Yup, and I think all of it can be tracked back to the tyranny of metrics (not the book, the subject). When "doing good science" became formalized into number of publications, or h factors, or any other system, users (scientists most interested in career advancement) gamed the system. And so we now have reams of papers that explain essentially nothing. But we have to read orders of magnitude more just to stand still, and are certainly going to miss the basic fact that it's already been done. But no matter, papers don't get retracted for reporting mild twists on unacknowledged old discoveries, so onward! And career advancement for the savvy operator is a feedback loop--they consume ever more students and postdocs, to do things that have already been done or are objectively worthless, shit out ever more papers, and are judged only on the gamed metrics. They gain editorships, at which point the lowly truth-seeker had better cite their work if they want to get their own work published! And higher their metrics go!
The kicker is that there is good stuff being done in the trenches, which in a pinch is held up as proof the broken system works. But it gets done in spite of the system, not because of it. And given our obsession with metrics, the same can be said of pretty much everything in modern society (American, at least, can't speak for others). So-called leaders, titans of industry, technology, economics--all falling down. But (insert your favorite politician), FAANG, 5G, "unemployment" figures--I must be wrong. And on it goes...
The phrase "faster than light" is kinda misleading in these discussions. An entangled pair of things can't be expressed as a product state; from a certain point of view, it is no longer two things. So even though you can poke here and apparently have an effect there, it's one object (composed of two strongly coupled objects) reacting.
The division into photons 1 and 2 is artificial when they're entangled. Any "impulse" that were traveling wouldn't be doing so through the vacuum or environment between the photons. So, is it "faster than light"? Only from the fiction that something must move through the space between them.
More so, the lack of protections for shouting fire seem to rest on the properties of (a) panic, AND (b) falsehood. I suspect one's shouting of "fire" in a crowded theater would indeed be protected, if there were in fact a fire that threatened everyone in the theater.
Conservatives have gleefully embraced the alternative reality of Hannity, Jones, et al. as it suited them. That the intrinsic falsehood of that alternative world should lead to the actual world filtering it out is just the other shoe dropping. Don't like it that reality has a liberal bias? Tough.
Of those 8.3% who said they could name a famous woman tech leader, only 4% actually could -- and a quarter of those respondents named "Siri" or "Alexa." Now, granted, this represents only about 10 people
1000 * (0.083 * 0.04) * 0.25 = 0.83, or perhaps 1 person, if math illiterate.
1000 * 0.04 * 0.25 = 10 as stated, if just plain illiterate. Even so, if they identified Siri or Alexa, then this 4% could not in fact identify a famous female tech leader, as these two names are not tech leaders.
So, what happened again? The inability to express the issue clearly kinda takes the sting out of the critique.
Besides, the more alarming thing was that all the other active respondents identified Tay.:)
No more lies! We'll see the mic booms on the "moon landing" footage, chemtrails will be exposed as whites-only obesity-promoting chemicals, and we'll learn the true extent of the HAARP array's mind control powers!
...Or did they mean fighting actually fake news? Pphht, doesn't seem like a very revenue-positive thing to do.
make the concept of running a Linux distro natively a thing of the past
It will be a thing of the past. You will run a Linux distro in a VM on Linux in the cloud, or a Linux distro in a VM on Azure, or a Linux distro in a VM on Android, or something else. For the most part, who cares? Of course for the really performance-minded, no, Linux-on-metal isn't going away anytime soon.
As for whether a company can extinguish the world's technical backbone, I doubt it without extinguishing the concept of FOSS altogether, which they would have done long ago if they could.
which it plans to combine with its AOL assets into a subsidiary called Oath, covering some 50 media brands and 1 billion people globally
Thank goodness Verizon was finally able to buy the dog shit for their Central Park diorama, because it just isn't a park without a healthy helping of dog turds.
OK, maybe Finance is worth something, possibly Sports I imagine, but what else does Yahoo bring? BRAND recognition? A savvy user base? Maybe they're conflating (defunct) accounts with "people"?
in its attempts to remain apolitical and objective, the march focused primarily on funding and communication aspects of its mission
..., thereby succeeding in its goals to remain apolitical and objective.
There is plenty to support in "social justice," but injecting it into every discussion regardless of scope is dumb. And for a self-styled scientist to do so brings into doubt their credibility as a scientist. A current pet peeve is claims that diverse groups are necessarily better, for example this one . The danger of this over-enthusiastic embrace of affirmative action is the contrapositive--if diverse teams are better/smarter/Xer, then less diverse teams are necessarily less good/smart/X. If that were true, then I can say that whites/males/{over-represented group of choice} are either demonstrably less X as individuals, or that they are demonstrably less capable of having team interactions that lead to better X. Once that door is open to racial/gender/etc discrimination, the barrel can be turned right back onto the very groups that were hoping to benefit from greater diversity.
Not to mention the poor sampling of diversity dimensions. Gender and race have some representation in the diversity debate, but what about LGBTQ? Geographic? Philosophical? Political?
Affirmative action is a defensible temporary policy to correct certain historical discriminations, given the intensity and horror of them. But when it becomes unquestioned policy applied across the board, it is an outright danger to the very things it attempts to fix.
Until the funding source to maintain the site runs dry, or vested interests persuade certain Congressional parties to defund it.
We had developed software under the DOE SciDAC program, which was distributed via an Outreach site. When funding got tight, that site was the first to go. Last I checked, the lone guy holding the line was keeping it up until the server died, at which point it was gone for good. If you want persistence, better make sure not to rely on a single point of failure. And these days, any.gov domain is prone to failure and/or corruption.
Well-written C can be very readable and maintainable. Machine-generated C rarely is.
Totally agree, based on the source I've seen dumped out by various tools. However, the question is whether it needs to be that way. I don't think it does, and that comes down to the research part. I know "Deep Neural Net" is treated like magic pixie dust to solve everything, but what if unreadable machine-generated code is a non-linear transformation away from readable, maintainable code? If it is, then would COBOL-to-maintainable-C be impossible? If we can't answer that with more than opinion, then it may be worth taking a look at seriously.
Yes, I realize the 95% correctness of a trained NN would result in awful correctness checking exercises, and there would undoubtedly be labor-intensive steps, but if it's that or "let the system collapse or rot", I'd choose moving ahead. Then again, maybe the problem's not that serious.
So true! Assemble multiple rockets in orbit, lash 'em together, and go. That 10-rockets-to-accelerate-no-deceleration plan might just get someone onto the surface of Mars in jellied form by his impeachment trial end date, if we start now.
Funding academic CS work on COBOL-to-C, or maybe even COBOL-to-C++, source transformations. They could start with any of the hits a "COBOL to C" Google search provides.
And just how the hell will I now forget to put a receptacle in place and dispense $5 of pre-made juice onto my kitchen counter by phone? BY PHONE, do you understand?? This is Internet-connected! IoT! AI! Deep Learning!
I'll bet it could order its own $1000 worth of juice refills from Amazon automatically, too, dammit.
The success marks a new era for the large-scale design of new magnetic materials
No, it doesn't. This is screening, and regardless of how much those in the field of drug "design" and materials "design" use those words, it's not Design, and it's not Engineering. It's Discovery, and it's great that physics and computation have gotten to the point where we can actually discover useful things in silico much faster than at the bench. But engineering requires an understanding of the underlying relationship between materials composition and desired quantitative property, and that is largely still lacking. If it weren't, you'd be screening 10s of compounds, not ~10^5.
Discovery and prediction are fine, no one will complain about the gold you get from panning for it. But don't dress it up in misleading language--true materials engineering and design are still a long way off.
But this won't happen -- or shouldn't happen -- unless we find ways of making techniques like deep learning more understandable to their creators and accountable to their users. Otherwise it will be hard to predict when failures might occur -- and it's inevitable they will.
Sounds a lot like humans. An observer has no hope of understanding why I make a decision, beyond shared social convention. And if they want to understand the process mechanistically, following impulses around the 1e14 estimated neural connections in a human brain? Forget it.
I agree that it's good to understand how a tool works, but we'll accept the deployment of these tools for the same reasons we accept our fellow beings hurtling around in 2+ ton wheeled projectiles--because most of the time, there isn't a problem, and more is gained from taking the risk than is lost from avoiding it. Legal responsibility needs to be made clear first, but as long as someone pays for fuck ups, probably OK
Maybe they should just introduce a troll icon, and a bot icon, along with a rating system a la Slashdot. Once a user crosses a threshold on "Troll" or "Bot" ratings, their icon automatically gets changed to one or the other.
In fact, just create icons to go along with every rating category, and change their icon based on how a user is rated.
Actually, come to think of it, Slashdot IS hardened Twitter, without the popularity (or icons).
But details are hard and require knowing stuff, brah. Not. For. Meeeee...
Seriously though, agree totally, but are all those VCs and angels really so brain dead as to not know this? Or are they, like the finance folks capitalizing on herd movement into the stock markets at the moment, just cashing in on the phenomenon until it implodes? I suspect the only ones who aren't in on the joke are the dimwits who genuinely believe their brilliant calorie/counting-or-messaging/Youtube/mashup-or-whatever phone app is going to change the world, and they aren't going to pay much attention to cranky ol' Linus.
...the DVD I received rattling around in the refrigerator-sized shipping box. Insufficient training data on how to ship DVDs to Luddites mistrustful of streaming services.
Oh, good. As someone who has to build applications for users on HPC systems, first I had make. It was simple, things were either in Makefile or not. Drop the compiler options and paths right in. Then I had autotools, where I could pass paths and switches to possibly undocumented options. Still manageable, eventually. Thanks heavens that mechanism to manage an architecture and OS zoo came out as the world consolidated to x86-64 and Linux. Then I had cmake, which I can get to work sometimes. Not to mention all the one-off solutions from individual development groups that just had to improve on the state-of-the-art, because.
Now we'll get to troubleshoot "X not found" from a black-box build system generated by a neural net? Good luck. Nothing leads to "higher quality code" like unmanageable complexity. I kind of look forward to going back to the good old days of one codebase building on the development machine only. I can just about hear the world's productivity dropping from this trendy BS.
The root problem (I speculate) is the same one that afflicts companies looking ahead only to the next quarter, schools teaching only prescribed and minimally challenging material, the slow strangulation of endless safety regulation, etc. We're short-sighted, and can't fathom even a slight risk of negativity. As long as the next increment turns out OK, we figure we'll be just fine. That works, so long as the path you're taking doesn't lead you off a cliff or to stagnation, but if you have to get over some barrier (financial loss, global competition with rising powers, new discoveries) to get onto the optimal track, the purely local gradient-based search won't work. Even if eventual failure of a system is demonstrated, we'll keep doing the same thing because we're too fearful of the unknown to do anything else.
On the other hand, the real breakthroughs have never been supported by conventional thinkers or their backers. Kuhn, etc. Would taxpayers accept 99% of research funding to add up to nothing for the remaining 1% to pay off 1000-fold or more? I doubt it. The angels and VCs might risk those odds, but not the standard research funding apparatus.
Slashdot Mirror
underscored the importance of "credible" information.
https://youtu.be/g5vnZec964c?t...>Amen!
The best way to stop this is via the law, preferably Federal law.
but, but, but, that would be State overreach!
Yup, and I think all of it can be tracked back to the tyranny of metrics (not the book, the subject). When "doing good science" became formalized into number of publications, or h factors, or any other system, users (scientists most interested in career advancement) gamed the system. And so we now have reams of papers that explain essentially nothing. But we have to read orders of magnitude more just to stand still, and are certainly going to miss the basic fact that it's already been done. But no matter, papers don't get retracted for reporting mild twists on unacknowledged old discoveries, so onward! And career advancement for the savvy operator is a feedback loop--they consume ever more students and postdocs, to do things that have already been done or are objectively worthless, shit out ever more papers, and are judged only on the gamed metrics. They gain editorships, at which point the lowly truth-seeker had better cite their work if they want to get their own work published! And higher their metrics go!
The kicker is that there is good stuff being done in the trenches, which in a pinch is held up as proof the broken system works. But it gets done in spite of the system, not because of it. And given our obsession with metrics, the same can be said of pretty much everything in modern society (American, at least, can't speak for others). So-called leaders, titans of industry, technology, economics--all falling down. But (insert your favorite politician), FAANG, 5G, "unemployment" figures--I must be wrong. And on it goes...
The phrase "faster than light" is kinda misleading in these discussions. An entangled pair of things can't be expressed as a product state; from a certain point of view, it is no longer two things. So even though you can poke here and apparently have an effect there, it's one object (composed of two strongly coupled objects) reacting.
The division into photons 1 and 2 is artificial when they're entangled. Any "impulse" that were traveling wouldn't be doing so through the vacuum or environment between the photons. So, is it "faster than light"? Only from the fiction that something must move through the space between them.
So, wormholes, all the way down.
More so, the lack of protections for shouting fire seem to rest on the properties of (a) panic, AND (b) falsehood. I suspect one's shouting of "fire" in a crowded theater would indeed be protected, if there were in fact a fire that threatened everyone in the theater.
Conservatives have gleefully embraced the alternative reality of Hannity, Jones, et al. as it suited them. That the intrinsic falsehood of that alternative world should lead to the actual world filtering it out is just the other shoe dropping. Don't like it that reality has a liberal bias? Tough.
They form a loose but auto-griddleable patty that's then plopped onto the bun before the whole package slides out of the machine
I look forward to sinking my teeth into a burger that could have been cooked, but is apparently not by default.
Of those 8.3% who said they could name a famous woman tech leader, only 4% actually could -- and a quarter of those respondents named "Siri" or "Alexa." Now, granted, this represents only about 10 people
So, what happened again? The inability to express the issue clearly kinda takes the sting out of the critique.
Besides, the more alarming thing was that all the other active respondents identified Tay. :)
No more lies! We'll see the mic booms on the "moon landing" footage, chemtrails will be exposed as whites-only obesity-promoting chemicals, and we'll learn the true extent of the HAARP array's mind control powers!
...Or did they mean fighting actually fake news? Pphht, doesn't seem like a very revenue-positive thing to do.
make the concept of running a Linux distro natively a thing of the past
It will be a thing of the past. You will run a Linux distro in a VM on Linux in the cloud, or a Linux distro in a VM on Azure, or a Linux distro in a VM on Android, or something else. For the most part, who cares? Of course for the really performance-minded, no, Linux-on-metal isn't going away anytime soon.
As for whether a company can extinguish the world's technical backbone, I doubt it without extinguishing the concept of FOSS altogether, which they would have done long ago if they could.
which it plans to combine with its AOL assets into a subsidiary called Oath, covering some 50 media brands and 1 billion people globally
Thank goodness Verizon was finally able to buy the dog shit for their Central Park diorama, because it just isn't a park without a healthy helping of dog turds.
OK, maybe Finance is worth something, possibly Sports I imagine, but what else does Yahoo bring? BRAND recognition? A savvy user base? Maybe they're conflating (defunct) accounts with "people"?
...but don't belong in every discussion.
in its attempts to remain apolitical and objective, the march focused primarily on funding and communication aspects of its mission
..., thereby succeeding in its goals to remain apolitical and objective.
There is plenty to support in "social justice," but injecting it into every discussion regardless of scope is dumb. And for a self-styled scientist to do so brings into doubt their credibility as a scientist. A current pet peeve is claims that diverse groups are necessarily better, for example this one . The danger of this over-enthusiastic embrace of affirmative action is the contrapositive--if diverse teams are better/smarter/Xer, then less diverse teams are necessarily less good/smart/X. If that were true, then I can say that whites/males/{over-represented group of choice} are either demonstrably less X as individuals, or that they are demonstrably less capable of having team interactions that lead to better X. Once that door is open to racial/gender/etc discrimination, the barrel can be turned right back onto the very groups that were hoping to benefit from greater diversity.
Not to mention the poor sampling of diversity dimensions. Gender and race have some representation in the diversity debate, but what about LGBTQ? Geographic? Philosophical? Political?
Affirmative action is a defensible temporary policy to correct certain historical discriminations, given the intensity and horror of them. But when it becomes unquestioned policy applied across the board, it is an outright danger to the very things it attempts to fix.
"Code.gov is here to stay."
Until the funding source to maintain the site runs dry, or vested interests persuade certain Congressional parties to defund it.
We had developed software under the DOE SciDAC program, which was distributed via an Outreach site. When funding got tight, that site was the first to go. Last I checked, the lone guy holding the line was keeping it up until the server died, at which point it was gone for good. If you want persistence, better make sure not to rely on a single point of failure. And these days, any .gov domain is prone to failure and/or corruption.
Well-written C can be very readable and maintainable. Machine-generated C rarely is.
Totally agree, based on the source I've seen dumped out by various tools. However, the question is whether it needs to be that way. I don't think it does, and that comes down to the research part. I know "Deep Neural Net" is treated like magic pixie dust to solve everything, but what if unreadable machine-generated code is a non-linear transformation away from readable, maintainable code? If it is, then would COBOL-to-maintainable-C be impossible? If we can't answer that with more than opinion, then it may be worth taking a look at seriously.
Yes, I realize the 95% correctness of a trained NN would result in awful correctness checking exercises, and there would undoubtedly be labor-intensive steps, but if it's that or "let the system collapse or rot", I'd choose moving ahead. Then again, maybe the problem's not that serious.
So true! Assemble multiple rockets in orbit, lash 'em together, and go. That 10-rockets-to-accelerate-no-deceleration plan might just get someone onto the surface of Mars in jellied form by his impeachment trial end date, if we start now.
Funding academic CS work on COBOL-to-C, or maybe even COBOL-to-C++, source transformations. They could start with any of the hits a "COBOL to C" Google search provides.
And just how the hell will I now forget to put a receptacle in place and dispense $5 of pre-made juice onto my kitchen counter by phone? BY PHONE, do you understand?? This is Internet-connected! IoT! AI! Deep Learning!
I'll bet it could order its own $1000 worth of juice refills from Amazon automatically, too, dammit.
The success marks a new era for the large-scale design of new magnetic materials
No, it doesn't. This is screening, and regardless of how much those in the field of drug "design" and materials "design" use those words, it's not Design, and it's not Engineering. It's Discovery, and it's great that physics and computation have gotten to the point where we can actually discover useful things in silico much faster than at the bench. But engineering requires an understanding of the underlying relationship between materials composition and desired quantitative property, and that is largely still lacking. If it weren't, you'd be screening 10s of compounds, not ~10^5.
Discovery and prediction are fine, no one will complain about the gold you get from panning for it. But don't dress it up in misleading language--true materials engineering and design are still a long way off.
But this won't happen -- or shouldn't happen -- unless we find ways of making techniques like deep learning more understandable to their creators and accountable to their users. Otherwise it will be hard to predict when failures might occur -- and it's inevitable they will.
Sounds a lot like humans. An observer has no hope of understanding why I make a decision, beyond shared social convention. And if they want to understand the process mechanistically, following impulses around the 1e14 estimated neural connections in a human brain? Forget it.
I agree that it's good to understand how a tool works, but we'll accept the deployment of these tools for the same reasons we accept our fellow beings hurtling around in 2+ ton wheeled projectiles--because most of the time, there isn't a problem, and more is gained from taking the risk than is lost from avoiding it. Legal responsibility needs to be made clear first, but as long as someone pays for fuck ups, probably OK
Probably does, from the alt-right perspective.
Maybe they should just introduce a troll icon, and a bot icon, along with a rating system a la Slashdot. Once a user crosses a threshold on "Troll" or "Bot" ratings, their icon automatically gets changed to one or the other.
In fact, just create icons to go along with every rating category, and change their icon based on how a user is rated.
Actually, come to think of it, Slashdot IS hardened Twitter, without the popularity (or icons).
Also, I see "below" as an adjective more and more. Irksome, even though it is more logical as a parallel construction with "above."
"Please look at the above paragraph" :) vs. "please look at the below paragraph" >:(
But details are hard and require knowing stuff, brah. Not. For. Meeeee...
Seriously though, agree totally, but are all those VCs and angels really so brain dead as to not know this? Or are they, like the finance folks capitalizing on herd movement into the stock markets at the moment, just cashing in on the phenomenon until it implodes? I suspect the only ones who aren't in on the joke are the dimwits who genuinely believe their brilliant calorie/counting-or-messaging/Youtube/mashup-or-whatever phone app is going to change the world, and they aren't going to pay much attention to cranky ol' Linus.
...the DVD I received rattling around in the refrigerator-sized shipping box. Insufficient training data on how to ship DVDs to Luddites mistrustful of streaming services.
AI will disrupt how developers build applications
Oh, good. As someone who has to build applications for users on HPC systems, first I had make. It was simple, things were either in Makefile or not. Drop the compiler options and paths right in. Then I had autotools, where I could pass paths and switches to possibly undocumented options. Still manageable, eventually. Thanks heavens that mechanism to manage an architecture and OS zoo came out as the world consolidated to x86-64 and Linux. Then I had cmake, which I can get to work sometimes. Not to mention all the one-off solutions from individual development groups that just had to improve on the state-of-the-art, because.
Now we'll get to troubleshoot "X not found" from a black-box build system generated by a neural net? Good luck. Nothing leads to "higher quality code" like unmanageable complexity. I kind of look forward to going back to the good old days of one codebase building on the development machine only. I can just about hear the world's productivity dropping from this trendy BS.
The root problem (I speculate) is the same one that afflicts companies looking ahead only to the next quarter, schools teaching only prescribed and minimally challenging material, the slow strangulation of endless safety regulation, etc. We're short-sighted, and can't fathom even a slight risk of negativity. As long as the next increment turns out OK, we figure we'll be just fine. That works, so long as the path you're taking doesn't lead you off a cliff or to stagnation, but if you have to get over some barrier (financial loss, global competition with rising powers, new discoveries) to get onto the optimal track, the purely local gradient-based search won't work. Even if eventual failure of a system is demonstrated, we'll keep doing the same thing because we're too fearful of the unknown to do anything else.
On the other hand, the real breakthroughs have never been supported by conventional thinkers or their backers. Kuhn, etc. Would taxpayers accept 99% of research funding to add up to nothing for the remaining 1% to pay off 1000-fold or more? I doubt it. The angels and VCs might risk those odds, but not the standard research funding apparatus.