shipped a lot of 10 E4500s from boston to NYC last yr. fedex dropped 3 of them. had smashed CPU heatsinks and dented front panels. 3 x 300,000 = nearly a million bucks worth of damage.
problem is photomechanical processses like this one cause strain and the material breaks in a day. its not (yet) a reusable transfer unless they figure out how to do repairs. could be useful for CDR type applictions though.
The main problem is that chemical rockets dont have the thrust to weight ratios to get into space cheaply. sure..that duct tape and baling wire model might eventually attain sub orbital capability but thats afar cry from lobbing tonnes of material up there to construct lasting structures. what we really need are fusion or fission engines based on cheap, clean thermonuclear reactions. Theres significant engineering to be done to build a robust nuclear engine to get us to space with the required cost and thrust to weight ratios. of course theres the environmental and other hazards (greenpeace supporters disrupting launches? terrorists blowing up nuclear engines?) to consider when you have regular nuclear lifts into space.
the thing is that we dont know the route to take in order to generate machines with emotions. since biological systems are grown together over millenia its VERY hard to reverse engineer a biological system of sufficient complexity.
the only way we can hope to get close is the shotgun approach -- fund tens of thousands of researchers, some of which take advantage of this to build cool toys to play with or are total morons and waste the funds. but the important thing is they try thousands of different approaches and maybe the human race gets lucky with ONE guy who figures it out. and voila -- we get computers which can reason. and all that wasted funding has been worth it.
obviously we may never get a guy who figures it out in which case the money is wasted but its like a lottery ticket where you only have to get lucky once. and when you do its worth it.
hyping AI or selling systems like kismet is one way to get additinal funding for it. that alone and the payoffs if we get lucky make it worth it.
it is everyday life. thats unfortunately a really bad thing since its basically the Cover Your Arse mentality thats dripping into our society. no one can give a concrete answer since bad things happen to those who do -- its CYA at its finest. and not something anyone should be proud of. much less the scientific community.
as a recent dropout from the IT industry into college level teaching i can confirm its true. i normally employ a standard 2 exams + 4 labs for each course when i taught several years ago. i tried doing the same (2+2+4) when i taught a course recently. the labs were ok...no complaints. i normally award full points to completing the labs sucessfully so the students were ok. my exams are usually bloody murder. as usual i put my first exam into the class halfway thru the course. most of the class flunked. of course with the 4 labs they were passing anyway so i wasnt too bothered. what i didnt expect was getting hauled into the office of the Chair and being blasted for introducing examinations which might "traumatize" students and make them sue the college for "emotional duress".
i then had to drop the last exam and make em do a large lab excercise. even with that i still had people fail.
so yeah..its started in colleges already and i expect it to move into universities fairly soon.
AS/400s have built in RAID-5 hardware. they also have redundant PSUs, built in UPS, the whole nine yards. they store data in EBCDIC not ASCII but its trivial to dump data from the built in DB/2 database into a PC since IBM AS/400s typically come with a PC expansion card that boots a virtual PC with windows which can see the filesystem. i have a theory that the midrange (its not a mainframe) somehow killed the process (or batch job) running that operated the front end of their data entry system...or the process died on its own. they probably dont know how to login as QSECOFR and restart the batch process to get their front end back and so they think their system crashed. i doubt its a hardware failure.... those things are bloody reliable and multiply redundant. more likely a crashed job or some other simple software error that makes it appear like the system "crashed" (i.e. not responding).
yeah. i had that and i fixed it. fix was a bitch though..involved hardware. it was on an F-50 dual CPU IBM box with AIX 4.3.x... where are you located ? if you can pay for transport i should be able to fix it.
note that carbon fibre parts are *expensive* and dangerous to work with even though they look nice. always wear a organic respirator and gloves to prevent carbon fibres entering your skin (which they do pretty easily) and the resin from overwhelming you with fumes. molds can typically be made for 400 bucks and the part for another 500-600 bucks. fibreglast.com gives all the info including technical brochures anyway.
hmm...power can probably be brought to these devices the same way millions of living things get it today -- photons. capture enough photons and you have electrical power to power your miniature CPU/robotic manipulator on your MEMS device. just think about your credit card sized solar powered calculators today. they get energy from ambient light..not even solar power...just ambient. if we can get photon collectors down small enough we can power all the whiz bang assemblers we can think of. power distribution can be handled to assemblers who work in shadow by reflecting ambient light from those which collect an excess of it. think small fireflys without chemicals..just reflect photons in random directions and use a small fraction of them. no need to store energy. granted its far off -- but not as far off as teleportation/time travel etc. its a hard (ok..VERY HARD) engineering problem more than anything else.
the NEC versa daylite has a 10.4 inch transreflective LCD. not great but it does 1024x768 and has a 600MHz CPU 128 megs ram, 10 gig hdd etc. here's a review: http://computers.cnet.com/hardware/0-1027-405-6 057 811.html its costly at $2500 or so..more expensive than most laptops with 10 inch LCDs.
yeah but it aint gonna happen. too many clueless PHBs wandering around who equate 9-5 with getting stuff done. cant blame em too..most people love being in a box their whole lives while the average programmer hates it. something about herding cats...
the thing though is that these labs had the financial firepower to do research tightly focused at the commercial market. now the only people doing research are the people in universities who dont have the budget to do the nifty whiz bang gadgets..they do pure research only. which means a lot of university developed research will not be developed into whiz bang gadgets and hence improve the market place or the company stocks. it just means the public wont be able to buy really kewl new technology and companies wont make huge profits selling this technology. the effects are usually felt in 3-5 years or so when AT&T will find itself without any new products to develop since the research base has gone. its a tragic loss. i hope someone can keep up the VNC GPLed tree at least.
humans can learn far more than any AI today. they make decisions by sifting thru vast quantities of rules and data. they have true random number generators built into them. They have vast quantities of processing power which is adaptable. ALL AI systems today cant even compare to ants much less compare to a human brain. AI is a dismal failure today at any rate. its just *hard*.
while i dont disagree that AI will find it difficult to see the big picture, growth between neurons is easy to simulate in a computer program. repetition simulates growth. just wait until enough repeated events occur to form a solid connection. metal activity is unguided. there is no reason to guide a self organizing system based on chaos. it just self organises. does anyone "guide" a tornado forming ? the rules are there, let chaos theory do the rest.
hmm..they have a crappy interface so they need to provide the API. i'd love to have something like cybertraderpro (try the free demo - http://www.cybertrader.com/cybertrader/contact_sim.asp ) with an API.
cost of foam is around $7 at home depot so its fairly trivial. messy to work with though.
Re:OT: Software for those wall-size displays?
on
The Root of All E-Mail
·
· Score: 2, Informative
NOCOL : http://www.netplex-tech.com/software/nocol/ usually. it takes data from router SNMP and displays it graphically. i would imagine some custom work goes on for converting it into a wall mounted display. some companies must be doing minor custom work on it as consultants.
umm...ok..im not too familiar with your systems but ive taken apart and put together an IBM AS/400 720e (running V4R4M0). Firstly, YOU ARE NOT SUPPOSED TO OPEN THE DAMN THINGS. IBM will send out techs for FREE to install hardware you want to install on em. They do NOT supply service manuals in general. i dont know where you got yours from. Most AS/400s (high end models) are LOCKED and cant be opened without breaking the lock with a screwdriver or using a special key from IBM. I had to break the lock with a screwdriver to open the damn thing. its the cheap plastic lock that prevents you from exposing the front panel. Secondly there arent any screws. its snap out assembly kinda like dells. The screws are on the modular boards, holding delicate connections together, which you arent supposed to take apart anyway. Thirdly, there arent any CPU cards. the CPU modules are placed in a a cage with the memory assembly which you arent supposed to take apart. The assembly snaps into the backplane slots as one module. Taking apart the CPU module assembly will destroy it completely if you arent really really careful. that thing is fragile and protected by metal casing for a good reason. Fourthly, its easy to check if your machine is under warranty. the thing has a modem built in which you can connect to a phone line and allows the machine to phone home, check the software and hardware licenses and warranty info right on your console. just login as QSECOFR and you should get in.
im afraid he's right. while your recursive algorithm may work to simplify a circuit board or your expert system with a large set of predefined rules may be able to work fairly well in a limited problem domain AI just wont work. the main problem is keeping logic coherency even with unexpected behaviour or inputs. a human programmer does that because he/she understands the entire problem domain with *all* behaviour models, expected or unexpected. AI software can never do that. Example : An expert system/genetic program/latest buzzword is controlling a passenger aircraft and has been flying people around for 50 years. its capable of handling navigational errors, failures in mechanical parts of the plane, regular flight and everything else including landings and takeoffs. A human is in the cockpit to monitor the aircraft in case of some weird problems and also to respond to radio traffic. The human gets a heart attack and dies. does the expert system note the lack of radio communication and land the plane ? nope. if a human pilot was substituted for the system he/she would land the plane which would be the correct response. bottom line is that AI just wont work. at least not in its present form today. eventually you cant specify enough of the problem domain to cover 100% of all possible cases and you have to miss out something.
and thats the real problem. that elusive "spark" of intelligence just doesnt exist in your AI systems.
Slashdot Mirror
shipped a lot of 10 E4500s from boston to NYC last yr. fedex dropped 3 of them. had smashed CPU heatsinks and dented front panels. 3 x 300,000 = nearly a million bucks worth of damage.
problem is photomechanical processses like this one cause strain and the material breaks in a day. its not (yet) a reusable transfer unless they figure out how to do repairs. could be useful for CDR type applictions though.
The main problem is that chemical rockets dont have the thrust to weight ratios to get into space cheaply. sure..that duct tape and baling wire model might eventually attain sub orbital capability but thats afar cry from lobbing tonnes of material up there to construct lasting structures.
what we really need are fusion or fission engines based on cheap, clean thermonuclear reactions. Theres significant engineering to be done to build a robust nuclear engine to get us to space with the required cost and thrust to weight ratios.
of course theres the environmental and other hazards (greenpeace supporters disrupting launches? terrorists blowing up nuclear engines?) to consider when you have regular nuclear lifts into space.
the thing is that we dont know the route to take in order to generate machines with emotions. since biological systems are grown together over millenia its VERY hard to reverse engineer a biological system of sufficient complexity.
the only way we can hope to get close is the shotgun approach -- fund tens of thousands of researchers, some of which take advantage of this to build cool toys to play with or are total morons and waste the funds. but the important thing is they try thousands of different approaches and maybe the human race gets lucky with ONE guy who figures it out. and voila -- we get computers which can reason. and all that wasted funding has been worth it.
obviously we may never get a guy who figures it out in which case the money is wasted but its like a lottery ticket where you only have to get lucky once. and when you do its worth it.
hyping AI or selling systems like kismet is one way to get additinal funding for it. that alone and the payoffs if we get lucky make it worth it.
it is everyday life. thats unfortunately a really bad thing since its basically the Cover Your Arse mentality thats dripping into our society. no one can give a concrete answer since bad things happen to those who do -- its CYA at its finest.
and not something anyone should be proud of. much less the scientific community.
as a recent dropout from the IT industry into college level teaching i can confirm its true. i normally employ a standard 2 exams + 4 labs for each course when i taught several years ago.
i tried doing the same (2+2+4) when i taught a course recently. the labs were ok...no complaints. i normally award full points to completing the labs sucessfully so the students were ok. my exams are usually bloody murder. as usual i put my first exam into the class halfway thru the course. most of the class flunked. of course with the 4 labs they were passing anyway so i wasnt too bothered.
what i didnt expect was getting hauled into the office of the Chair and being blasted for introducing examinations which might "traumatize" students and make them sue the college for "emotional duress".
i then had to drop the last exam and make em do a large lab excercise. even with that i still had people fail.
so yeah..its started in colleges already and i expect it to move into universities fairly soon.
heh. knowledge is never wasted though.
AS/400s have built in RAID-5 hardware. they also have redundant PSUs, built in UPS, the whole nine yards. they store data in EBCDIC not ASCII but its trivial to dump data from the built in DB/2 database into a PC since IBM AS/400s typically come with a PC expansion card that boots a virtual PC with windows which can see the filesystem. .... those things are bloody reliable and multiply redundant. more likely a crashed job or some other simple software error that makes it appear like the system "crashed" (i.e. not responding).
i have a theory that the midrange (its not a mainframe) somehow killed the process (or batch job) running that operated the front end of their data entry system...or the process died on its own. they probably dont know how to login as QSECOFR and restart the batch process to get their front end back and so they think their system crashed. i doubt its a hardware failure
yeah. i had that and i fixed it. fix was a bitch though..involved hardware. it was on an F-50 dual CPU IBM box with AIX 4.3.x... where are you located ? if you can pay for transport i should be able to fix it.
note that carbon fibre parts are *expensive* and dangerous to work with even though they look nice. always wear a organic respirator and gloves to prevent carbon fibres entering your skin (which they do pretty easily) and the resin from overwhelming you with fumes.
molds can typically be made for 400 bucks and the part for another 500-600 bucks. fibreglast.com gives all the info including technical brochures anyway.
hmm...power can probably be brought to these devices the same way millions of living things get it today -- photons. capture enough photons and you have electrical power to power your miniature CPU/robotic manipulator on your MEMS device.
just think about your credit card sized solar powered calculators today. they get energy from ambient light..not even solar power...just ambient. if we can get photon collectors down small enough we can power all the whiz bang assemblers we can think of. power distribution can be handled to assemblers who work in shadow by reflecting ambient light from those which collect an excess of it. think small fireflys without chemicals..just reflect photons in random directions and use a small fraction of them. no need to store energy.
granted its far off -- but not as far off as teleportation/time travel etc. its a hard (ok..VERY HARD) engineering problem more than anything else.
the NEC versa daylite has a 10.4 inch transreflective LCD. not great but it does 1024x768 and has a 600MHz CPU 128 megs ram, 10 gig hdd etc. :6 057 811.html ..more expensive than most laptops with 10 inch LCDs.
here's a review
http://computers.cnet.com/hardware/0-1027-405-
its costly at $2500 or so
yeah but it aint gonna happen. too many clueless PHBs wandering around who equate 9-5 with getting stuff done.
cant blame em too..most people love being in a box their whole lives while the average programmer hates it. something about herding cats...
the thing though is that these labs had the financial firepower to do research tightly focused at the commercial market. now the only people doing research are the people in universities who dont have the budget to do the nifty whiz bang gadgets..they do pure research only. which means a lot of university developed research will not be developed into whiz bang gadgets and hence improve the market place or the company stocks. it just means the public wont be able to buy really kewl new technology and companies wont make huge profits selling this technology. the effects are usually felt in 3-5 years or so when AT&T will find itself without any new products to develop since the research base has gone.
its a tragic loss. i hope someone can keep up the VNC GPLed tree at least.
exair has some for sale from $200 to $995.
they are cool but also very inefficient.
humans can learn far more than any AI today. they make decisions by sifting thru vast quantities of rules and data. they have true random number generators built into them. They have vast quantities of processing power which is adaptable. ALL AI systems today cant even compare to ants much less compare to a human brain.
AI is a dismal failure today at any rate. its just *hard*.
while i dont disagree that AI will find it difficult to see the big picture, growth between neurons is easy to simulate in a computer program.
repetition simulates growth. just wait until enough repeated events occur to form a solid connection.
metal activity is unguided. there is no reason to guide a self organizing system based on chaos. it just self organises. does anyone "guide" a tornado forming ? the rules are there, let chaos theory do the rest.
you might get flamed a lot out here -- but for me personally the review was VERY helpful and confirmed my opinion of Asus.
thanks. nicely done.
hmm..they have a crappy interface so they need to provide the API.m .asp ) with an API.
i'd love to have something like cybertraderpro (try the free demo - http://www.cybertrader.com/cybertrader/contact_si
cost of foam is around $7 at home depot so its fairly trivial. messy to work with though.
NOCOL : http://www.netplex-tech.com/software/nocol/ usually.
it takes data from router SNMP and displays it graphically.
i would imagine some custom work goes on for converting it into a wall mounted display.
some companies must be doing minor custom work on it as consultants.
umm...ok..im not too familiar with your systems but ive taken apart and put together an IBM AS/400 720e (running V4R4M0).
Firstly, YOU ARE NOT SUPPOSED TO OPEN THE DAMN THINGS. IBM will send out techs for FREE to install hardware you want to install on em. They do NOT supply service manuals in general. i dont know where you got yours from. Most AS/400s (high end models) are LOCKED and cant be opened without breaking the lock with a screwdriver or using a special key from IBM. I had to break the lock with a screwdriver to open the damn thing. its the cheap plastic lock that prevents you from exposing the front panel.
Secondly there arent any screws. its snap out assembly kinda like dells. The screws are on the modular boards, holding delicate connections together, which you arent supposed to take apart anyway.
Thirdly, there arent any CPU cards. the CPU modules are placed in a a cage with the memory assembly which you arent supposed to take apart. The assembly snaps into the backplane slots as one module. Taking apart the CPU module assembly will destroy it completely if you arent really really careful. that thing is fragile and protected by metal casing for a good reason.
Fourthly, its easy to check if your machine is under warranty. the thing has a modem built in which you can connect to a phone line and allows the machine to phone home, check the software and hardware licenses and warranty info right on your console. just login as QSECOFR and you should get in.
AT & T global. accept no substitutes.
they are the only one i found for virtually all the countries i have travelled to.
expensive as hell though.
im afraid he's right. while your recursive algorithm may work to simplify a circuit board or your expert system with a large set of predefined rules may be able to work fairly well in a limited problem domain AI just wont work.
the main problem is keeping logic coherency even with unexpected behaviour or inputs. a human programmer does that because he/she understands the entire problem domain with *all* behaviour models, expected or unexpected. AI software can never do that.
Example : An expert system/genetic program/latest buzzword is controlling a passenger aircraft and has been flying people around for 50 years. its capable of handling navigational errors, failures in mechanical parts of the plane, regular flight and everything else including landings and takeoffs. A human is in the cockpit to monitor the aircraft in case of some weird problems and also to respond to radio traffic. The human gets a heart attack and dies. does the expert system note the lack of radio communication and land the plane ? nope. if a human pilot was substituted for the system he/she would land the plane which would be the correct response.
bottom line is that AI just wont work. at least not in its present form today. eventually you cant specify enough of the problem domain to cover 100% of all possible cases and you have to miss out something.
and thats the real problem. that elusive "spark" of intelligence just doesnt exist in your AI systems.
so true. i'll second this one. having working in both types a truly open (no cublicle) arrangement is the best with instant mangement accessibility.