There are a fairly large set of questions to be asked about this sort of project, especially in which AI is being utilised as the sole method for optimising signals or navigation. The primary one is an age old question which is just going to get harder: Is it a feature or a bug?
Take a neural network as the sole navigational utility. Sure its been trained through 100,000 generations to work out the opimal path in realtime to fly out to saturn and back, but when it finally coes down to it, do you trust it? There is no algorithm you can check, there is no definite way of predicting what it may do if it encounters some previously unthought of situation.
Imagine, you turn on your ship, give your target, ship starts flying there no problem, then a meteor flies past you which happens to look remarkably like a square. Neural network gets a flashback to its initial training when simple squares were used to indicate planets because it was simpler, and it makes a massive erroneous gavitational adjustment and starts flying towards the sun.
Thats bad enough, but you can't even tell whether its idea of flying towards the sun is a good idea it has suddenly had about a slingshot that it could do to get you to your destination faster, or whether it has just gone nuts and is trying to get you killed.
The same, although less extreme cases apply to most things, if the AI is the only thing doing signal adjustment on your cellphone, maybe it'll flip out for no reason that is discernable. What do you do then? you can't "fix" the bug, its buried deep in such a complex neural network that it'd be like trying to figure out why a mute human with no body language drew a picture of a frog when told to draw a picture of an apple.
At least to start with, I think we are going to find that neural networks will only be good for tweaking certain aspects of a standard algorithmic system, and while these limitations are in place, they won't be able to show such huge advantages in signal tracking etc that is proclaimed for them. It will be some time yet before we can figure out ways of making AI safe.
As a rebuttle to certain points (Not necessarily because I believe it is so, but just because I don't believe it is):
1. As I noted previously, a hash table has a certain volume up to which collisions are so minimal that there is no performance impact, therefore it may well be that the brain, a product of an evolution during the majority of which a lifespan of 40 years was impressive, has developed a bucket/hash equivalent where it starts getting bad performance from inserts/retrievals around 25/30.
2. As you noted, reinforcement has a great deal to do with it as well, people at 25 have more cross-references to use to support a memory than 16 year olds, plus the fact that 16 year olds are usually swimming in hormones which may well affect certain varieties of memory.
This is not to say that I don't totally agree with you on your picture of the brain, however I see no reason from your arguments that the comparison with a hash table in a simplistic fashion is not valid.
Furthermore, the point I was trying to make in the end is this: I don't believe reviving a bunch of previously dead braincells would achieve anything towards increasing mental performance. The functions and structures of the brain by the time this is relevant are already in place, and the sudden addition of a bunch of braincells would, IMHO, be like malloc()'ing 200kb more space for your hash bucket points, a waste of time at best, since the hash algorithm never goes there anyway.
In a system as complex as the brain, it is quite possible that such a revival would have seriously adverse effects as the new braincells struggle the adjust to the weightings of the network around it, often firing spontaneously where previously there was no fire at all, and in large numbers, this could do extremely weird, if not entirely bad, things.
In the single case of the treatment of certain brain diseases, I believe it may be effective, because entire sections of the brain may die out, and having anything here is better than nothing, but for the case of mental capability, I suspect the technology is of little use.
I had always been of the belief that the reason learning ability and linking slows down as one ages is not due to the loss of brain cells as such, but more the fact that you've already got a lot of information stored in there and it takes longer to make leaps and jumps around in it.
For a programming analogy take your average hash table, the first bunch of inserts and recalls are 1 step, because the bucket at the hash point is empty, so its the first reply you get, but as you fill up a hash table, you begin to get collisions, so you have to do step-searches through the buckets, or jump to overflow lists etc, slowing everything down. The hash is more useful because it contains more data, but to get any data out of it takes a longer time, and to insert new data in also takes a longer time.
Admittedly the brain is of a considerably different structure to that of your average hash table, but it seems an appropriate analogy to me.
Having read a significant proportion of this discussion so far and seen a significant number of people arguing against M2, I'd like to put in an argument for it.
M1 is in place for several reasons: 1. Removing spam and flames 2. Adding notice to well written articles
Where exactly does this fail? several points have been raised: 1. People push their own agenda, moderating up articles by friends or whose topic they agree with, rather than the reasoning behind it.
2. Its a simple fact that some people have very different views from others.
In a collaborative moderation scheme such as this, it is entirely necessary to limit these bad effects as much as possible. The first attempt here on slashdot was to minimise the damage moderators could do, by giving them an extremely limited amount of power.
I believe M2 represents a better solution. While it cannot have effect instantly, since it will take a short amount of time for karma ratings to settle properly and for people to get used to the system, it now gives us the ability to not only limit those moderators who moderate inappropriately, but also to give more power to those who regularly moderate well, thus reducing the need to have more and more people moderate (reducing its special status and making it a chore) while at the same time, increasing the likelyhood that any given moderation point awarded is from a moderator whose moderating methods meet the approval of the community.
In this case, the presence of M3 or higher would never be necessary, M2 represents the will of the community as a whole, and M1 represents the average of that will, with a human intellect.
All in all, an excellent system, and as a collaborative system (as opposed to personal moderation, which is another issue entirely) I believe it only requires some tweaking to achieve excellent results.
I did say heavily dynamic. Very complex pages with realtime generated graphs and tabled data, pulled from SQL databases. Include on top of this user accounting and tracking, a large degree of modularity within the code due to the complexity of the system etc, and 7 hits a second is, in my opinion anyway, quite impressive. A significant number of tricks were utilised to decrease the CPU hit including an httpd accelerator (squid in reverse), various partial pregeneration of pages where possible, and rewrites to cached copies of graphs and data where possible.
It was written in PHP, and I'm more than happy with the performance, in fact (I'm not the author of much of the code itself) I'm more often than not, very impressed with the speed with which it operates. I am looking forward to being able to move it over to Zend, this should increase our capacity even more, without any hardware additions.
While serving up plain HTML is no biggie, and any old box will do for that, serving dynamic content can be orders of magnatude harder. If you have heavily dynamic material (And your concept suggests that perhaps you do) then you will need a fairly capable box in order to respond quickly to requests. It is here and in bandwidth that the critical bottlenecks lie.
To give a suggestion of the CPU power required, the company I work for has several heavily loaded servers:
A celeron 350/128mb ram, maxes out at approximately 7 hits/second (Heavily dynamic material)
A celeron 350/128mb ram, maxes out at ~17hits/sec (Quite heavily dynamic material).
I just don't know how many hits/sec the/. effect generates:)
Note that these servers have been specially configured to handle the traffic involved, it is unlikely that you will go to the same levels of specialisation, so leave some extra space.
It seems like a remarkable conclusion. I don't believe that they have really put enough detail into the article to judge it correctly. I make the claim that vision cannot be entirely serial, I play an arcade game called House of the Dead in two player mode by myself, and having played this for some time now, I have little or no difficulty maintaining accurate aim from both guns on varying targets on the screen.
While it may well be true that the highest level of vision is serial, this particular level of vision must be quite tightly defined, for, going back to the House of the Dead example, I always shoot for the head, which is by no means just a simple object recognition in such a game.
I suspect more research really needs to be done in the area, and more importantly, that conclusions need to be very accurately defined, rather than making such broad statements.
Trying 169.207.154.108... Connected to crack.linuxppc.org. Escape character is '^]'. GET/cgi-bin/cachemgr.cgi?wtf=9 HTTP/1.1 Host: crack.linuxppc.org
Connection closed by foreign host.
Whats the deal there? no perm-denied..no no-such-page. Is this definitely a stock install?
Re:Not again ... (Unable to run on Slackware)
on
Mozilla M6 released
·
· Score: 1
Ok, here's the deal as far as I'm aware. I'm loath To report it as a bug, since its only a bug with the binary, not with compiled source.
The libstdc++ library contains a symbol, __iostream8 or similar, which normally only has one "_" in front of it. However the library which is being compiled against by the binary builder is an old, broken one, which gives this particular symbol TWO underscores. Both Debian and Redhat appear to distrubute, as part of the standard distrubution, a copy of the old, broken, version as well as the newer versions which fix the bug. Thus the binaries work fine on them, but not on slackware and possibly a couple of other distributions, because they DON'T include a copy of the broken lib. There isn't anything you can do afaik, Your only option is to attempt to compile the source yourself, and thus gain a binary which uses the correct symbol for iostream8
Slashdot Mirror
A genetic algorithm is never intelligent, its results can be though :) Thats you buddy :)
There are a fairly large set of questions to be asked about this sort of project, especially in which AI is being utilised as the sole method for optimising signals or navigation. The primary one is an age old question which is just going to get harder: Is it a feature or a bug?
Take a neural network as the sole navigational utility. Sure its been trained through 100,000 generations to work out the opimal path in realtime to fly out to saturn and back, but when it finally coes down to it, do you trust it? There is no algorithm you can check, there is no definite way of predicting what it may do if it encounters some previously unthought of situation.
Imagine, you turn on your ship, give your target, ship starts flying there no problem, then a meteor flies past you which happens to look remarkably like a square. Neural network gets a flashback to its initial training when simple squares were used to indicate planets because it was simpler, and it makes a massive erroneous gavitational adjustment and starts flying towards the sun.
Thats bad enough, but you can't even tell whether its idea of flying towards the sun is a good idea it has suddenly had about a slingshot that it could do to get you to your destination faster, or whether it has just gone nuts and is trying to get you killed.
The same, although less extreme cases apply to most things, if the AI is the only thing doing signal adjustment on your cellphone, maybe it'll flip out for no reason that is discernable. What do you do then? you can't "fix" the bug, its buried deep in such a complex neural network that it'd be like trying to figure out why a mute human with no body language drew a picture of a frog when told to draw a picture of an apple.
At least to start with, I think we are going to find that neural networks will only be good for tweaking certain aspects of a standard algorithmic system, and while these limitations are in place, they won't be able to show such huge advantages in signal tracking etc that is proclaimed for them. It will be some time yet before we can figure out ways of making AI safe.
As a rebuttle to certain points (Not necessarily because I believe it is so, but just because I don't believe it is):
1. As I noted previously, a hash table has a certain volume up to which collisions are so minimal that there is no performance impact, therefore it may well be that the brain, a product of an evolution during the majority of which a lifespan of 40 years was impressive, has developed a bucket/hash equivalent where it starts getting bad performance from inserts/retrievals around 25/30.
2. As you noted, reinforcement has a great deal to do with it as well, people at 25 have more cross-references to use to support a memory than 16 year olds, plus the fact that 16 year olds are usually swimming in hormones which may well affect certain varieties of memory.
This is not to say that I don't totally agree with you on your picture of the brain, however I see no reason from your arguments that the comparison with a hash table in a simplistic fashion is not valid.
Furthermore, the point I was trying to make in the end is this: I don't believe reviving a bunch of previously dead braincells would achieve anything towards increasing mental performance. The functions and structures of the brain by the time this is relevant are already in place, and the sudden addition of a bunch of braincells would, IMHO, be like malloc()'ing 200kb more space for your hash bucket points, a waste of time at best, since the hash algorithm never goes there anyway.
In a system as complex as the brain, it is quite possible that such a revival would have seriously adverse effects as the new braincells struggle the adjust to the weightings of the network around it, often firing spontaneously where previously there was no fire at all, and in large numbers, this could do extremely weird, if not entirely bad, things.
In the single case of the treatment of certain brain diseases, I believe it may be effective, because entire sections of the brain may die out, and having anything here is better than nothing, but for the case of mental capability, I suspect the technology is of little use.
I had always been of the belief that the reason learning ability and linking slows down as one ages is not due to the loss of brain cells as such, but more the fact that you've already got a lot of information stored in there and it takes longer to make leaps and jumps around in it.
For a programming analogy take your average hash table, the first bunch of inserts and recalls are 1 step, because the bucket at the hash point is empty, so its the first reply you get, but as you fill up a hash table, you begin to get collisions, so you have to do step-searches through the buckets, or jump to overflow lists etc, slowing everything down. The hash is more useful because it contains more data, but to get any data out of it takes a longer time, and to insert new data in also takes a longer time.
Admittedly the brain is of a considerably different structure to that of your average hash table, but it seems an appropriate analogy to me.
Having read a significant proportion of this discussion so far and seen a significant number of people arguing against M2, I'd like to put in an argument for it.
M1 is in place for several reasons:
1. Removing spam and flames
2. Adding notice to well written articles
Where exactly does this fail? several points have been raised:
1. People push their own agenda, moderating up articles by friends or whose topic they agree with, rather than the reasoning behind it.
2. Its a simple fact that some people have very different views from others.
In a collaborative moderation scheme such as this, it is entirely necessary to limit these bad effects as much as possible. The first attempt here on slashdot was to minimise the damage moderators could do, by giving them an extremely limited amount of power.
I believe M2 represents a better solution. While it cannot have effect instantly, since it will take a short amount of time for karma ratings to settle properly and for people to get used to the system, it now gives us the ability to not only limit those moderators who moderate inappropriately, but also to give more power to those who regularly moderate well, thus reducing the need to have more and more people moderate (reducing its special status and making it a chore) while at the same time, increasing the likelyhood that any given moderation point awarded is from a moderator whose moderating methods meet the approval of the community.
In this case, the presence of M3 or higher would never be necessary, M2 represents the will of the community as a whole, and M1 represents the average of that will, with a human intellect.
All in all, an excellent system, and as a collaborative system (as opposed to personal moderation, which is another issue entirely) I believe it only requires some tweaking to achieve excellent results.
I did say heavily dynamic. Very complex pages with realtime generated graphs and tabled data, pulled from SQL databases. Include on top of this user accounting and tracking, a large degree of modularity within the code due to the complexity of the system etc, and 7 hits a second is, in my opinion anyway, quite impressive. A significant number of tricks were utilised to decrease the CPU hit including an httpd accelerator (squid in reverse), various partial pregeneration of pages where possible, and rewrites to cached copies of graphs and data where possible.
It was written in PHP, and I'm more than happy with the performance, in fact (I'm not the author of much of the code itself) I'm more often than not, very impressed with the speed with which it operates. I am looking forward to being able to move it over to Zend, this should increase our capacity even more, without any hardware additions.
While serving up plain HTML is no biggie, and any old box will do for that, serving dynamic content can be orders of magnatude harder. If you have heavily dynamic material (And your concept suggests that perhaps you do) then you will need a fairly capable box in order to respond quickly to requests. It is here and in bandwidth that the critical bottlenecks lie.
/. effect generates :)
To give a suggestion of the CPU power required, the company I work for has several heavily loaded servers:
A celeron 350/128mb ram, maxes out at approximately 7 hits/second (Heavily dynamic material)
A celeron 350/128mb ram, maxes out at ~17hits/sec (Quite heavily dynamic material).
I just don't know how many hits/sec the
Note that these servers have been specially configured to handle the traffic involved, it is unlikely that you will go to the same levels of specialisation, so leave some extra space.
It seems like a remarkable conclusion. I don't believe that they have really put enough detail into the article to judge it correctly. I make the claim that vision cannot be entirely serial, I play an arcade game called House of the Dead in two player mode by myself, and having played this for some time now, I have little or no difficulty maintaining accurate aim from both guns on varying targets on the screen.
While it may well be true that the highest level of vision is serial, this particular level of vision must be quite tightly defined, for, going back to the House of the Dead example, I always shoot for the head, which is by no means just a simple object recognition in such a game.
I suspect more research really needs to be done in the area, and more importantly, that conclusions need to be very accurately defined, rather than making such broad statements.
I don't believe this is an option, since the SSI appears (as far as I can tell) to be simply including a text file generated by some other script.
Trying 169.207.154.108... /cgi-bin/cachemgr.cgi?wtf=9 HTTP/1.1
Connected to crack.linuxppc.org.
Escape character is '^]'.
GET
Host: crack.linuxppc.org
Connection closed by foreign host.
Whats the deal there? no perm-denied..no no-such-page. Is this definitely a stock install?
Ok, here's the deal as far as I'm aware. I'm loath
To report it as a bug, since its only a bug with
the binary, not with compiled source.
The libstdc++ library contains a symbol, __iostream8 or similar, which normally only has
one "_" in front of it. However the library which
is being compiled against by the binary builder
is an old, broken one, which gives this particular
symbol TWO underscores. Both Debian and Redhat
appear to distrubute, as part of the standard distrubution, a copy of the old, broken, version
as well as the newer versions which fix the bug. Thus the binaries work fine on them, but not on slackware and possibly a couple of other distributions, because they DON'T include a copy of the broken lib. There isn't anything you can do
afaik, Your only option is to attempt to compile the source yourself, and thus gain a binary which uses the correct symbol for iostream8