There is no "new paradigm" that could do what you predict. There are hard, established theoretical results that say that many important algorithms cannot benefit from more cores. Does nobody know the basics anymore?
Yes, so? That is what you say and it has nothing to do whith what I just said. For most tasks, it does not matter how good the implementation is, more cores do not help.
You seem to know nothing about algorithm performance analysis. Please stop claiming nonsense. You are just disgracing yourself.
The fact of the matter is that for many important classes of algorithms, using more cores slows things down and that is a hard, theoretical fact, so there is no way around it. For many others, you get speed-ups per additional core that are so bad that the communication overhead kills all advantages. This has been known for something like 50 years or longer.
Well, we should cross that bridge when we get to it. That will very likely not happen. So far QC has scaled extremely badly over the decades. A sub-linear scaling seems to be what is going on. Even if that scaling continues (unlikely) and does not hit a wall, classical computers will stay faster for many decades.
Also, your "predictive trees" are nonsense, at least as a QC application. Stop believing the hype.
Sounds more to me that he cannot let go of the old, failed ideas and face the reality that there is no silver bullet. There is plenty of aging CS professors around with that problem.
Boy, do I miss the old MC68000, where I could just look up how long something takes and could do actually fully synchronous coding for a significant performance boost.
We have hit that wall. This technology is mature and everything it can do well, it can do now at close to maximum-possible speeds. Sure, software sucks today and coders are mostly incompetent and there is some speed increase to be expected from that angle, but that is it. That there was an area of seemingly exponential growth does in no way imply it will continue without limits. And it does not.
It does not really matter, because we will not get it going much faster than today anyways. There is no "new architecture" or "new language" that will change that. Massive parallel systems failed last century and did so several times. Vector architectures have just hit the same brick wall as conventional ones. There really is nothing else.
Various alternate architectures have been tried out over the decades. A lot of other programming models have been tried out as well. They all basically failed or live on only in niches because people could not hack coding for them.
Performance increases for most tasks are over. Deal with it and stop proposing silver bullets. It only makes you look stupid.
You know what most of modern CS is based on? Mathematics, sometimes done thousands of years ago. This is relevant and that is why we keep these people around.
Only works for fully formalized proofs and a human has to explain the proof to the system in detail. That means it takes a lot of time to do and needs a human that fully understands the proof.
Also, this is not actually AI in any meaningful sense, it only gets in there because of the AI hype.
Well. I agree to your points. While hinting at things may or may not work, saying them clearly does certainly not work. Personally, I have mostly given up on people and say what I think clearly now. Fortunately, not many even listen, so the risk for me is small. And yes, that one guy you could (maybe) get rid of is only a symptom. Actually getting rid of him would not solve anything.
I agree. The "think of the women" argument is just a smokescreen. This is about power, plain and simple and it is by people that want power without having done anything to merit it. As they cannot get it in any legitimate or honorable way, they try to steal it. Unfortunately, they seem to have success with that.
We will see. Even if Linux has a lot of institutional support, it can still be forked successfully by a dozen competent and dedicated people that want no part of this disgusting nonsense.
Looks like it. Maybe he also though that his creation could float successfully without him for quite a while even while under attack by these retrogrades, and if so, I think he is right. Even if these people manage to destroy the kernel in the next 20 years or so, they will not destroy the Unix API and most FOSS software is not tied to Linux specifically. Things that are (think, for example, the systemd atrocity) are of no or negative value generally anyways.
We are talking about pretty smart people here. There will not be a lot of effect in this "pushing" and it is still _their_ decision, not anybody else's.
Slashdot Mirror
There is no "new paradigm" that could do what you predict. There are hard, established theoretical results that say that many important algorithms cannot benefit from more cores. Does nobody know the basics anymore?
The article is just dumb.
Yes, so? That is what you say and it has nothing to do whith what I just said. For most tasks, it does not matter how good the implementation is, more cores do not help.
Then they each get less performance than if they had their own computer due to various bottle-necks. That is not scaling.
You seem to know nothing about algorithm performance analysis. Please stop claiming nonsense. You are just disgracing yourself.
The fact of the matter is that for many important classes of algorithms, using more cores slows things down and that is a hard, theoretical fact, so there is no way around it. For many others, you get speed-ups per additional core that are so bad that the communication overhead kills all advantages. This has been known for something like 50 years or longer.
Well, we should cross that bridge when we get to it. That will very likely not happen. So far QC has scaled extremely badly over the decades. A sub-linear scaling seems to be what is going on. Even if that scaling continues (unlikely) and does not hit a wall, classical computers will stay faster for many decades.
Also, your "predictive trees" are nonsense, at least as a QC application. Stop believing the hype.
Magic?
No. Not going to materialize and even if it did, it would not accelerate most tasks, just a small set of very specific ones.
I completely agree to that.
Turns out that in this the past is not a reliable predictor of the future, as is true in any other area.
There really is nothing more to add.
Sounds more to me that he cannot let go of the old, failed ideas and face the reality that there is no silver bullet. There is plenty of aging CS professors around with that problem.
More cores are worthless for most tasks. Takes some actual knowledge to see that though.
Boy, do I miss the old MC68000, where I could just look up how long something takes and could do actually fully synchronous coding for a significant performance boost.
We have hit that wall. This technology is mature and everything it can do well, it can do now at close to maximum-possible speeds. Sure, software sucks today and coders are mostly incompetent and there is some speed increase to be expected from that angle, but that is it. That there was an area of seemingly exponential growth does in no way imply it will continue without limits. And it does not.
It does not really matter, because we will not get it going much faster than today anyways. There is no "new architecture" or "new language" that will change that. Massive parallel systems failed last century and did so several times. Vector architectures have just hit the same brick wall as conventional ones. There really is nothing else.
Various alternate architectures have been tried out over the decades. A lot of other programming models have been tried out as well. They all basically failed or live on only in niches because people could not hack coding for them.
Performance increases for most tasks are over. Deal with it and stop proposing silver bullets. It only makes you look stupid.
I fail to detect any connection between what I wrote and what you wrote. Are you on drugs?
You know what most of modern CS is based on? Mathematics, sometimes done thousands of years ago. This is relevant and that is why we keep these people around.
In addition, interactive proof assist systems do not even use neural networks.
Only works for fully formalized proofs and a human has to explain the proof to the system in detail. That means it takes a lot of time to do and needs a human that fully understands the proof.
Also, this is not actually AI in any meaningful sense, it only gets in there because of the AI hype.
Well. I agree to your points. While hinting at things may or may not work, saying them clearly does certainly not work. Personally, I have mostly given up on people and say what I think clearly now. Fortunately, not many even listen, so the risk for me is small. And yes, that one guy you could (maybe) get rid of is only a symptom. Actually getting rid of him would not solve anything.
About time.
I agree. The "think of the women" argument is just a smokescreen. This is about power, plain and simple and it is by people that want power without having done anything to merit it. As they cannot get it in any legitimate or honorable way, they try to steal it. Unfortunately, they seem to have success with that.
We will see. Even if Linux has a lot of institutional support, it can still be forked successfully by a dozen competent and dedicated people that want no part of this disgusting nonsense.
Looks like it. Maybe he also though that his creation could float successfully without him for quite a while even while under attack by these retrogrades, and if so, I think he is right. Even if these people manage to destroy the kernel in the next 20 years or so, they will not destroy the Unix API and most FOSS software is not tied to Linux specifically. Things that are (think, for example, the systemd atrocity) are of no or negative value generally anyways.
We are talking about pretty smart people here. There will not be a lot of effect in this "pushing" and it is still _their_ decision, not anybody else's.