'll need to endow the robot with some reasoning about the environment, which is rather difficult with the C/assembly tools that are used for microcontroller programming.
lego kits are fine as an introduction to robotics, and to disillusion people - that sensors are really noisy, that the environment rarely follows your assumptions about it, that you can't trust your perception, and that to duplicate even the most simple of human behaviors you need much more cognitive processing than basic systems provide. but robotics research, while facing more difficult problems, also doesn't work under the constraints of lego kits. consider the systems that are in operation right now, like museum tour-guide robots in bonn, the mars pathfinder project, and delivery bots at many hospitals and research facilities. it's a world of difference.
so it's about building hardware-based parallelized neural networks. nothing to get excited about, this stuff has been around for a while. especially the wetware version.:)
btw, even though connectionist approaches (such as neural networks) work beautifully on problems where tight coupling of perception and action is necessary, such as 'teaching' a hand to grab an object, i wouldn't bet on making complete intelligent systems out of them. neural networks are notorious for being "black boxes" that are difficult to engineer for complex situations, because they provide no information to the designer about why they're doing what they're doing. imagine writing a program in a language you barely know, and without a debugger.:)
yeah. lisp forever. or something.:)
ps. "synthetic epistemology"? oh great, like we're not having enough trouble with natural epistemology...:)
the magazine certainly was read and enjoyed by techies. the technolust section was the bomb, featuring every gadget one could possibly imagine. the political articles, while heavier on cyberhappy rhetoric than actual content, provided a great introduction to the complex interactions between the dreams of cyberspace and the laws of politics. and they were always eager to talk of the new cool thing at the media lab.:)
but it wasn't the visionary magazine some think it was. the articles, while skillfully written and even more wonderfully illustrated, with funky, fluorescent lettering on high-quality paper, were still written by journalists who wrote of things they knew little of. and it showed. it showed through detailed speculations of possible future impacts of technologies that contained no interesting details of how they work in the present. it showed in clueless articles on science and pseudoscience (the article about supposed discovery of antigravity especially stuck in my throat). but more importantly, it showed in the cult of celebrity (or "cyber-celebrity", later replaced by "making-money-on-the-'net-celebrity") that filled the magazine. it's easier to write about a person than about the work they do, the latter would require grokking the technical.
wired had always been enjoyable. but not enjoyable in a "wow, this is just what i needed, i think i'm gonna start working on this" sort of a way. there was never enough detail for that. it was rather inspiring in a "wow, this is a neat toy, if i ever get rich i'm gonna get me one of these" way, or even in a "this story isn't quite right, but it's good that someone's telling these stories to the masses" way. or at least they did for a while, until they got replaced a year ago with the stories of the internet as el dorado.
wired, you served your role. you've been an interesting read. you've been a people magazine that aspired to be the economist. but my subscription ends in december, and i'm not going to renew.
i think the article phrased the problem in a somewhat confusing way, which resulted in the flurry of comments about the artwork aspects of game design and how they should be opened, whatever that might mean. but that's not necessarily a good way of thinking about it.
open source flourished by providing open infrastructure for building complex projects, from networking tools and standards to good programming tools. it may sound obvious, but it seems the same should happen here. there's no need for open games with public domain artwork and such, even just the notion is absurd.
but there is a need for infrastructure. for example, an open source engine, with sophisticated networking (minimal packet size, massive multiplayer support, etc), a couple pluggable graphics engines (for 3d shooters, orthogonal rts, etc.), and hooks for plugins for ai and such.
people will come to this project if it frees them from rewriting the same engine over and over again, and lets concentrate on the aspects that need to be customized - such as graphics.
now how to make such an engine that's general across genres and still efficient is a whole different topic...:)
First (pedantic), `alife' is the _wrong_ term to use for modular/threaded programming techniques (Brooks' subsumption architechture etc.) The former refers to virus simulation/evolution, the latter being a methodology inspired by the modular nature of the brain.
you're right, but genetic programming is precisely what game programmers are usually doing - they use (or plan on using) some sort of evolutionary technique to breed a 'perfect' behavior algorithm (which probably means breeding a perfect finite state machine to control the creature).
frankly, most of computer game ai is pretty simple - it's mostly finite state machines for behavior control, sometimes augmented with 'fuzzy' (i.e. nondeterministic) state progression. there are some attempts at doing more that that - afaik, crash bandicoot and jedi knight both use subsumption for behavior control - but those are few and far between.
take a look at the telemetry log - it contains lines such as:
;; [:EXEC-ACT:INTERESTING 43645139.016];; Simulating NEB1 status throw failure;;
this is not only lisp/scheme syntax, but the colon-prefix notation is typical lisp. it's also one of the ways one could represent phenomena in lisp-based inference engines.
granted, the evidence is circumstantial, but fits lisp better than any other language!:)
Is this really an AI? Aren't there a set of laws that define what is truly an AI and what is just Agent software [...]?
oh, but it is ai. this system uses a full inference engine - formal inference being one of the 'classic areas' in ai research.
here's what happens. you give the computer information about the components of the system (in this case, all the parts of the proble that it needs to know about: probably thrusters, sensors, etc.), information about what inputs/outputs these components handle, and what are the effects of those components on other components. these details should be numerous, but fairly simple - after all, one thingy can be causally directly connected to only so many other thingies.
and then once you have this network in place, you can tell the machine to achieve some goal - for example, once it reaches one a.u. from the earth it should take photos of the earth every hour.
the computer will now perform 'inference' - grovel through its network of dependencies, find all conditions that need to be satisfied for the task to be successful (open camera lens, etc.:), will satisfy them, and perform the task. furthermore, if equipment fails (as it invariably does), the internal network will get updated with the information that the goal is not achievable because of some X, and the engine will find some other way of achieving or maintaining the goal state.
judging from the telemetry log, i'm pretty sure that's the type of engine they use. high intelligence it ain't, but hey, it's better than doing everything manually the way nasa used to do it... but the term 'agent' is really ill-fitted to this application (not the least because 'agent' is one of the most widely abused terms in computer science today) - so the whole probe may be an autonomous agent in a sense that it has a concept of 'survival goals' and takes actions to make sure they are achieved, but it would be much clearer to call it an inference engine...
So unfortunately, teaching it to compile your kernel is out.
oh really?:)
don't know whether this applies to the commercial version, but our robotics group at northwestern implemented the scheme language for the dogs, and our pets are now fully programmable. you can even have them crunch equations for you in their spare time.:) (yeah, yeah, i heard all about beowoof clusters. stop that!:)
btw, the dogs make a really cool robotics research platform. if you're in orlando this summer, i'd check out the aaai conference - several universities have had the dogs for a while now, and i hope there will be some interesting applications of the dogs to Real Robotics Research (tm) exhibited at the conference...
i think the premise of this article is just flawed. censorship implies limiting communication because the message is not to the censor's liking. now, wb canceled the finale because it revolved around a high school massacre. but wb couldn't care less about whether massacres are or are not to be broadcasted. that sounds not like censorship, but more like trying not to be blatantly distasteful in the eyes of the general public.
wait a minute... what is their legal basis for persecution? at least in u.s. (and ipix seems to be an american company) you can't just go after people because they're implementing your algorithms. that is, if your invention is not patented, it's up for grabs. afaik, if they have just a u.s. patent, they still can't force a german developer to honor it. only if they have a patent in germany as well, they could force dersch to pay royalties. does anyone know what basis they have for making those threats?
yes, i've been captivated by this for a while, too. each group defines 'social skills' around what feels normal for them, and since the group that has the necessary social skills to be in leadership positions differs diametrically from the ultrageek group, they naturally find their manners to be utterly incomprehensible, and label them antisocial. but that's only to be expected - don't ultrageeks feel the same way towards suits and those of their kind?
the same goes for communicationsfolk (journalists and such), who usually don't feel comfortable with either suits or geeks - and who get the same reaction from them. there's a reason why popular media regard geeks as ridiculous, and geeks regard the entertainment industry as composed of overblown, overpaid idiots. going further, this scales both up (to countries and large social groups) and down (to individual social circles). to define oneself by differences from others is only a natural human tendency.
that's why when some halfwit says "X doesn't have any social skills", it has to be understood as symbolic of "in the context of the societies in which i participate and the norms which i hold to be right because they are present for my community, X doesn't have any social skills.":)
How do you explain that money doesn't enumerate all value?
oh, but there is correlation! if people buy something, that means it has value to them!
but that's not the main point. the problem is not just schools not spending much, but also the industry not providing much of interest.
case in point, i recently talked with an old high school teacher of mine, and at one point he started telling me about the college-level calculus courses they're planning, and the difficulties they have finding the right software. his experience was that the market is pretty much split into silly puzzle-style programs on one end, and mathematica on the other - nothing in between. college freshmen trying to find good tutoring software to help them with calc can relate, i'm sure. no wonder people don't buy programs, if there's little available.
and secondly, regarding the poor quality of school software - i've worked for a while in one of the big public school systems, and educational software costs a lot. depending on software, it can get into upper double- and triple-digits per machine for a specialized program (times several dozen machines per lab). on one hand it's understandable, because a limited-market title is naturally going to be more expensive to offset production costs - but if you're an educator faced with a choice of spending your entire yearly software budget on one or two specialized titles, or several types of general productivity and math/statistics software, which one will you choose? considering present pricing and selection, no wonder we have curricula based around m$ office and mathcad.
which is not to say educators are not to blame. all too often they mistake typing and ms office classes for 'computer science education' - a distinction which i'm afraid is also often lost on college freshmen trying pick a major. labs are often run by people without much it training, who learn as they go. but that is more excusable. teachers are primarily educators, and shouldn't be expected to be necessarily computer savvy. somewhere higher up in the administration there should be people who keep up with technology, evaluate educational software, and advise the teachers, but we've still a way to go until then.
which brings us right back to not having good software to begin with.:)
quantum model of the internet :)
on
Infinite Space
·
· Score: 2
Maybe the net will grow and grow like the real universe until it reaches a limit and all the gravitational pull makes it collapse upon itself.:)
it's all because of the exchange of virtual particles called attenti-ons. when a web site comes into the vicinity of another website in the cyberspace, tiny virtual 'particles' of attention (but they're not really just attention particles, remember the medium-message duality?) that are being constantly generated and annihilated within the website finally come close enough to each other that instead of staying inside one web site, they swap places, and a large number of such brief exchanges results in an attraction between the two web sites that is stronger than charge-based repulsion. hence the creation of multi-site nuclei (i.e. portals).
this is difficult to observe, because the attenti-ons are so tiny that they cannot be observed by bombardment with our statistical photons, but the results of their interactions can be observed by formation of such macroscopic structures as 'URL links' - which are really just models for exchange of attenti-ons.
it's also quite interesting to observe when a tiny web site traveling at a very high hype velocity collides with a large but stationary site - the results, such as slashdotting, can be quite spectacular, including a release of a large number of attenti-ons, and sometimes even flame-ons. but that's a whole different story.:)
r now where's that quark model of attention i've been working on?:)
one comment about what's been said about the futility of getting a ph.d.:
people who get ph.d. don't care about earning less money - matter of fact, when you count in all the raises, promotions, etc. they'd get during that time, they lose many hundred thousand. people get ph.d. for the knowledge, and to work on problems that have never been attempted before.
grad students don't want to be like gates or jobs or torvalds. they want to be ritchie and knuth and minsky. and there's a world of difference.
CS comes in a couple different flavors, depending on whether the university's CS department came out of math or EE.
oh, there's at least one more approach - for the lack of a better term, i'd call it programming not as math or electronics, but as communication.
starts off by asking the students - given that you want to express something, what's the best way of expressing it? for example, what's the best way of expressing a factorial? iteration? recursion? why? and what about a fibonacci? why is recursion bad for a fibonacci? so now you have two of those expressions - what's the best way of keeping them together? what if both of them have some common functionality? and so on and so forth.
this way advanced concepts are introduced right off the bat without worrying about quirks of low-level languages or mathematical correctness, and students will cover many data structures, abstraction issues, complexity, and in many cases will write a simple compiler or a symbolic reasoning system by the end of the second quarter of their freshman year.
this is certainly the way advocated in the wizard book (structure and interpretation of computer programs, by abelson & sussman), and methinks it works much better than the alternatives.
i think there's a fundamental distinction that needs to be made here.
censorship happens when one person filters content for someone else. that's generally not a good idea, and i agree with parts of katz's posting condemning censorship.
but that's completely different from people filtering content for themselves! that sort of filtering happens all the time, and not just on cognitive level - and i don't see a reason to condemn it. anyone who ever blocked out kiddies cross-flaming on usenet will agree.
of course, an argument could be made that intelligent people shouldn't block out speech just because they don't agree with the message, but that's completely different from saying that filtering speech for one's own use equals censorship.
Slashdot Mirror
there's a classic paper by kenneth colby (citation below), in which he programmed an eliza-style chatterbot that tried to detect positive and negative undertones (positive attitude, agreement, interest, disinterest, humiliation, malevolence, etc.) in the person's text. those characteristics would then change the state of the bot's simple 'emotion system' (really simple, with clear internal variables for things like mistrust, fear, anger, etc.) which in turn would modulate the output - so the bot would appear withdrawn, angry, happy, and so on.
:)
he then did a sort-of turing test where psychologists were supposed to differentiate conversation logs with the bot from conversation logs with real paranoid patients. and the beautiful thing was - the bot was a more convincing paranoid patient than the paranoid patients themselves!
the problem, of course, is that it's easy to write a paranoid bot, because paranoia is a very convenient excuse for not always necessarily making sense.
reference:
---
COLBY K.M. (1973) Simulations of Belief Systems. In: SCHANK R., COLBY K.M. (Eds.), Computer Models of Thought and Language. S. Francisco, Freeman.
---
part one - information.
:)
:)
re each small candle - looks like he'll have a song in something like an amnesty international compilation. pretty cool, though he's done that before.
on the other hand, as admitted before (on the web site and in the new york times), he's been working on an opera (!!!) called ca ira for quite a while. it's about the french revolution, and planned to be released somewhere around fall 2000.
not to mention of course the recent tour of the u.s. (anyone else seen the chicago show?), and the planned 2000 tour...
part two - critique.
re waters' lyrics - i think they've grown quite a bit. if that means the loss of quasi-intellectual vagueness, so be it. it's for the better.
just look at the final cut - musically it's a poor brother of the wall, but the lyrics carry so much more significance! it's no longer a bitter semi-autobiographical satire of the society, but an powerful political commentary. the lyrics have been given greater prominence than in previous albums, and while some people may bitch about it, i consider it a big plus.
(then of course there's radio k.a.o.s. yeah, i know. it sucked. the story was okay, but the delivery - just the fact that he had to write out the synopsis before the lyrics started should've been a warning sign. oh well. there's one in every phonography.
pros and cons of hitchhiking, on the other hand, is sheer brilliance. the leitmotif of the whole album - hitchhiking as a metaphore for relationships, and travel as a way of exploring the world and exploring yourself - works incredibly well with the lyrics. the examination of painful ways in which people treat each other is especially compelling. btw, this might be the most verbose album in the history of rock. but it works very well.
and finally, amused to death. it's definitely not a pop album. it stands against consummerism, against war, against militarism, and against unexamined patriotism (which can be a medicine especially difficult to swallow here in the u.s.). the problem is, it exposes the audience (and the society at large) as a mindless mob, following the path of convenience and thoughtlessness, unaware that it only leads towards self-destruction. but it's not a message that anyone would like. nobody wants to be told they're not as hot as they think they are. [1]
but i'll admit readily, these last two are not easy albums. almost anyone can pick up dark side and connect with the music - because the lyrics were designed to speak of simple facts of life using simple words. [2] but the same is not true of amused, pros and cons, or even the final cut. those require not just listening, but reading the lyrics, and working to understand them. less like space rock, more like literature.
---
[1] btw, in regard to this, the album goes very much in the spirit of the original neil postman book, but unlike the book, it's quite coherent.
[2] see schaffer's a saucerful of secrets - the pink floyd odyssey for a great account of the creation of dark side (not to mention the remainder of the history of the band).
re dogma,
the plot twist, characters, and dialogue (especially dialogue) were excellent, the numerous pokes at the catholic church were jaded but quite funny, but i wish kevin smith just stayed with the ace he had up his sleeve. instead, when he was done tearing the dogma apart, for some reason he decided to replace it with his own vision of what god and spirituality are all about. and the problem is - his version is unbelievably butchered.
let's take god, for instance. one moment god is wrathful and proud, the next moment she is benevolent and curious about the world she created; one moment she is bound by rules of human logic, another moment she exercises omnipotence; one moment she is the god of the old testament, the next moment of the new testament, and yet at others a very pantheist omnibeing permeating all of existence. never mind further consistency problems with other members of his pantheon (e.g. loki was a nordic god, not a biblical entity).
i really wish he hadn't done that. ideas like god having to obey human church's laws about atonement of sins are really quite funny, but how can you present a self-consistent version of spirituality based on that? you just can't. not even if you're kevin smith.
it always amazes me why people declare berners-lee to be a 'genius' or otherwise brilliant person for creating the basis of world wide web.
the sad thing is, http is neither the only, nor even the first hypertext protocol (anyone remember gopher?) - it just happened to be the one that people started using, thanks as much (if not more!) to the availability of client software such as ncsa mosaic. the ingenuity of the web comes from the ways in which people use it, not from the protocols on which it's built.
the only problem is the industry at large ignores them, and it's unlikely that anything short of legislative intervention will change that...
Even this little robot cylinder thing doesn't do more than navigate a path through a virtual field. Nothing most slashdot readers couldn't code in under an hour. What ever happened to breathroughs in AI? Why are the serious researchers leaving it for other fields? Should we forget the dream?
cye is not intended as a general ai solution. i saw the cye base in action at this year's aaai conference, and my understanding that the base was a proof of solution to a traditionally difficult problem in robotics - that of dead-reckoning (estimating your position based only on measurement of previous movement).
dead-reckoning is difficult because carpeted floors, which are the default for most office environments, wreak havoc on traditional wheel-and-shaft-encoder combinations - that is because the fibers in the carpet introduce very tiny amounts of variation into robot's movement, but those variations add up really really fast. the insight in cye is the wheel design, which helps solve that problem (i don't remember the encoders or brains being anything out of the ordinary, though).
btw, regarding "nothing that most slashdotters couldn't code up in an hour" - for dead-reckoning code that may be true. but i wouldn't extrapolate to other problems in robotics. i've seen many highly competent programmers getting their egos crushed by the sheer difficulty of interfacing a machine to the real world. it's a very difficult, and sadly largely underconstrained problem.
r
ps. and regarding the sad state of ai in entertainment - the times, they are a-changin'. check the july issue of game developer magazine (treebased medium, unfortunately) for a cool article on the state of ai in games, which also includes a few notes on the work between the entertainment industry and the academia.
all right, let's start a holy war. :)
first of all, artificial intelligence is not a theoretical field, but a science. unlike in theory fields (math, theory of cs), you can't just expect everything to be neat and clear and derivable from first principles. just remember how many centuries (millenia?) it took to come up with a reasonable model of the atom. we can't expect a reasonable model of the mind to just pop up overnight. for every brilliant insight there is a dozen detours. such is the way of science.
and secondly, re your comment of ai being an ill-defined field - it's not the field that's ill-defined, it's the definition of intelligence that keeps changing on us! back in the 50s, when first ai systems were born, people actually considered intelligence to be equivalent to formal inference, spatial reasonoing, and so on. but as computers started getting good at those, the definition kept changing, as if to exclude what computers were doing - people started realizing: what about emotions, what about social skills, what about pragmatics? but this is a vicious epistemological circle - ai trying to model intelligence which is constantly being redefined because of ai's successes. to blame ai for this circle would be as foolish as blaming mathematics for people's fear of differential equations. the question should be how to break it.
on one hand i agree that robo soccer is somewhat of a toy domain, and that work done on that problem may not scale up to real-life situations. but on the other, i think your extrapolation of the state of ai from that one project is a bit short-sighted.
i can completely understand your frustration with the lego kits. when the robot's perception consists of only touch sensors, IRs, and maybe a sonar or two, it's going to hit a performance wall really early. the simple fact is, there's only so much you can do with such limited perception. in order to do anything of research quality, you will need much more than that - real-time vision and a ring of sonars, for example. not to mention you'll need to endow the robot with some reasoning about the environment, which is rather difficult with the C/assembly tools that are used for microcontroller programming.
lego kits are fine as an introduction to robotics, and to disillusion people - that sensors are really noisy, that the environment rarely follows your assumptions about it, that you can't trust your perception, and that to duplicate even the most simple of human behaviors you need much more cognitive processing than basic systems provide. but robotics research, while facing more difficult problems, also doesn't work under the constraints of lego kits. consider the systems that are in operation right now, like museum tour-guide robots in bonn, the mars pathfinder project, and delivery bots at many hospitals and research facilities. it's a world of difference.
so it's about building hardware-based parallelized neural networks. nothing to get excited about, this stuff has been around for a while. especially the wetware version. :)
:)
:)
:)
btw, even though connectionist approaches (such as neural networks) work beautifully on problems where tight coupling of perception and action is necessary, such as 'teaching' a hand to grab an object, i wouldn't bet on making complete intelligent systems out of them. neural networks are notorious for being "black boxes" that are difficult to engineer for complex situations, because they provide no information to the designer about why they're doing what they're doing. imagine writing a program in a language you barely know, and without a debugger.
yeah. lisp forever. or something.
ps. "synthetic epistemology"? oh great, like we're not having enough trouble with natural epistemology...
it's about time to retire wired.
:)
the magazine certainly was read and enjoyed by techies. the technolust section was the bomb, featuring every gadget one could possibly imagine. the political articles, while heavier on cyberhappy rhetoric than actual content, provided a great introduction to the complex interactions between the dreams of cyberspace and the laws of politics. and they were always eager to talk of the new cool thing at the media lab.
but it wasn't the visionary magazine some think it was. the articles, while skillfully written and even more wonderfully illustrated, with funky, fluorescent lettering on high-quality paper, were still written by journalists who wrote of things they knew little of. and it showed. it showed through detailed speculations of possible future impacts of technologies that contained no interesting details of how they work in the present. it showed in clueless articles on science and pseudoscience (the article about supposed discovery of antigravity especially stuck in my throat). but more importantly, it showed in the cult of celebrity (or "cyber-celebrity", later replaced by "making-money-on-the-'net-celebrity") that filled the magazine. it's easier to write about a person than about the work they do, the latter would require grokking the technical.
wired had always been enjoyable. but not enjoyable in a "wow, this is just what i needed, i think i'm gonna start working on this" sort of a way. there was never enough detail for that. it was rather inspiring in a "wow, this is a neat toy, if i ever get rich i'm gonna get me one of these" way, or even in a "this story isn't quite right, but it's good that someone's telling these stories to the masses" way. or at least they did for a while, until they got replaced a year ago with the stories of the internet as el dorado.
wired, you served your role. you've been an interesting read. you've been a people magazine that aspired to be the economist. but my subscription ends in december, and i'm not going to renew.
i think the article phrased the problem in a somewhat confusing way, which resulted in the flurry of comments about the artwork aspects of game design and how they should be opened, whatever that might mean. but that's not necessarily a good way of thinking about it.
:)
open source flourished by providing open infrastructure for building complex projects, from networking tools and standards to good programming tools. it may sound obvious, but it seems the same should happen here. there's no need for open games with public domain artwork and such, even just the notion is absurd.
but there is a need for infrastructure. for example, an open source engine, with sophisticated networking (minimal packet size, massive multiplayer support, etc), a couple pluggable graphics engines (for 3d shooters, orthogonal rts, etc.), and hooks for plugins for ai and such.
people will come to this project if it frees them from rewriting the same engine over and over again, and lets concentrate on the aspects that need to be customized - such as graphics.
now how to make such an engine that's general across genres and still efficient is a whole different topic...
First (pedantic), `alife' is the _wrong_ term to use for modular/threaded programming techniques (Brooks' subsumption architechture etc.) The former refers to virus simulation/evolution, the latter being a methodology inspired by the modular nature of the brain.
you're right, but genetic programming is precisely what game programmers are usually doing - they use (or plan on using) some sort of evolutionary technique to breed a 'perfect' behavior algorithm (which probably means breeding a perfect finite state machine to control the creature).
frankly, most of computer game ai is pretty simple - it's mostly finite state machines for behavior control, sometimes augmented with 'fuzzy' (i.e. nondeterministic) state progression. there are some attempts at doing more that that - afaik, crash bandicoot and jedi knight both use subsumption for behavior control - but those are few and far between.
btw, for some papers on future directions and game ai research, i'd check out the web page from the 1999 aaai symposium on computer games and artificial intelligence.
Full disclosure: I'm doing contract PR work for Franz. Stories like this are like candy to me. :) --Tom pr@franz.com
and you folks let harlequin grab the deep space project? geez...
:)
It was intelligent enough to reboot the computer when it froze, it obviously could do the work of the MCSE SysAdmins where I work.
;)
but that's provably impossible for a computer - it would imply solving the halting problem!
take a look at the telemetry log - it contains lines such as:
;; [:EXEC-ACT :INTERESTING 43645139.016] ;; Simulating NEB1 status throw failure ;;
:)
this is not only lisp/scheme syntax, but the colon-prefix notation is typical lisp. it's also one of the ways one could represent phenomena in lisp-based inference engines.
granted, the evidence is circumstantial, but fits lisp better than any other language!
Is this really an AI? Aren't there a set of laws that define what is truly an AI and what is just Agent software [...]?
oh, but it is ai. this system uses a full inference engine - formal inference being one of the 'classic areas' in ai research.
here's what happens. you give the computer information about the components of the system (in this case, all the parts of the proble that it needs to know about: probably thrusters, sensors, etc.), information about what inputs/outputs these components handle, and what are the effects of those components on other components. these details should be numerous, but fairly simple - after all, one thingy can be causally directly connected to only so many other thingies.
and then once you have this network in place, you can tell the machine to achieve some goal - for example, once it reaches one a.u. from the earth it should take photos of the earth every hour.
the computer will now perform 'inference' - grovel through its network of dependencies, find all conditions that need to be satisfied for the task to be successful (open camera lens, etc.:), will satisfy them, and perform the task. furthermore, if equipment fails (as it invariably does), the internal network will get updated with the information that the goal is not achievable because of some X, and the engine will find some other way of achieving or maintaining the goal state.
judging from the telemetry log, i'm pretty sure that's the type of engine they use. high intelligence it ain't, but hey, it's better than doing everything manually the way nasa used to do it... but the term 'agent' is really ill-fitted to this application (not the least because 'agent' is one of the most widely abused terms in computer science today) - so the whole probe may be an autonomous agent in a sense that it has a concept of 'survival goals' and takes actions to make sure they are achieved, but it would be much clearer to call it an inference engine...
So unfortunately, teaching it to compile your kernel is out.
:)
:) (yeah, yeah, i heard all about beowoof clusters. stop that! :)
oh really?
don't know whether this applies to the commercial version, but our robotics group at northwestern implemented the scheme language for the dogs, and our pets are now fully programmable. you can even have them crunch equations for you in their spare time.
btw, the dogs make a really cool robotics research platform. if you're in orlando this summer, i'd check out the aaai conference - several universities have had the dogs for a while now, and i hope there will be some interesting applications of the dogs to Real Robotics Research (tm) exhibited at the conference...
i think the premise of this article is just flawed. censorship implies limiting communication because the message is not to the censor's liking. now, wb canceled the finale because it revolved around a high school massacre. but wb couldn't care less about whether massacres are or are not to be broadcasted. that sounds not like censorship, but more like trying not to be blatantly distasteful in the eyes of the general public.
wait a minute... what is their legal basis for persecution? at least in u.s. (and ipix seems to be an american company) you can't just go after people because they're implementing your algorithms. that is, if your invention is not patented, it's up for grabs. afaik, if they have just a u.s. patent, they still can't force a german developer to honor it. only if they have a patent in germany as well, they could force dersch to pay royalties. does anyone know what basis they have for making those threats?
yes, i've been captivated by this for a while, too. each group defines 'social skills' around what feels normal for them, and since the group that has the necessary social skills to be in leadership positions differs diametrically from the ultrageek group, they naturally find their manners to be utterly incomprehensible, and label them antisocial. but that's only to be expected - don't ultrageeks feel the same way towards suits and those of their kind?
:)
the same goes for communicationsfolk (journalists and such), who usually don't feel comfortable with either suits or geeks - and who get the same reaction from them. there's a reason why popular media regard geeks as ridiculous, and geeks regard the entertainment industry as composed of overblown, overpaid idiots. going further, this scales both up (to countries and large social groups) and down (to individual social circles). to define oneself by differences from others is only a natural human tendency.
that's why when some halfwit says "X doesn't have any social skills", it has to be understood as symbolic of "in the context of the societies in which i participate and the norms which i hold to be right because they are present for my community, X doesn't have any social skills."
How do you explain that money doesn't enumerate all value?
:)
oh, but there is correlation! if people buy something, that means it has value to them!
but that's not the main point. the problem is not just schools not spending much, but also the industry not providing much of interest.
case in point, i recently talked with an old high school teacher of mine, and at one point he started telling me about the college-level calculus courses they're planning, and the difficulties they have finding the right software. his experience was that the market is pretty much split into silly puzzle-style programs on one end, and mathematica on the other - nothing in between. college freshmen trying to find good tutoring software to help them with calc can relate, i'm sure. no wonder people don't buy programs, if there's little available.
and secondly, regarding the poor quality of school software - i've worked for a while in one of the big public school systems, and educational software costs a lot. depending on software, it can get into upper double- and triple-digits per machine for a specialized program (times several dozen machines per lab). on one hand it's understandable, because a limited-market title is naturally going to be more expensive to offset production costs - but if you're an educator faced with a choice of spending your entire yearly software budget on one or two specialized titles, or several types of general productivity and math/statistics software, which one will you choose? considering present pricing and selection, no wonder we have curricula based around m$ office and mathcad.
which is not to say educators are not to blame. all too often they mistake typing and ms office classes for 'computer science education' - a distinction which i'm afraid is also often lost on college freshmen trying pick a major. labs are often run by people without much it training, who learn as they go. but that is more excusable. teachers are primarily educators, and shouldn't be expected to be necessarily computer savvy. somewhere higher up in the administration there should be people who keep up with technology, evaluate educational software, and advise the teachers, but we've still a way to go until then.
which brings us right back to not having good software to begin with.
Maybe the net will grow and grow like the real universe until it reaches a limit and all the gravitational pull makes it collapse upon itself. :)
:)
:)
it's all because of the exchange of virtual particles called attenti-ons. when a web site comes into the vicinity of another website in the cyberspace, tiny virtual 'particles' of attention (but they're not really just attention particles, remember the medium-message duality?) that are being constantly generated and annihilated within the website finally come close enough to each other that instead of staying inside one web site, they swap places, and a large number of such brief exchanges results in an attraction between the two web sites that is stronger than charge-based repulsion. hence the creation of multi-site nuclei (i.e. portals).
this is difficult to observe, because the attenti-ons are so tiny that they cannot be observed by bombardment with our statistical photons, but the results of their interactions can be observed by formation of such macroscopic structures as 'URL links' - which are really just models for exchange of attenti-ons.
it's also quite interesting to observe when a tiny web site traveling at a very high hype velocity collides with a large but stationary site - the results, such as slashdotting, can be quite spectacular, including a release of a large number of attenti-ons, and sometimes even flame-ons. but that's a whole different story.
r
now where's that quark model of attention i've been working on?
one comment about what's been said about the futility of getting a ph.d.:
people who get ph.d. don't care about earning less money - matter of fact, when you count in all the raises, promotions, etc. they'd get during that time, they lose many hundred thousand. people get ph.d. for the knowledge, and to work on problems that have never been attempted before.
grad students don't want to be like gates or jobs or torvalds. they want to be ritchie and knuth and minsky. and there's a world of difference.
CS comes in a couple different flavors, depending on whether the university's CS department came out of math or EE.
oh, there's at least one more approach - for the lack of a better term, i'd call it programming not as math or electronics, but as communication.
starts off by asking the students - given that you want to express something, what's the best way of expressing it? for example, what's the best way of expressing a factorial? iteration? recursion? why? and what about a fibonacci? why is recursion bad for a fibonacci? so now you have two of those expressions - what's the best way of keeping them together? what if both of them have some common functionality? and so on and so forth.
this way advanced concepts are introduced right off the bat without worrying about quirks of low-level languages or mathematical correctness, and students will cover many data structures, abstraction issues, complexity, and in many cases will write a simple compiler or a symbolic reasoning system by the end of the second quarter of their freshman year.
this is certainly the way advocated in the wizard book (structure and interpretation of computer programs, by abelson & sussman), and methinks it works much better than the alternatives.
i think there's a fundamental distinction that needs to be made here.
censorship happens when one person filters content for someone else. that's generally not a good idea, and i agree with parts of katz's posting condemning censorship.
but that's completely different from people filtering content for themselves! that sort of filtering happens all the time, and not just on cognitive level - and i don't see a reason to condemn it. anyone who ever blocked out kiddies cross-flaming on usenet will agree.
of course, an argument could be made that intelligent people shouldn't block out speech just because they don't agree with the message, but that's completely different from saying that filtering speech for one's own use equals censorship.