(As for the BlueJ feature itself, I'm not exactly sure what's supposed to be new about it anyway. People have been doing that kind of testing since the days of Smalltalk.)
As TFA says:
We have never explicitly claimed invention of this technique, although it was new to us at the time. But I have never undertaken an exhaustive search for prior art, so it is entirely possible that there are earlier implementations. Key concepts are influenced by a variety of other systems anyway, such as Smalltalk and Self.
Please read more carefully. I didn't say the Linux binary would run on the Mac -- do you really think someone who had long experience with both would say such a thing? I said the app could be downloaded for the Mac. Between Fink, DarwinPorts, and the fact that most sites offering a particular downloadable Linux app also have a Mac version, I can't recall an occasion when an app that was available for Linux, and which I wanted to use on the Mac, could not be trivially obtained for the latter.
Plus, there are a lot of commercial packages that are available on Windows and the Mac, but not on Linux.
Unless a major goal of the OP is to learn about Linux specifically, I think it's a no-brainer.
As someone who has used both Linux and the Mac for years and years, I have to second this. Apple has put a massive amount of effort into making a Unix-like OS easier to use, and Linux is nowhere near catching up. Any app that runs on Linux can also be downloaded for the Mac. For the non-programmer I see no advantage to Linux whatsoever.
Ask anyone who has submitted a patent application whether they felt their patent was frivolous. I imagine you'd find the vast majority of them holding the belief that they did something novel.
Yes, but in the software field, that belief is likely to be mistaken. A colleague of mine and I once interviewed a senior IBM engineer for a position at the startup we work for. He described a patent he held for a memory allocation technique. Both my colleague and I, being familiar with the garbage collection literature, knew immediately of prior art dating back to the late 1970's. The technique we were thinking of is not in any sense obscure -- in fact it is now used in practically all production-quality collectors for languages other than C/C++ -- but this senior IBM engineer was nonetheless, evidently, entirely unaware of it. My colleague and I were unimpressed by his ignorance of such fundamental work, but our opinion of him aside, the fact is that software is a vast field and one can't begin to keep track of all of it.
I like what someone else here posted: a big part of the problem is that there is a burden of proof on the PTO to show that an invention is obvious, but there is no burden of proof on the applicant to show that it is nonobvious. In the case of the metal cylinder around the battery, the fact that no solution previously existed despite the economic importance of the problem counts as objective evidence of nonobviousness. Absent such evidence, I think the law should tell the PTO to take the safe course and reject the patent.
I can't comment on his overall intelligence, but I can see he doesn't understand algorithmic trading. You don't have to understand why your system works, and the market will tell you when it's broken (it will start losing money). You don't just build it and forget about it; you have to keep improving it.
Maybe, as someone else said, he didn't want to spend his time on that. But the fact that his potential employers don't understand neural nets should not have been the reason he didn't take the job.
If you're thinking in terms of "stock picking" you're not getting the picture; you're in a completely different time frame.
Algorithmic trading is mostly done in time frames of minutes; sometimes as short as a few seconds, or as long as a few days. Stock picking, on the other hand, as it is normally thought of, is concerned with time frames in the months-to-years range. In a nutshell, short-term trading works because those with longer time frames tend to occasionally act in a somewhat herd-like manner, and this behavior can be detected and exploited. It's not about stock picking; it's about watching for moments in which the stock-pickers are reacting predictably. So it has nothing to do with fundamentals, and everything to do with investor psychology.
This relates to a common misunderstanding about technical analysis. People unfamiliar with that field tend to think that TA-ers claim to be able to predict long-term chart movements. But TA is not about that; it's about noticing when the next move of the crowd is somewhat predictable -- not very predictable, necessarily, but enough so that playing the odds is still a winning proposition after commissions. So if you can identify a class of situations where the crowd's next move is predictable even just maybe 55% of the time, you can make money. 55% accuracy doesn't sound that great, but consider: if you have a profitable system, your profit is exponential in the number of trades you can do (since the larger your trading account, the more you can bet on each trade). If each trade takes only a few minutes, the potential profit is quite sizable, even if the average gain of each trade is small, percentage-wise.
This is how you have to think to understand algorithmic trading. It's a world very foreign to the typical long-term investor.
BTW, you're right, of course, that profitable systems don't necessarily stay profitable. It is, as TFA said, an ongoing arms race.
Yes, Linux is terrible about this. I have a machine that's still running SuSE 9.0 because I have a commercial app on it that doesn't link with recent glibc versions. Eventually I might have to upgrade the box anyway and run SuSE 9.0 under VMware for the benefit of this app.
Although the idea of setting up this Slashdot interview is clever, and my fellow Slashdotters are asking good questions, I have to say that this whole exercise is moot. Diebold wants us to trust them, but the system should be set up in such a way that we wouldn't have to trust them.
If we have a verifiable paper trail, nothing Diebold PR says matters. And if we don't, then nothing they can say should matter either, because even if we could trust them now (and I don't) there's no way to know that we can trust them into the indefinite future.
This is a long-standing complaint of mine, too. I would really like a 17-inch MBP at 1920x1200.
However, the current 17-inch resolution is, in fact, 1680x1050. So you can get the same resolution you have now -- you just have to spend a bit more money and carry a slightly larger machine (but I have a 17-inch PowerBook G4 and don't find the size to be a problem at all).
I agree with the GP -- 1 DIMM slot per core doesn't cut it.
Denser RAM is also more expensive per GB -- if you can even find it. Is anyone even selling 2GB DDR2-800 DIMMs yet?
I have used KDE for years, and have been pretty fond of it. (I have not liked Gnome when I tried it, but the last time was a while ago.) There's only one complaint I have. Every so often someone on the KDE team gets the bright idea that the way to implement some feature is to trap some key combination in the window manager or X server or some low-level place where the behavior is unconfigurable. In one release, a couple of years ago, the Windows key brought up the K menu, and there was no way to change this, not even using xmodmap. And now I find that in the new KDE used in SLED 10, Alt-Space brings up the Kerry Beagle Search dialog, and there's absolutely nothing I can do to change this -- it's not a configurable shortcut. In fact, it's clearly implemented at some low level, because it has a bug: not only does the dialog pop up, but somehow the Alt key's key-release event is lost, so Alt stays on unless I press and release it again.
Dear KDE team: please make it a part of your culture that you never ever ever do this. All special key combinations must be trapped in such a way that they are configurable, period. (In fairness, these things do seem to get fixed in subsequent releases... but I have to live with them in the meantime.)
Since I'm sure you're wondering why I care -- well, you must have no Emacs users on your team; Alt-Space (or Meta-Space in Emacs-speak) and Control-Meta-Space are both important Emacs commands which I use frequently. Interestingly, both of them are now bringing up the Kerry Beagle dialog, which is additional evidence that the key combination is being trapped incorrectly.
This whole glossy screen thing seems like a very unfortunate fad. I know, glossy is better for watching movies -- is that really the primary use that most laptops are being sold for???
I think we (those of us who don't always use our laptops in the dark) need to let the manufacturers know that there is still a significant fraction of the market for whom glossy screens are unacceptable, period. It's just nuts that you can't get a 17" 1920x1200 laptop from Dell with a matte-finish screen.
So, I learned that my machine was not built for constant processor work
Hogwash. That was a defective machine. I've run F@H on plenty of machines -- including Dells -- and have never seen such a failure. (And if one had failed, I would still consider it defective.)
You're right. I have one. The controls suck for fine frame selection. It's the only thing about the product I really hate. I have an old VCR that's much better.
Remember, all matter (at least, outside of a neutron star) is almost entirely empty space. Its density is controlled by how the electrons interact with one another and with the nuclei -- that's what keeps everything from collapsing into black holes. As the atomic number increases (the nuclei have more protons, and thus more positive charge) the electrostatic forces pull the electrons into tighter orbits, and, other things being equal, you get a denser material.
That's the very simple explanation -- there are other factors that affect density, like molecular conformation and crystal structure -- but it makes the point. It's all about the electrons.
Well, I was only half serious -- I'm running Solaris on my quad Opteron, but most people probably wouldn't want to. Indeed, I would really rather run Linux myself, except that the Linux kernel has a terrible performance problem on the particular workload I care most about. It's a data mining application written in Lisp; I routinely run process sizes between 5GB and 20GB. Even at the low end of that range, the Linux kernel, under certain circumstances, bogs down badly and the machine becomes unresponsive. (I can explain in more detail if you're curious.)
But Solaris runs just great on this and every other workload I've managed to throw at it. I haven't benchmarked it vs. Linux on more typical workloads, but subjectively it seems just as fast, or close.
How can a motherboard have all this stuff and leave out ECC? I would never buy a motherboard without ECC. Don't people want their machines to stay up more than a week at a time???
Slashdot Mirror
Yes, I am definitely going to have to remember this phrase! It's an instant classic!
"Is your code stable?" "Yes, it's a complete virtual impossibility that any bugs remain."
(As for the BlueJ feature itself, I'm not exactly sure what's supposed to be new about it anyway. People have been doing that kind of testing since the days of Smalltalk.)
As TFA says:
We have never explicitly claimed invention of this technique, although it was new to us at the time. But I have never undertaken an exhaustive search for prior art, so it is entirely possible that there are earlier implementations. Key concepts are influenced by a variety of other systems anyway, such as Smalltalk and Self.
This seems geeky even for Slashdot...
Please read more carefully. I didn't say the Linux binary would run on the Mac -- do you really think someone who had long experience with both would say such a thing? I said the app could be downloaded for the Mac. Between Fink, DarwinPorts, and the fact that most sites offering a particular downloadable Linux app also have a Mac version, I can't recall an occasion when an app that was available for Linux, and which I wanted to use on the Mac, could not be trivially obtained for the latter.
Plus, there are a lot of commercial packages that are available on Windows and the Mac, but not on Linux.
Unless a major goal of the OP is to learn about Linux specifically, I think it's a no-brainer.
As someone who has used both Linux and the Mac for years and years, I have to second this. Apple has put a massive amount of effort into making a Unix-like OS easier to use, and Linux is nowhere near catching up. Any app that runs on Linux can also be downloaded for the Mac. For the non-programmer I see no advantage to Linux whatsoever.
"Score: 0, Flamebait"??? Come on, mods! Surely "flamebait" doesn't just mean "an unpopular opinion"!
Ask anyone who has submitted a patent application whether they felt their patent was frivolous. I imagine you'd find the vast majority of them holding the belief that they did something novel.
Yes, but in the software field, that belief is likely to be mistaken. A colleague of mine and I once interviewed a senior IBM engineer for a position at the startup we work for. He described a patent he held for a memory allocation technique. Both my colleague and I, being familiar with the garbage collection literature, knew immediately of prior art dating back to the late 1970's. The technique we were thinking of is not in any sense obscure -- in fact it is now used in practically all production-quality collectors for languages other than C/C++ -- but this senior IBM engineer was nonetheless, evidently, entirely unaware of it. My colleague and I were unimpressed by his ignorance of such fundamental work, but our opinion of him aside, the fact is that software is a vast field and one can't begin to keep track of all of it.
I like what someone else here posted: a big part of the problem is that there is a burden of proof on the PTO to show that an invention is obvious, but there is no burden of proof on the applicant to show that it is nonobvious. In the case of the metal cylinder around the battery, the fact that no solution previously existed despite the economic importance of the problem counts as objective evidence of nonobviousness. Absent such evidence, I think the law should tell the PTO to take the safe course and reject the patent.
I can't comment on his overall intelligence, but I can see he doesn't understand algorithmic trading. You don't have to understand why your system works, and the market will tell you when it's broken (it will start losing money). You don't just build it and forget about it; you have to keep improving it.
Maybe, as someone else said, he didn't want to spend his time on that. But the fact that his potential employers don't understand neural nets should not have been the reason he didn't take the job.
If you're thinking in terms of "stock picking" you're not getting the picture; you're in a completely different time frame.
Algorithmic trading is mostly done in time frames of minutes; sometimes as short as a few seconds, or as long as a few days. Stock picking, on the other hand, as it is normally thought of, is concerned with time frames in the months-to-years range. In a nutshell, short-term trading works because those with longer time frames tend to occasionally act in a somewhat herd-like manner, and this behavior can be detected and exploited. It's not about stock picking; it's about watching for moments in which the stock-pickers are reacting predictably. So it has nothing to do with fundamentals, and everything to do with investor psychology.
This relates to a common misunderstanding about technical analysis. People unfamiliar with that field tend to think that TA-ers claim to be able to predict long-term chart movements. But TA is not about that; it's about noticing when the next move of the crowd is somewhat predictable -- not very predictable, necessarily, but enough so that playing the odds is still a winning proposition after commissions. So if you can identify a class of situations where the crowd's next move is predictable even just maybe 55% of the time, you can make money. 55% accuracy doesn't sound that great, but consider: if you have a profitable system, your profit is exponential in the number of trades you can do (since the larger your trading account, the more you can bet on each trade). If each trade takes only a few minutes, the potential profit is quite sizable, even if the average gain of each trade is small, percentage-wise.
This is how you have to think to understand algorithmic trading. It's a world very foreign to the typical long-term investor.
BTW, you're right, of course, that profitable systems don't necessarily stay profitable. It is, as TFA said, an ongoing arms race.
Yes, Linux is terrible about this. I have a machine that's still running SuSE 9.0 because I have a commercial app on it that doesn't link with recent glibc versions. Eventually I might have to upgrade the box anyway and run SuSE 9.0 under VMware for the benefit of this app.
Although the idea of setting up this Slashdot interview is clever, and my fellow Slashdotters are asking good questions, I have to say that this whole exercise is moot. Diebold wants us to trust them, but the system should be set up in such a way that we wouldn't have to trust them.
If we have a verifiable paper trail, nothing Diebold PR says matters. And if we don't, then nothing they can say should matter either, because even if we could trust them now (and I don't) there's no way to know that we can trust them into the indefinite future.
This is a long-standing complaint of mine, too. I would really like a 17-inch MBP at 1920x1200.
However, the current 17-inch resolution is, in fact, 1680x1050. So you can get the same resolution you have now -- you just have to spend a bit more money and carry a slightly larger machine (but I have a 17-inch PowerBook G4 and don't find the size to be a problem at all).
I agree with the GP -- 1 DIMM slot per core doesn't cut it. Denser RAM is also more expensive per GB -- if you can even find it. Is anyone even selling 2GB DDR2-800 DIMMs yet?
I have used KDE for years, and have been pretty fond of it. (I have not liked Gnome when I tried it, but the last time was a while ago.) There's only one complaint I have. Every so often someone on the KDE team gets the bright idea that the way to implement some feature is to trap some key combination in the window manager or X server or some low-level place where the behavior is unconfigurable. In one release, a couple of years ago, the Windows key brought up the K menu, and there was no way to change this, not even using xmodmap. And now I find that in the new KDE used in SLED 10, Alt-Space brings up the Kerry Beagle Search dialog, and there's absolutely nothing I can do to change this -- it's not a configurable shortcut. In fact, it's clearly implemented at some low level, because it has a bug: not only does the dialog pop up, but somehow the Alt key's key-release event is lost, so Alt stays on unless I press and release it again.
Dear KDE team: please make it a part of your culture that you never ever ever do this. All special key combinations must be trapped in such a way that they are configurable, period. (In fairness, these things do seem to get fixed in subsequent releases... but I have to live with them in the meantime.)
Since I'm sure you're wondering why I care -- well, you must have no Emacs users on your team; Alt-Space (or Meta-Space in Emacs-speak) and Control-Meta-Space are both important Emacs commands which I use frequently. Interestingly, both of them are now bringing up the Kerry Beagle dialog, which is additional evidence that the key combination is being trapped incorrectly.
Except for that -- keep up the good work!
... with GIB.
Well, I don't do much image work, so color accuracy means little to me -- and I just hate glare.
I suppose if the manufacturers would take the next logical step and add an anti-reflective coating, we could all be happy.
This whole glossy screen thing seems like a very unfortunate fad. I know, glossy is better for watching movies -- is that really the primary use that most laptops are being sold for???
I think we (those of us who don't always use our laptops in the dark) need to let the manufacturers know that there is still a significant fraction of the market for whom glossy screens are unacceptable, period. It's just nuts that you can't get a 17" 1920x1200 laptop from Dell with a matte-finish screen.
Elude: to avoid or escape; to escape the understanding of: The correct spelling eluded the editor.
Hogwash. That was a defective machine. I've run F@H on plenty of machines -- including Dells -- and have never seen such a failure. (And if one had failed, I would still consider it defective.)
You're right. I have one. The controls suck for fine frame selection. It's the only thing about the product I really hate. I have an old VCR that's much better.
Remember, all matter (at least, outside of a neutron star) is almost entirely empty space. Its density is controlled by how the electrons interact with one another and with the nuclei -- that's what keeps everything from collapsing into black holes. As the atomic number increases (the nuclei have more protons, and thus more positive charge) the electrostatic forces pull the electrons into tighter orbits, and, other things being equal, you get a denser material.
That's the very simple explanation -- there are other factors that affect density, like molecular conformation and crystal structure -- but it makes the point. It's all about the electrons.
(It's for a data mining project. No, I haven't filled all the swap yet, but I'm still coding...)
But Solaris runs just great on this and every other workload I've managed to throw at it. I haven't benchmarked it vs. Linux on more typical workloads, but subjectively it seems just as fast, or close.
That's right! Clearly, the only way to go on a box like that is Solaris!
How can a motherboard have all this stuff and leave out ECC? I would never buy a motherboard without ECC. Don't people want their machines to stay up more than a week at a time???