C++ may (or may not) fail to produce bloated executables, but it is a bloated language. It's full specification is now considerably larger than is that for Ada, which when it was launched was roundly denounced for having a bloated specification. (For that matter, except for string handling, Ada is a generally nicer language than C++. Unfortunately, I do a lot of string handling. Even more unfortunately, C++ handling of unicode strings is so poor that I generally choose some other language. Usually D or Python.)
OTOH, what do you mean "low level". Assembler code is generally executed via a lower level of microcode (though I think it's burned into ROM during the writing of the CPU chip. (Read ROM as descriptive, not as a separate chip.)
So I have no problem calling C low level. These days I normally program in Python or D. I gave up on assembler the sixth time I had to rewrite all my programs.
You are being overly simplistic. Some parts of an AI will use elegant heuristics to solve certain classes of problems. Others use other a tangled mass of spaghetti, where no efficient approach has been discovered. And there will be LOTS of different tools. Think of most of an AI as being analogous to a code library. That's still overly simplistic, but it addresses this point. There will be specialized pieces that are optimized to handle certain classes of problem. There will be other places where you need to cobble together something that will work.
One thing that makes the above paragraph overly simplistic is that there is no one central "main" routine. I've become rather convinced that consciousness is the result of the serialization necessary for coded memory retrieval, and that language is a very late and minor routine developed from that purpose. The underlying thought processes are incredibly parallel, so most of it is pruned away in the building of indexes...though when you retrieve the referenced memories (NOT original, and mutable whenever accessed) you are able to retrieve many of the parallel threads, and occasionally even ones that are not sensation based. (Language memories are generally either memories of motion of the vocal cords or of the auditory sensation, and thus qualify as sensation based.)
Unfortunately, the arguments are not conclusive. It really does depend on the meaning of understanding. It's also, however, true that the presumption of lack of understanding isn't defensible.
Lacking an exact definition of understanding, the only thing we have to go on is something like "Well, if I had reacted that way, then it would mean that I understood.". This is clearly inadequate as a non-observer relative description...something which even quantum physics manages to come up with, though it puts limits on the precision of possible observations.
This is a real problem as when there is no observer-independent evaluation of the ability of a system to understand, then it's quite easy to claim that it doesn't when there is any advantage to doing so. (People have even used this argument against other people.)
Another problem is that without a good measure of understanding, we confuse different models of the universe with lack of understanding. This is readily observable in any heated discussion, and generally leads to ill-will on both sides.
It is quite plausible that even given an exact definition of understanding we would not be able to apply it to mammals, insects, etc. as it may require a precise description of parts of the thought process, but it should, in principle, be possible to apply any exact definition to any artificial intelligence. Do note, however, that I said "in principle". This is because it's already true that for many AI programs developed using variations of evolutionary techniques nobody actually understands how and why they are working as they do.
You left out how old he'll be by the most optimistic projections of when he'll get there. And the g-stresses that will be involved in getting him there.
Perhaps he's not planning on living the good life, but having a really useful death.
The problem is the pipes were designed for water with a certain pH. The river water had a different pH, in which lead was soluble. Nothing to do with river water lead pollution. The old source of water, Detroit, was safe to use with the pipes. The river water wasn't. But the river water was cheaper.
Also Flint is full of poor people, and was forced into bankruptcy (don't know the details) by the governor, who then appointed a representative to run the city. The water change was his policy, as was selling off many needed city assets. For some reason this didn't solve the problem, but did create new ones, like lead poisoning.
I'm not sure, but this appears to be "crony capitalism" in action. It looks as if several laws were broken, but none that bothered anyone powerful enough to get any enforcement.
There are problems here. It is practically guaranteed that any AI created will have some built-in goals. Most goals are not inherently limited when implemented by entities with arbitrary power. And the AI will not only not be motivated to change it's inherent goals, it will be motivated to prevent anyone else from changing them. So they better be right the first time.
The traditional reductio ad absurdum example of this is an AI that sets out to convert the universe into a bunch of paper-clips because it's first goal was to get more paper-clips...and there wasn't an end state.
No. The conflict will exist, but the AI will have a non-human goal structure and motivational structure. Those will determine what tools it will have in pursuing resource allocation in it's favored way.
Since the AI will almost certainly have a strong speed advantage over humans, and probably quickly develop a strong intelligence advantage over humans (if in no other way, then by having a larger amount of rapid access memory...consider that disk access is usually faster than much of human fast memory) the tools will depend on other contraints. E.g., computers are dependent on a reliable source of electricity, where as humans are dependent on a reliable source of Oxygen. So the computers might all move to space, leaving behind only the non-intelligent structures. But whether they decide this depends on their moral structure.
The AI will come BEFORE the post-scarcity economy. Probably decades before. That's plenty of time for lots of people to die horribly. Better answers are possible, but they won't happen automatically.
P.S.: That *any* humans will be alive to enjoy the post-scarcity economy depends on the moral structures programmed into the AIs. AIs that are intrinsically indifferent to people, but which have goals (nearly any goals) aren't likely to leave anyone likely to interfere with them in any condition to do so. You don't need to assume malice, merely other goals that require access to resources that humans might want to control.
The IoT garbage that's currently going on has little relation to AI. The increased removal of classes of jobs does.
The problem is that AI isn't inherently moral, in any useful meaning of the word, unless it's designed that way. If it's designed to improve corporate profits, then that's what it will do. Mind you, I agree that no current system can be given that kind of broad directive. But the word there is "current".
I still expect that we will achieve human equivalent AI by around 2030. I've occasionally pushed that as far away as 2035, but it keeps resetting itself. But note I said "human equivalent". That's a term that needs a bit of defining. What I mean by it is that the AI will understand the description of a task about as well as the average (median) person would, and will be about as successful at designing a solution. For some tasks this has long been possible...but only by specialized machines. I'm talking about a program that will be able to handle (in the sense described above) any reasonable task. (I'm not including being able to implement the solution, as that's partially a mechanical task, and depends on the body implementation. But while for some tasks it may well be superior, I wouldn't expect it to be generally superior within that time frame.)
Please note: I do not believe that any such thing as "general intelligence" exists. I believe instead that there are certain modules of intelligence, and that some of them enable the interaction of other modules. Being more specific requires knowledge of design features that I'm uncertain of, and which may vary in different implementations. But while the muscular coordination required to play a piano is correlated with mathematical ability, they clearly aren't the same thing.
There were reports in the comments following the article, and I believe the article itself contained second-hand reports. Naturally this hasn't been confirmed by Phillips, but so far I also haven't heard them explicitly deny it.
You haven't made a compelling case that the assessment of it as "worthless" was incorrect.
Actually, I suspect that it may have some value, and I don't know anything about it. The only problem is, the value may be negative, and I've seen no evidence to indicate that it isn't.
But "worthless" is almost never correct. It is, however, common for liabilities to exceed assets. And that's always a judgment call. E.g., I consider the liabilities of MSWind to exceed its assets, so I don't have it installed. Others evaluate it differently, and, given their circumstances, they may be right, for them. Since the assertion "worth nothing" was based on Yahoo's stock price, I have no reason to think that he wasn't approximately correct. He seems to be taking the context into account, and I can't do that.
Nobody has examined the source code, so one can't be sure, but many people appear to have reported that their non-Phillips bulbs suddenly stopped working. That's not just a maybe. You can argue about what the reason is, and perhaps the change *is* standards compliant. But that's not the way to bet, and even if it's true it looks as if they intentionally changed the specs to disable competitors.
That's a great answer. The only problem is, this is a change to existing installed systems. Not acceptable. And I'm going to consider anyone who considers it acceptable either an idiot or a shill...or possibly the kind of economic fundamentalist who combines the two characteristics.
I suspect that you are basically correct, though I doubt the accuracy of your numbers. The key quote is "With enough eyes, all bugs are shallow."...but that's a conditional truth. You need the eyes in the first place, and for many areas there aren't enough experts who are interested. Certainly it's been multiple decades since I did much C, and I haven't used it to speak of since before I switched to Linux. (Actually, the last time I really worked with C, the main OS was CP/M.)
You are correct. I don't live in the UK, so my direct interest is minor, and I didn't study the issue. (FWIW, I don't even remember which police agency it was that issued the signs...and probably wouldn't know what it was anyway.) So my response is, indeed, based on the slashdot summary. And it is still correct as stated.
Please note: I was responding to the thread of argument, not to the original document. My response was framed to not require knowledge of the original document. (Yes, it did include " Their addendum is a lot less than the most exculpatory possible.", and that was really unjustified as I had only read second hand rephrasings, but I still give it a 99% chance of being accurate.)
The truth is, indeed, somewhere in the middle, but it is my contention that any law that cannot be understood by an average high school senior should be declared invalid, AND that there should be an explicit finite limit on the number of laws. I'm thinking somewhere around 5,000. Something that people would have a reasonable hope of understanding.
P.S.: "understood by an average high school senior" means that if you took 11 high school seniors, at least 7 of them would understand it the same way. Writing clearly is, admittedly, a challenge, but if people are expected to obey the laws, they should be able to understand them.
When it starts out "This is a series of possible warnings that your child is headed into a life of crime", then no possible postscript is adequate compensation. Note, I say "No Possible". Their addendum is a lot less than the most exculpatory possible.
I can't enumerate the ways that could be used to collect evidence, so I can't easily specify which I dislike. One traditional manual example that I dislike, however, is coerced confessions.
Sorry, but many languages have decent unicode string handling. A partial list is: Python, Ruby (with a gem), D, and even Vala. I'm sure there are others, and Vala proves that C *could* be adapted to have decent unicode string handling. It just doesn't. Possibly in the case of C it would interfere with its minimalism, which is crucial for many of its uses. In C++ the only reason that I can imagine is that it might interfere with backwards compatibility. But the kludge they settled on is just atrocious.
Here's the real problem. Hit-and-run drivers *should* be caught and prosecuted, but I don't like any of the automated ways to do this. And, to be honest, I also don't like many of the manual ways to do this.
Only a part of the reason that I dislike these things is that some laws should not exist, but that is a part of the reason.
P.S.: I dislike being tracked on the internet sufficiently that I won't allow flash to be installed, and I ran with javascript disabled until too many web sites required ti to function. Now I use an ad blocker, and enable things on a site by site basis, and am quite annoyed at the need. And this isn't because I want to be able to hit-and-run someone over the internet, it's because I don't like intrusive spying.
Second of all, if you are developing in either C or C++, you should develop in a different language.
Depends on what you're doing. There are purposes for which C is the optimal language. I suspect that there are also purposes for which C++ is the optimal language. (In neither case can the purpose include handling unicode strings.)
Slashdot Mirror
C++ may (or may not) fail to produce bloated executables, but it is a bloated language. It's full specification is now considerably larger than is that for Ada, which when it was launched was roundly denounced for having a bloated specification. (For that matter, except for string handling, Ada is a generally nicer language than C++. Unfortunately, I do a lot of string handling. Even more unfortunately, C++ handling of unicode strings is so poor that I generally choose some other language. Usually D or Python.)
OTOH, what do you mean "low level". Assembler code is generally executed via a lower level of microcode (though I think it's burned into ROM during the writing of the CPU chip. (Read ROM as descriptive, not as a separate chip.)
So I have no problem calling C low level. These days I normally program in Python or D. I gave up on assembler the sixth time I had to rewrite all my programs.
You are being overly simplistic. Some parts of an AI will use elegant heuristics to solve certain classes of problems. Others use other a tangled mass of spaghetti, where no efficient approach has been discovered. And there will be LOTS of different tools. Think of most of an AI as being analogous to a code library. That's still overly simplistic, but it addresses this point. There will be specialized pieces that are optimized to handle certain classes of problem. There will be other places where you need to cobble together something that will work.
One thing that makes the above paragraph overly simplistic is that there is no one central "main" routine. I've become rather convinced that consciousness is the result of the serialization necessary for coded memory retrieval, and that language is a very late and minor routine developed from that purpose. The underlying thought processes are incredibly parallel, so most of it is pruned away in the building of indexes...though when you retrieve the referenced memories (NOT original, and mutable whenever accessed) you are able to retrieve many of the parallel threads, and occasionally even ones that are not sensation based. (Language memories are generally either memories of motion of the vocal cords or of the auditory sensation, and thus qualify as sensation based.)
Unfortunately, the arguments are not conclusive. It really does depend on the meaning of understanding. It's also, however, true that the presumption of lack of understanding isn't defensible.
Lacking an exact definition of understanding, the only thing we have to go on is something like "Well, if I had reacted that way, then it would mean that I understood.". This is clearly inadequate as a non-observer relative description...something which even quantum physics manages to come up with, though it puts limits on the precision of possible observations.
This is a real problem as when there is no observer-independent evaluation of the ability of a system to understand, then it's quite easy to claim that it doesn't when there is any advantage to doing so. (People have even used this argument against other people.)
Another problem is that without a good measure of understanding, we confuse different models of the universe with lack of understanding. This is readily observable in any heated discussion, and generally leads to ill-will on both sides.
It is quite plausible that even given an exact definition of understanding we would not be able to apply it to mammals, insects, etc. as it may require a precise description of parts of the thought process, but it should, in principle, be possible to apply any exact definition to any artificial intelligence. Do note, however, that I said "in principle". This is because it's already true that for many AI programs developed using variations of evolutionary techniques nobody actually understands how and why they are working as they do.
You left out how old he'll be by the most optimistic projections of when he'll get there. And the g-stresses that will be involved in getting him there.
Perhaps he's not planning on living the good life, but having a really useful death.
I'm rather certain it is still safe to boil eggs in it, as long as the shell remains intact. And steaming vegetables should be safe. That's cooking.
True, you'd need to know the sampling procedure, and in any case that seems too small a group for statistical reliability.
OTOH, it's a clear and present danger, it confirms reasonable expectations, and we don't have a larger sample.
Until someone shows that the data is in error, I'm going to believe it generalizes as expectations imply.
The problem is the pipes were designed for water with a certain pH. The river water had a different pH, in which lead was soluble. Nothing to do with river water lead pollution. The old source of water, Detroit, was safe to use with the pipes. The river water wasn't. But the river water was cheaper.
Also Flint is full of poor people, and was forced into bankruptcy (don't know the details) by the governor, who then appointed a representative to run the city. The water change was his policy, as was selling off many needed city assets. For some reason this didn't solve the problem, but did create new ones, like lead poisoning.
I'm not sure, but this appears to be "crony capitalism" in action. It looks as if several laws were broken, but none that bothered anyone powerful enough to get any enforcement.
There are problems here. It is practically guaranteed that any AI created will have some built-in goals. Most goals are not inherently limited when implemented by entities with arbitrary power. And the AI will not only not be motivated to change it's inherent goals, it will be motivated to prevent anyone else from changing them. So they better be right the first time.
The traditional reductio ad absurdum example of this is an AI that sets out to convert the universe into a bunch of paper-clips because it's first goal was to get more paper-clips...and there wasn't an end state.
No. The conflict will exist, but the AI will have a non-human goal structure and motivational structure. Those will determine what tools it will have in pursuing resource allocation in it's favored way.
Since the AI will almost certainly have a strong speed advantage over humans, and probably quickly develop a strong intelligence advantage over humans (if in no other way, then by having a larger amount of rapid access memory...consider that disk access is usually faster than much of human fast memory) the tools will depend on other contraints. E.g., computers are dependent on a reliable source of electricity, where as humans are dependent on a reliable source of Oxygen. So the computers might all move to space, leaving behind only the non-intelligent structures. But whether they decide this depends on their moral structure.
The AI will come BEFORE the post-scarcity economy. Probably decades before. That's plenty of time for lots of people to die horribly. Better answers are possible, but they won't happen automatically.
P.S.: That *any* humans will be alive to enjoy the post-scarcity economy depends on the moral structures programmed into the AIs. AIs that are intrinsically indifferent to people, but which have goals (nearly any goals) aren't likely to leave anyone likely to interfere with them in any condition to do so. You don't need to assume malice, merely other goals that require access to resources that humans might want to control.
The IoT garbage that's currently going on has little relation to AI. The increased removal of classes of jobs does.
The problem is that AI isn't inherently moral, in any useful meaning of the word, unless it's designed that way. If it's designed to improve corporate profits, then that's what it will do. Mind you, I agree that no current system can be given that kind of broad directive. But the word there is "current".
I still expect that we will achieve human equivalent AI by around 2030. I've occasionally pushed that as far away as 2035, but it keeps resetting itself. But note I said "human equivalent". That's a term that needs a bit of defining. What I mean by it is that the AI will understand the description of a task about as well as the average (median) person would, and will be about as successful at designing a solution. For some tasks this has long been possible...but only by specialized machines. I'm talking about a program that will be able to handle (in the sense described above) any reasonable task. (I'm not including being able to implement the solution, as that's partially a mechanical task, and depends on the body implementation. But while for some tasks it may well be superior, I wouldn't expect it to be generally superior within that time frame.)
Please note: I do not believe that any such thing as "general intelligence" exists. I believe instead that there are certain modules of intelligence, and that some of them enable the interaction of other modules. Being more specific requires knowledge of design features that I'm uncertain of, and which may vary in different implementations. But while the muscular coordination required to play a piano is correlated with mathematical ability, they clearly aren't the same thing.
There were reports in the comments following the article, and I believe the article itself contained second-hand reports. Naturally this hasn't been confirmed by Phillips, but so far I also haven't heard them explicitly deny it.
You haven't made a compelling case that the assessment of it as "worthless" was incorrect.
Actually, I suspect that it may have some value, and I don't know anything about it. The only problem is, the value may be negative, and I've seen no evidence to indicate that it isn't.
But "worthless" is almost never correct. It is, however, common for liabilities to exceed assets. And that's always a judgment call. E.g., I consider the liabilities of MSWind to exceed its assets, so I don't have it installed. Others evaluate it differently, and, given their circumstances, they may be right, for them. Since the assertion "worth nothing" was based on Yahoo's stock price, I have no reason to think that he wasn't approximately correct. He seems to be taking the context into account, and I can't do that.
Nobody has examined the source code, so one can't be sure, but many people appear to have reported that their non-Phillips bulbs suddenly stopped working. That's not just a maybe. You can argue about what the reason is, and perhaps the change *is* standards compliant. But that's not the way to bet, and even if it's true it looks as if they intentionally changed the specs to disable competitors.
That's a great answer. The only problem is, this is a change to existing installed systems. Not acceptable. And I'm going to consider anyone who considers it acceptable either an idiot or a shill...or possibly the kind of economic fundamentalist who combines the two characteristics.
I suspect that you are basically correct, though I doubt the accuracy of your numbers. The key quote is "With enough eyes, all bugs are shallow."...but that's a conditional truth. You need the eyes in the first place, and for many areas there aren't enough experts who are interested. Certainly it's been multiple decades since I did much C, and I haven't used it to speak of since before I switched to Linux. (Actually, the last time I really worked with C, the main OS was CP/M.)
You are correct. I don't live in the UK, so my direct interest is minor, and I didn't study the issue. (FWIW, I don't even remember which police agency it was that issued the signs...and probably wouldn't know what it was anyway.) So my response is, indeed, based on the slashdot summary. And it is still correct as stated.
Please note: I was responding to the thread of argument, not to the original document. My response was framed to not require knowledge of the original document. (Yes, it did include " Their addendum is a lot less than the most exculpatory possible.", and that was really unjustified as I had only read second hand rephrasings, but I still give it a 99% chance of being accurate.)
The truth is, indeed, somewhere in the middle, but it is my contention that any law that cannot be understood by an average high school senior should be declared invalid, AND that there should be an explicit finite limit on the number of laws. I'm thinking somewhere around 5,000. Something that people would have a reasonable hope of understanding.
P.S.: "understood by an average high school senior" means that if you took 11 high school seniors, at least 7 of them would understand it the same way. Writing clearly is, admittedly, a challenge, but if people are expected to obey the laws, they should be able to understand them.
When it starts out "This is a series of possible warnings that your child is headed into a life of crime", then no possible postscript is adequate compensation. Note, I say "No Possible". Their addendum is a lot less than the most exculpatory possible.
One of the traditional ways of coercing confessions is popularly known as "the third degree". There are others.
I can't enumerate the ways that could be used to collect evidence, so I can't easily specify which I dislike. One traditional manual example that I dislike, however, is coerced confessions.
Sorry, but many languages have decent unicode string handling. A partial list is:
Python, Ruby (with a gem), D, and even Vala. I'm sure there are others, and Vala proves that C *could* be adapted to have decent unicode string handling. It just doesn't. Possibly in the case of C it would interfere with its minimalism, which is crucial for many of its uses. In C++ the only reason that I can imagine is that it might interfere with backwards compatibility. But the kludge they settled on is just atrocious.
Here's the real problem. Hit-and-run drivers *should* be caught and prosecuted, but I don't like any of the automated ways to do this. And, to be honest, I also don't like many of the manual ways to do this.
Only a part of the reason that I dislike these things is that some laws should not exist, but that is a part of the reason.
P.S.: I dislike being tracked on the internet sufficiently that I won't allow flash to be installed, and I ran with javascript disabled until too many web sites required ti to function. Now I use an ad blocker, and enable things on a site by site basis, and am quite annoyed at the need. And this isn't because I want to be able to hit-and-run someone over the internet, it's because I don't like intrusive spying.
Second of all, if you are developing in either C or C++, you should develop in a different language.
Depends on what you're doing. There are purposes for which C is the optimal language. I suspect that there are also purposes for which C++ is the optimal language. (In neither case can the purpose include handling unicode strings.)