I think that this was debunked (by Lenat himself) in the article "Why AM and Eurisko appear to work", which is a mea-culpa that shows that Lenat is a proper scientist (we need more of them!) but that he was wrong. The critical insight I drew from the work is that insightful intelligence can arise from an interaction between a human and a machine, which is not to say that pure machine intelligence is impossible, but that Lenats experiments show insight arising and being codified from a collaboration.
The reason is that there are a limited number of bits in the Universe and a limited amount of seconds. If you multiply these two numbers and then multiply it by the instructions per second rate of a computer thats thought to be as fast as is physically possible then you get a number like
10^128
or so.
This is a very, very large number, but it does represent an absolute limit on tractablity of computing problems because what it says is if you could build a parallel computer that was physically optimal and ran it for the life time of the universe you would not get an answer to a problem that required more instructions than 10^128 to compute.
It's a while since first year CS (10 years) but as I remember it (and I can't be bothered to look on Google) the halting problem actually shows that it's not possible to use one program on a turing machine to check that all other programs will finish because there is at least one setting that causes the checking process to continue indefinitely.
That's not what you say here, and the difference is important.
Also model checking is well known to be exponentially complex and intractable *in the worst case* the nice thing is that the worst case hardly ever seems to crop up and the cases that do crop up in a nice implementation can be checked very fast. This is what the NASA chaps are doing her.
BTW. I'm pretty sure that I couldn't get a job as a janitor at NASA (although I am quite niffy with a mop and bucket), so I think a little more respect and consideration of opinion for them is in order from you!
It's a little while since I looked at Minsky's book, but as I remember it his point was that perceptron learning, which could only be applied one layer at a time could only separate linearly divisable functions. He showed that a multi layer perceptron with particular settings could separate an XOR problem (for example), but at the time there was no algorithm to learn these settings.
Later Rumelhart and Hinton invented back propagation that could over come this issue by learning the kind of classifiers that Minsky was describing - ie. non linearally separable spaces.
A recent revolution has come in the form of the realisation that structural risk minimisation can also be done automatically, as well as statistical risk minimisation in classifier learning. That is that we can not only minimise the error rate of the classifier (statistical risk) but also minimise the risk that we are over fitting to the data set and not the domain theory (structural risk). Algorithms that do this are things like instance based learners, support vector machines and various ensemble learners like boosters and roc learners.
I don't know of much more progress in supervised learning after this point - it's mostly held to be solved I think now. The challenges are more in things like inference based learning, community based learning and unsupervised learning of various types.
Remember maintenance! You are 100% correct for fire and forget projects, but introducting an opaque layer into the stack could mean trouble later on. People will start injecting data directly into the wire - for example from network devices or information appliances - and subtle issues and bugs could creep in. I thought this was one of the reasons we ditched CORBA?
Well - I understand your frustration, but perhaps he wasn't being as dumb as it seems on first blush.
CIOs are challenged by the need to find people who understand poorly documented proprietary data formats, and by finding techies who are techie enough to write parsers for them. XML, with standard ways to use the open source parsing libraries such as the Apache ones, support from every vendor you can think about and a course in every CS department, is very attractive to people in the position of dealing with systems ten years down the road!
What the editorial meant (I think) is that the system doesn't *roam* between landline basestations. I think it does route calls that you make in your house or office into your landline via a voip infrastructure - so you can use all your saved numbers on your cell phone while sitting and watching TV!
I think that the big bonus is for corporate customers who are often in different offices of their multi national super company, but only reachable by their cell. Using this kind of system they'll get a big cost reduction, which right now is pretty welcome.
I read on the ESA site that Huygens landed and then illuminated it's environment with a twenty watt bulb. There are comments about how surprised and please the mission scientists are that the bulb was still lit after several hours of the probe being on the surface.
My guess is that the illumination was included because the surface of Titan must be quite dark by our standards. I have two reasons for thinking that this might be the case : the fact that sol is much further from Saturn than from earth and the density of titans atmosphere that has cloaked it from us until this moment.
I think that the CCD designers (I think at Uni. Arizona?) must have faced three challenges that would have been very difficult to resolve. First the unit must be very sensitive to produce images from the craft on relatively short exposures as it fell. Secondly the unit must be low power, or at least relatively low power, finally the unit had to be light.
I suspect that the unit cost several million Euros, although I have seen no information, but a spend of that order would be justified by the cost of the lauch and trip to Titan and the penalty of device failure. I would be surprised if there was a chance of improving performance significantly for less than a ten million euro spend at the time, or they would have taken it and dumped another instrument, or canned another mission.
As I understood it they fed the dimming factor (from particles directly) and the reflection factor (from the clouds) and the seasonality of the factors (because we use less energy in the summer in Europe than in winter due to the temperate nature of the climate in most parts) into a climate simulator, and this is what indicated that the African droughts had been caused by a deflection of rainfall south in Africa.
Of course the situation is complex, but the model made a prediction that was not there previously when they incorporated this information... and that prediction was seen in actual data....
Well, if you look at the animations it *seems* like there is some flow from left to right across the frame. About 60% of the way up it looks like there is a stream bed down which fluid is flowing. The rocks look like they are sitting in hollows that are consistent with this interpretation.
Sadly the program last night indicated that there is evidence that the massive famines of the eighties in the developing world were the result of climate change, and of course a lot of children did die in those. I was upset enough about that at the time without thinking that it was actually the result of actions that I personally took.
Travis, D.J., Carleton, A.M., and R. Lauritsen (2002). "Jet Contrails and Climate: Anomalous Increases in U.S. Diurnal Temperature Range for September 11-14, 2001." Nature, Vol. 418, p. 601.
It is in Nature, the best, most heavily reviewed and authorative journal that there is.
I think that the key element in your post is "new novel"; a novel is *a work of fiction*. ie. it is a story that is made up, full of made up assertions and claims.
If you were asking me to referecen Crichtons article in Nature then I would be interested....
You are wrong. The fact is that the particulates are precipitating water from the atmosphere and generating clouds with a higher reflectivitity than normal ones.
I did watch this, and I watched the other shows that you discuss. You are seriously misrepresenting things.
* Killer Atkins: yup, this was tosh.
* The Bible Code: debunked the idea that there were number codes in the bible.
* Mega Waves: I think that the awful events of the 26th December should be sufficient to show the quality of show here.
Last nights show was a sensational presentation of science, but it was science that has been published in peer reviewed journals, in fact there was a discussion about how a journal article had very little impact on the community because although the evidence was later proved to be correct it was so shocking that most of the community discounted it.
Tenured scientist from prestigious universities were filmed supporting the programs assertions directly. Yes, there were also pictures of burning trees and starving children, but don't for one second discount the significance of people from major universities with major careers going on camera to wave the flag for an imminent collapse in our ecology.
DR DAVID TRAVIS (University of Wisconsin, Whitewater) DR GERALD STANHILL (Agricultural Research Organisation, Israel) DR BEATE LIEPERT (Lamont-Doherty Earth Observatory) PROF GRAHAM FARQUHAR (Australian National University) DR MICHAEL RODERICK (Australian National University) PROF VEERABHADRAN RAMANATHAN (University of California) DR LEON ROTSTAYN (CSIRO Atmospheric Research) DR PETER COX (Hadley Centre, Met Office)
Final question, do you by any chance know any children? Because if you do I ask you to consider how they are going to feel starving to death in a desert because you, and others, are prepared to discount any evidence whatsoever while it doesn't fit with your world view.
I have a work PC so I have a commercial AV tool. I have a lot of friends though, who are home users only and don't have much cash. I have often helped them set open office and other foss programs, but virus protection has been a problem.
Could you give more information on AVG and could anyone else who reads this who has used it and has an opinion also give some information?
Of course a business model will be found, because different people will pay. Gamers will pay for more responsive services, home workers for QoS, personalisation and management (net nanny things, perhaps) will sell to young families. It's a cert that people will be able to think of other things and there will be providers to pay for them, but only where nice mr telco is around to foot the overall bill for the build out, support and initial application deployment to seed the market for every f'n pirate to storm in afterwards!
Well, and the laws of economics as they are currently understood.
Look, what you are asking for is not just connectivity, its service. You want hdtv? 4MBs & Mpeg4 baby, that will bring a hdtv stream to your very tv, oh yes.
But what you want is all the HDTV, all the time, in every language and so on. It's a big challenge in terms of servers, in terms of discovery and distribution, in terms of copyright. These are the true issues at the moment, as well as the DSLAMS to manage the distribution of bandwidth at a local level in a cost effective way.
- who is going to pay for the investment in kit? - who is going to pay for the maintenance of the kit? - who is going to pay for the optics in the ducts? - who is going to dig the f* ducts? - who is going to pay for the kit to parlay it into the Ethernet in your house? - what are we going to pipe it round your house with? - who do you ring when it don't work? The TV company, the film company? The telco?
Slashdot Mirror
I think that this was debunked (by Lenat himself) in the article "Why AM and Eurisko appear to work", which is a mea-culpa that shows that Lenat is a proper scientist (we need more of them!) but that he was wrong. The critical insight I drew from the work is that insightful intelligence can arise from an interaction between a human and a machine, which is not to say that pure machine intelligence is impossible, but that Lenats experiments show insight arising and being codified from a collaboration.
Wasn't this one of the things that Church and Turing proved? I vaugely remember an excited lecturer waving his arms in front of a OHP...
Right - off to the pub to hammer my continuous non-digital analogue computing surface into shape.
Cell phones!
Surely Atom?
. html
http://www.ietf.org/html.charters/atompub-charter
Well, no, you probably won't.
The reason is that there are a limited number of bits in the Universe and a limited amount of seconds. If you multiply these two numbers and then multiply it by the instructions per second rate of a computer thats thought to be as fast as is physically possible then you get a number like
10^128
or so.
This is a very, very large number, but it does represent an absolute limit on tractablity of computing problems because what it says is if you could build a parallel computer that was physically optimal and ran it for the life time of the universe you would not get an answer to a problem that required more instructions than 10^128 to compute.
It's a while since first year CS (10 years) but as I remember it (and I can't be bothered to look on Google) the halting problem actually shows that it's not possible to use one program on a turing machine to check that all other programs will finish because there is at least one setting that causes the checking process to continue indefinitely.
That's not what you say here, and the difference is important.
Also model checking is well known to be exponentially complex and intractable *in the worst case* the nice thing is that the worst case hardly ever seems to crop up and the cases that do crop up in a nice implementation can be checked very fast. This is what the NASA chaps are doing her.
BTW. I'm pretty sure that I couldn't get a job as a janitor at NASA (although I am quite niffy with a mop and bucket), so I think a little more respect and consideration of opinion for them is in order from you!
It's a little while since I looked at Minsky's book, but as I remember it his point was that perceptron learning, which could only be applied one layer at a time could only separate linearly divisable functions. He showed that a multi layer perceptron with particular settings could separate an XOR problem (for example), but at the time there was no algorithm to learn these settings.
Later Rumelhart and Hinton invented back propagation that could over come this issue by learning the kind of classifiers that Minsky was describing - ie. non linearally separable spaces.
A recent revolution has come in the form of the realisation that structural risk minimisation can also be done automatically, as well as statistical risk minimisation in classifier learning. That is that we can not only minimise the error rate of the classifier (statistical risk) but also minimise the risk that we are over fitting to the data set and not the domain theory (structural risk). Algorithms that do this are things like instance based learners, support vector machines and various ensemble learners like boosters and roc learners.
I don't know of much more progress in supervised learning after this point - it's mostly held to be solved I think now. The challenges are more in things like inference based learning, community based learning and unsupervised learning of various types.
And of course the dirty word - applications.
"I've never had a real life music exec in here" said Steve "apart from those Beatles fellas wanting their cash..."
Remember maintenance! You are 100% correct for fire and forget projects, but introducting an opaque layer into the stack could mean trouble later on. People will start injecting data directly into the wire - for example from network devices or information appliances - and subtle issues and bugs could creep in. I thought this was one of the reasons we ditched CORBA?
Well - I understand your frustration, but perhaps he wasn't being as dumb as it seems on first blush.
CIOs are challenged by the need to find people who understand poorly documented proprietary data formats, and by finding techies who are techie enough to write parsers for them. XML, with standard ways to use the open source parsing libraries such as the Apache ones, support from every vendor you can think about and a course in every CS department, is very attractive to people in the position of dealing with systems ten years down the road!
unless you get BT communicator?
e ss ionID=@@@@0367275130.1106859812@@@@&BV_EngineID=cc cdadddjlfjdhlcflgcefkdffndfkk.0
http://www.bt.com/btcommunicator/index.jsp?BV_S
hmmmm took me 10 minutes to find on BT.com... still free calls for a month!
gee....
What the editorial meant (I think) is that the system doesn't *roam* between landline basestations. I think it does route calls that you make in your house or office into your landline via a voip infrastructure - so you can use all your saved numbers on your cell phone while sitting and watching TV!
I think that the big bonus is for corporate customers who are often in different offices of their multi national super company, but only reachable by their cell. Using this kind of system they'll get a big cost reduction, which right now is pretty welcome.
I read on the ESA site that Huygens landed and then illuminated it's environment with a twenty watt bulb. There are comments about how surprised and please the mission scientists are that the bulb was still lit after several hours of the probe being on the surface.
My guess is that the illumination was included because the surface of Titan must be quite dark by our standards. I have two reasons for thinking that this might be the case : the fact that sol is much further from Saturn than from earth and the density of titans atmosphere that has cloaked it from us until this moment.
I think that the CCD designers (I think at Uni. Arizona?) must have faced three challenges that would have been very difficult to resolve. First the unit must be very sensitive to produce images from the craft on relatively short exposures as it fell. Secondly the unit must be low power, or at least relatively low power, finally the unit had to be light.
I suspect that the unit cost several million Euros, although I have seen no information, but a spend of that order would be justified by the cost of the lauch and trip to Titan and the penalty of device failure. I would be surprised if there was a chance of improving performance significantly for less than a ten million euro spend at the time, or they would have taken it and dumped another instrument, or canned another mission.
As I understood it they fed the dimming factor (from particles directly) and the reflection factor (from the clouds) and the seasonality of the factors (because we use less energy in the summer in Europe than in winter due to the temperate nature of the climate in most parts) into a climate simulator, and this is what indicated that the African droughts had been caused by a deflection of rainfall south in Africa.
Of course the situation is complex, but the model made a prediction that was not there previously when they incorporated this information... and that prediction was seen in actual data....
Well, if you look at the animations it *seems* like there is some flow from left to right across the frame. About 60% of the way up it looks like there is a stream bed down which fluid is flowing. The rocks look like they are sitting in hollows that are consistent with this interpretation.
Yes, sorry, that was a bit OTT.
Sadly the program last night indicated that there is evidence that the massive famines of the eighties in the developing world were the result of climate change, and of course a lot of children did die in those. I was upset enough about that at the time without thinking that it was actually the result of actions that I personally took.
Bad, bad, bad.
Yes they do.
They also say "do something about it now, before the methane hydrates are released and the rainforests catch fire."
Of course they say it in a very calm voice, but their eyes, their eyes are, well, jumpy...
Travis, D.J., Carleton, A.M., and R. Lauritsen (2002).
"Jet Contrails and Climate: Anomalous Increases in U.S. Diurnal Temperature Range for September 11-14, 2001."
Nature, Vol. 418, p. 601.
It is in Nature, the best, most heavily reviewed and authorative journal that there is.
Respect this science, or you will die!
I think that the key element in your post is "new novel"; a novel is *a work of fiction*. ie. it is a story that is made up, full of made up assertions and claims.
If you were asking me to referecen Crichtons article in Nature then I would be interested....
You are wrong. The fact is that the particulates are precipitating water from the atmosphere and generating clouds with a higher reflectivitity than normal ones.
So it makes sense.
Read the transcript.
I did watch this, and I watched the other shows that you discuss. You are seriously misrepresenting things.
* Killer Atkins: yup, this was tosh.
* The Bible Code: debunked the idea that there were number codes in the bible.
* Mega Waves: I think that the awful events of the 26th December should be sufficient to show the quality of show here.
Last nights show was a sensational presentation of science, but it was science that has been published in peer reviewed journals, in fact there was a discussion about how a journal article had very little impact on the community because although the evidence was later proved to be correct it was so shocking that most of the community discounted it.
Tenured scientist from prestigious universities were filmed supporting the programs assertions directly. Yes, there were also pictures of burning trees and starving children, but don't for one second discount the significance of people from major universities with major careers going on camera to wave the flag for an imminent collapse in our ecology.
DR DAVID TRAVIS (University of Wisconsin, Whitewater)
DR GERALD STANHILL (Agricultural Research Organisation, Israel)
DR BEATE LIEPERT (Lamont-Doherty Earth Observatory)
PROF GRAHAM FARQUHAR (Australian National University)
DR MICHAEL RODERICK (Australian National University)
PROF VEERABHADRAN RAMANATHAN (University of California)
DR LEON ROTSTAYN (CSIRO Atmospheric Research)
DR PETER COX (Hadley Centre, Met Office)
Final question, do you by any chance know any children? Because if you do I ask you to consider how they are going to feel starving to death in a desert because you, and others, are prepared to discount any evidence whatsoever while it doesn't fit with your world view.
In science we have words for people like you.
They are not very nice.
I have a work PC so I have a commercial AV tool. I have a lot of friends though, who are home users only and don't have much cash. I have often helped them set open office and other foss programs, but virus protection has been a problem.
Could you give more information on AVG and could anyone else who reads this who has used it and has an opinion also give some information?
Thanks.
Nice - yup this is the issue, in the can, bingo.
Of course a business model will be found, because different people will pay. Gamers will pay for more responsive services, home workers for QoS, personalisation and management (net nanny things, perhaps) will sell to young families. It's a cert that people will be able to think of other things and there will be providers to pay for them, but only where nice mr telco is around to foot the overall bill for the build out, support and initial application deployment to seed the market for every f'n pirate to storm in afterwards!
Hoist the jolly rodger! arrr arrr!
Well, and the laws of economics as they are currently understood.
Look, what you are asking for is not just connectivity, its service. You want hdtv? 4MBs & Mpeg4 baby, that will bring a hdtv stream to your very tv, oh yes.
But what you want is all the HDTV, all the time, in every language and so on. It's a big challenge in terms of servers, in terms of discovery and distribution, in terms of copyright. These are the true issues at the moment, as well as the DSLAMS to manage the distribution of bandwidth at a local level in a cost effective way.
- who is going to pay for the investment in kit?
- who is going to pay for the maintenance of the kit?
- who is going to pay for the optics in the ducts?
- who is going to dig the f* ducts?
- who is going to pay for the kit to parlay it into the Ethernet in your house?
- what are we going to pipe it round your house with?
- who do you ring when it don't work? The TV company, the film company? The telco?
Look, a business case is a business case is a business case.
Either you can make a return or you can't. Risk (will it work, will it have the benefit you claim) can be factored in).
But your problem is that no one, no one in the world, will invest $10K for a projected $15K save over one, two or three years.