No, his numbers are quite reasonable. You Americans have no idea about efficiency and you probably waste more power than the average European household uses in total.
I share a one-bed (ie 4 rooms, I would guess about 100m^2) flat with my girlfriend. This is a typical sized household for the UK (although the average size is obviously larger). We average 5 kWh per day (so ~1600 kWh per month). We don't live in the dark, the flat is warm over the winter despite the horrific lack of insulation and we are hardly living in the stone age. From the couch I can see three games consoles, three computers, flat-screen tv etc...
Your power consumption is seven times larger than ours (ignoring fuel consumption which is a major component). You are not conservative by any measure, just because you think you are at the low-usage end of the most wasteful, polluting nation on earth. Do you ever wonder why the rest of the world wants you to hold back on the raping the planet?
Once upon a time cars were pretty simple. The most effective way to fix a car that had broken was to find a mechanic. This was a man trained in the models of how cars work. He would sift through the collection of parts (data) in the car until he noticed an anomaly that he would charge you outrageously for.
Now cars have become so complex that these models are no longer needed. Instead you can just examine the millions of cars that either work or don't work right there on teh interweb. One you find a correlation between your car and another car you can then fix the difference without knowing anything about models of "how cars work"!
Err, maybe that analogy was a little too accurate as it has made his argument sound shit?
I used to think that I could translate most dialects of bullshit into english but this threw me off guard. The most reasonable explanation is that Chris Anderson is a tool and doesn't know what he is talking about.
For example, data is now "paradigm agnostic". Seriously, wtf? When was data ever not "paradigm agnostic" and when did we develop the need for a term to describe it. Data is data. It is raw, and unanalysed, and as such the notion of a paradigm is completely irrelevant.
Interesting, what are you using? We've got Condor as a job scheduler on our clusters. It's more of a pain to get C code up and running as the checkpointing is properly integrated into the JVM.
Personally I run prolog on our local cluster, but err, I guess that's just me. For some tasks java isn't going to be too much of a hit. Any numerically heavy stuff; i.e. dense linear algebra should be coded up in C because it will be a lot faster. For that kind of code we spend a lot of time writing inline assembly blocks because it improves the performance alot compared to gcc or icc.
For the kind of simulations mentioned in the blurb (agents and economic activity) I'm not sure that java would be any slower than C. The problem is no longer scheduling word-level arithmetic operations on the processor for maximum throughput. Instead it is much more similar to graph mutation, where you have a lot of indirect references and you need to search and then mutate data structures. From what I've seen this code runs just as fast in java as in C.
It's not really a straight analogy. Is releasing code with a few bugs really in the same league as deliberately falsifying results to commit fraud?
In general, academic fraud is harder to commit in CS (and maths) than in disciplines that have real physical experiments. The data in CS is more reproducible, and the content of papers is more devoted to proving the correctness of techniques and approaches. Maths is even further along the scale where academic fraud would be impossible.
What is easily done in CS, and is far too prevalent is plagiarism. The subject is drowning in it, and any professional oath for CS should include a description of what plagiarism is, and why the person taking the oath with avoid it.
Yikes that was a huge rambling hike through your views on the neats vs scruffies debate.
Quite an important correction though, you completely murdered your description of Hutter's results. Not only did he not claim that compression research would lead to an AI but he went to great pains to explain that an optimal compressor would not be an AI.
The point was that solving the compression problem is a prerequisite for an intelligent system. So the research into compression should really happen first because we will need some of the lessons to be learnt before attempting the harder problem of AI.
The reduction that he proved was that if you can build a decent AI then it must be an optimal compressor at the very least. So performing research on how to handle compresion would be a good place to start.
The choice of natural language as the corpus is quite important because a half-decent compressor needs to do sophisticated semantic analysis that is relevant for anyone pursuing hard AI. Lastly, I don't know if you've done any work in the area but applying a Baysian approach to this type of problem requires well known priors. The more prior knowledge of the domain that you allow the less "intelligence" is needed to solve the problem. Learning the prior distributions is difficult and complex, but something like Hutter's approach of picking a specific problem as a benchmark is probably a good way of attacking it.
That's not the same. When there is a success made in any of the fields that you mention it remains part of that field. A solved part of that field. Every success made in AI is no longer AI, so there are no successes or progress made "within the field". It's quite a substantial difference when it comes down to the perception of the field.
Chess was considered the ultimate AI problem back in the 40s and 50s. When we knew little about the game and how to solve it, it seemed that intelligence must be required to solve it. Now that machines are better at chess than humans we've redefined as a problem that is susceptible to brute force. It is not considered a success in the AI field, just another refinement of what is not AI.
Unlike you Thomas Kuhn understood the value of good research and the place of academia. Of course you've just read what you wanted to hear into The Structure of Scientific Revolutions to justify your own failure.
Come on Louis try and defend your work. It's always fun to hear you push your version of Computer Science. You can give Archimedies Plutonium a run for his money on your better days.
Aren't you going to try and defend your own definition of Universality? I'm sure you'll be pushing it into another slashdot discussion soon...
That's interesting. I was thinking of making the same change recently. Most of the subversion repositories that I use are either single user or only a small number, so the distributed nature of giu didn't appeal for the normal reasons.
But after a trip away without an internet connection I realised that not being able to take the repository with me and then handle merges when I got back was a bit of a limitation. Git looks interesting because it would get around this issue... but I just spent some time hacking up a tool to merge dumpfiles in the correct date order for revisions. If it works out I may just use it to keep a mirror of the repository on my laptop and merge it back in when necessary. Never upgrade your tools unless you really need to...:)
Perhaps that is why you never made it as a scientist and had to resort to cranking out the shit on your blog. Because you failed to understand when opinions become verified and objective facts.
Nice. You fail to accept your errors because as I explained at the beginning you are a crank. All it took was a little rational argument to prove it.
I've argued you to this point before, but every few years it is worth doing it again so that a few passing readers get to find out what you're like without wasting too much of their time.
Us people eh? You are trying your hardest to cover up your inadequacies with cheap jibs. But lets stick firmly to your area of ignorance.
It is not a question of people not understanding. What you are trying (poorly) to explain is emulation, not simulation. By using a term that has a well defined technical meaning, and trying to redefine it to suit your argument you just look foolish.
Execution speed is not an issue for universality. If a particular machine can run any computation possible, regardless of speed then it is universal. Adding parallelisation does not change the set of computations that can be executed. This insight was Turing's great contribution to the field, and it is why he is held in such high regard. To go further and show constructions of non-computable functions earnt him his title as the father of CS.
Adding parallelism or temporality to abstract models has been done many times. It does not expand the set of computations, and only affects the performance and expressiveness. It does not change the computational model, but merely changes how informative descriptions of this model are. This is a subtle point, although it is well covered on all undergraduate degree programs. Perhaps you should take the time to learn what has gone before, instead of assuming that you know better. Although your arrogance is astounding, your grasp of the subject is sorely lacking.
Now, lets try again. Either produce an example of a computation that cannot be simulated on a Turing Machine, or admit that you have misused the term. Once you accept that you have failed to use the right terminology then why don't you try to explain which part of the temporal behaviour you think can be modeled in a better way. Of course whatever 'improvement' you describe can be simulated on a Turing Machine...
Simulation is not concerned with speed. That is why it is a measure of computational power - how many computations can be undertaken, rather than the performance of them.
If you assume that the universe is some sort of computation (which is way beyond what anyone can prove) then it can still be simulated by a Turing Machine. There is still an open question of whether or not a Quantum Computer can compute anything beyond what a Turing Machine can, but most people in the field are betting that it cannot, that it can 'merely' run some types of computation faster.
I'm not taking offense, I've actually read most of your 'work' and I know that you don't actually know what you are talking about. Try again, name a single computation that a Turing Machine cannot simulate.
Your logic is terrible. It is not unethical to use wifi because "you may then do something illegal with that wifi". It would be unethical to do something illegal on your neighbour's wifi. If you still have trouble understanding this then consider a car analogy: "It is unethical to drive a car because while driving that car you may do something unethical".
A customer pays an ISP to use bandwidth. Many ISPs (mine included) don't have terms to say that you must not offer wifi access to other people. If your neighbour lets you use his wifi, and the bandwidth has been paid for then nobody is freeloading and nobody is harmed.
You need to learn what a logical implication is before you try and use one.
Yeah, my first job gave me some similar experiences. From the old-school unix hackers who turned up their noses at the crappy 386 dos boxes that most of us used and showed me how to log into the big scary sun boxes to get some real work done, to the shit-hot programmer who would rewrite the software stacks over his lunch hour and then wonder why it took other people weeks to do the same work.
Some time after meeting these guys (when I'd had long enough for what they taught me to sink in) I realised that I wasn't the smartest guy in the world. Nice thing is I've just reached the age where I can do the same for students.
If the Turing machine is not universal then name a single computation that cannot be simulated on it.
(Oh dear, is that just me being a crusty old academic and actually understanding the terms that you are blindly throwing around).
If you can name even one, then write a paper about it and become famous. Rather than regurgitating the same old tired shit on forums and being infamous.
He is quite a well-known kook and it is always quite amusing to watch him try and defend himself. If you follow the link to his blog and read his first posting about why Turing Machines are not really universal then you will get the drift.
It's amusing to hear his interpretation of Kuhn as well. I can see how somebody that lacks an education in the field that he pursues (his ideas have been tried many times before) would interpret Kuhn that way, but it must still have taken an amazing set of mental blinkers to do so.
Yikes I remember being 18 as well. Don't worry by the time you grow up you'll be amazed at how much everyone else has learnt. I'm only half taking the piss. When I first turned up at uni as a fresh faced undergraduate you've perfectly described my own self-image. Now that I'm an older, more cynical postdoc I see the world differently.
One thing that will make a real difference for you is to find your natural peer group. Until then, like the AC said: ask for lessons in humility.
Slashdot Mirror
*cough*. Very true. I'll get my coat...
No, his numbers are quite reasonable. You Americans have no idea about efficiency and you probably waste more power than the average European household uses in total.
I share a one-bed (ie 4 rooms, I would guess about 100m^2) flat with my girlfriend. This is a typical sized household for the UK (although the average size is obviously larger). We average 5 kWh per day (so ~1600 kWh per month). We don't live in the dark, the flat is warm over the winter despite the horrific lack of insulation and we are hardly living in the stone age. From the couch I can see three games consoles, three computers, flat-screen tv etc...
Your power consumption is seven times larger than ours (ignoring fuel consumption which is a major component). You are not conservative by any measure, just because you think you are at the low-usage end of the most wasteful, polluting nation on earth. Do you ever wonder why the rest of the world wants you to hold back on the raping the planet?
Once upon a time cars were pretty simple. The most effective way to fix a car that had broken was to find a mechanic. This was a man trained in the models of how cars work. He would sift through the collection of parts (data) in the car until he noticed an anomaly that he would charge you outrageously for.
Now cars have become so complex that these models are no longer needed. Instead you can just examine the millions of cars that either work or don't work right there on teh interweb. One you find a correlation between your car and another car you can then fix the difference without knowing anything about models of "how cars work"!
Err, maybe that analogy was a little too accurate as it has made his argument sound shit?
I used to think that I could translate most dialects of bullshit into english but this threw me off guard. The most reasonable explanation is that Chris Anderson is a tool and doesn't know what he is talking about.
For example, data is now "paradigm agnostic". Seriously, wtf? When was data ever not "paradigm agnostic" and when did we develop the need for a term to describe it. Data is data. It is raw, and unanalysed, and as such the notion of a paradigm is completely irrelevant.
Interesting, what are you using? We've got Condor as a job scheduler on our clusters. It's more of a pain to get C code up and running as the checkpointing is properly integrated into the JVM.
Personally I run prolog on our local cluster, but err, I guess that's just me. For some tasks java isn't going to be too much of a hit. Any numerically heavy stuff; i.e. dense linear algebra should be coded up in C because it will be a lot faster. For that kind of code we spend a lot of time writing inline assembly blocks because it improves the performance alot compared to gcc or icc.
For the kind of simulations mentioned in the blurb (agents and economic activity) I'm not sure that java would be any slower than C. The problem is no longer scheduling word-level arithmetic operations on the processor for maximum throughput. Instead it is much more similar to graph mutation, where you have a lot of indirect references and you need to search and then mutate data structures. From what I've seen this code runs just as fast in java as in C.
It's not really a straight analogy. Is releasing code with a few bugs really in the same league as deliberately falsifying results to commit fraud?
In general, academic fraud is harder to commit in CS (and maths) than in disciplines that have real physical experiments. The data in CS is more reproducible, and the content of papers is more devoted to proving the correctness of techniques and approaches. Maths is even further along the scale where academic fraud would be impossible.
What is easily done in CS, and is far too prevalent is plagiarism. The subject is drowning in it, and any professional oath for CS should include a description of what plagiarism is, and why the person taking the oath with avoid it.
Well, you could see the modern schooling system as a method of making the Turing Test simpler...
Yikes that was a huge rambling hike through your views on the neats vs scruffies debate.
Quite an important correction though, you completely murdered your description of Hutter's results. Not only did he not claim that compression research would lead to an AI but he went to great pains to explain that an optimal compressor would not be an AI.
The point was that solving the compression problem is a prerequisite for an intelligent system. So the research into compression should really happen first because we will need some of the lessons to be learnt before attempting the harder problem of AI.
The reduction that he proved was that if you can build a decent AI then it must be an optimal compressor at the very least. So performing research on how to handle compresion would be a good place to start.
The choice of natural language as the corpus is quite important because a half-decent compressor needs to do sophisticated semantic analysis that is relevant for anyone pursuing hard AI. Lastly, I don't know if you've done any work in the area but applying a Baysian approach to this type of problem requires well known priors. The more prior knowledge of the domain that you allow the less "intelligence" is needed to solve the problem. Learning the prior distributions is difficult and complex, but something like Hutter's approach of picking a specific problem as a benchmark is probably a good way of attacking it.
That's not the same. When there is a success made in any of the fields that you mention it remains part of that field. A solved part of that field. Every success made in AI is no longer AI, so there are no successes or progress made "within the field". It's quite a substantial difference when it comes down to the perception of the field.
Chess was considered the ultimate AI problem back in the 40s and 50s. When we knew little about the game and how to solve it, it seemed that intelligence must be required to solve it. Now that machines are better at chess than humans we've redefined as a problem that is susceptible to brute force. It is not considered a success in the AI field, just another refinement of what is not AI.
Ah, so as always you return to your anal fixation. I see that the rumours of your ass fascination were not exaggerated.
Yawn. Is that the best you can come up with?
Unlike you Thomas Kuhn understood the value of good research and the place of academia. Of course you've just read what you wanted to hear into The Structure of Scientific Revolutions to justify your own failure.
Come on Louis try and defend your work. It's always fun to hear you push your version of Computer Science. You can give Archimedies Plutonium a run for his money on your better days.
Aren't you going to try and defend your own definition of Universality? I'm sure you'll be pushing it into another slashdot discussion soon...
Out of interest, why do you use multiple clients on the same working copy? Isn't that somewhat unusual?
That's interesting. I was thinking of making the same change recently. Most of the subversion repositories that I use are either single user or only a small number, so the distributed nature of giu didn't appeal for the normal reasons.
:)
But after a trip away without an internet connection I realised that not being able to take the repository with me and then handle merges when I got back was a bit of a limitation. Git looks interesting because it would get around this issue... but I just spent some time hacking up a tool to merge dumpfiles in the correct date order for revisions. If it works out I may just use it to keep a mirror of the repository on my laptop and merge it back in when necessary. Never upgrade your tools unless you really need to...
He does physics as well :) I've only seen his "work" on computer psuedo-science
Perhaps that is why you never made it as a scientist and had to resort to cranking out the shit on your blog. Because you failed to understand when opinions become verified and objective facts.
Nice. You fail to accept your errors because as I explained at the beginning you are a crank. All it took was a little rational argument to prove it.
I've argued you to this point before, but every few years it is worth doing it again so that a few passing readers get to find out what you're like without wasting too much of their time.
Us people eh? You are trying your hardest to cover up your inadequacies with cheap jibs. But lets stick firmly to your area of ignorance.
It is not a question of people not understanding. What you are trying (poorly) to explain is emulation, not simulation. By using a term that has a well defined technical meaning, and trying to redefine it to suit your argument you just look foolish.
Execution speed is not an issue for universality. If a particular machine can run any computation possible, regardless of speed then it is universal. Adding parallelisation does not change the set of computations that can be executed. This insight was Turing's great contribution to the field, and it is why he is held in such high regard. To go further and show constructions of non-computable functions earnt him his title as the father of CS.
Adding parallelism or temporality to abstract models has been done many times. It does not expand the set of computations, and only affects the performance and expressiveness. It does not change the computational model, but merely changes how informative descriptions of this model are. This is a subtle point, although it is well covered on all undergraduate degree programs. Perhaps you should take the time to learn what has gone before, instead of assuming that you know better. Although your arrogance is astounding, your grasp of the subject is sorely lacking.
Now, lets try again. Either produce an example of a computation that cannot be simulated on a Turing Machine, or admit that you have misused the term. Once you accept that you have failed to use the right terminology then why don't you try to explain which part of the temporal behaviour you think can be modeled in a better way. Of course whatever 'improvement' you describe can be simulated on a Turing Machine...
You don't really understand the terms do you?
Simulation is not concerned with speed. That is why it is a measure of computational power - how many computations can be undertaken, rather than the performance of them.
If you assume that the universe is some sort of computation (which is way beyond what anyone can prove) then it can still be simulated by a Turing Machine. There is still an open question of whether or not a Quantum Computer can compute anything beyond what a Turing Machine can, but most people in the field are betting that it cannot, that it can 'merely' run some types of computation faster.
I'm not taking offense, I've actually read most of your 'work' and I know that you don't actually know what you are talking about. Try again, name a single computation that a Turing Machine cannot simulate.
But you are not biting anything. I pointed out that your logic was flawed, so there is no debate about the quality of your points.
If you insist on arguing about something that you have already shown to be wrong on, then:
1. I never mentioned borrowing someone else's car. I was pointing out that simple substitution into your original claim showed that it was circular.
2. As explained, using bandwidth for free is not necessarily freeloading. Again, as explained nobody is subsidising your use if it has been paid for.
Your logic is terrible. It is not unethical to use wifi because "you may then do something illegal with that wifi". It would be unethical to do something illegal on your neighbour's wifi. If you still have trouble understanding this then consider a car analogy: "It is unethical to drive a car because while driving that car you may do something unethical".
A customer pays an ISP to use bandwidth. Many ISPs (mine included) don't have terms to say that you must not offer wifi access to other people. If your neighbour lets you use his wifi, and the bandwidth has been paid for then nobody is freeloading and nobody is harmed.
You need to learn what a logical implication is before you try and use one.
Yeah, my first job gave me some similar experiences. From the old-school unix hackers who turned up their noses at the crappy 386 dos boxes that most of us used and showed me how to log into the big scary sun boxes to get some real work done, to the shit-hot programmer who would rewrite the software stacks over his lunch hour and then wonder why it took other people weeks to do the same work.
Some time after meeting these guys (when I'd had long enough for what they taught me to sink in) I realised that I wasn't the smartest guy in the world. Nice thing is I've just reached the age where I can do the same for students.
If the Turing machine is not universal then name a single computation that cannot be simulated on it.
(Oh dear, is that just me being a crusty old academic and actually understanding the terms that you are blindly throwing around).
If you can name even one, then write a paper about it and become famous. Rather than regurgitating the same old tired shit on forums and being infamous.
He is quite a well-known kook and it is always quite amusing to watch him try and defend himself. If you follow the link to his blog and read his first posting about why Turing Machines are not really universal then you will get the drift.
It's amusing to hear his interpretation of Kuhn as well. I can see how somebody that lacks an education in the field that he pursues (his ideas have been tried many times before) would interpret Kuhn that way, but it must still have taken an amazing set of mental blinkers to do so.
Yikes I remember being 18 as well. Don't worry by the time you grow up you'll be amazed at how much everyone else has learnt. I'm only half taking the piss. When I first turned up at uni as a fresh faced undergraduate you've perfectly described my own self-image. Now that I'm an older, more cynical postdoc I see the world differently.
One thing that will make a real difference for you is to find your natural peer group. Until then, like the AC said: ask for lessons in humility.
Because you like to park in handicapped spaces while handicapped people make handicapped faces