You want to actually quote a statement I made about multi-threading that was incorrect?
Feel free to assert my lack of understanding about multi-threading. Feel free to explain how you are the expert on exotics. Guess it's good to keep your mind occupied between making the customers' lattes.
Hmm, citing Cato on derivatives is like citing my 4 year old on cats: yes, you'll get a strong opinion, but it won't be very useful to you.
Last week we added >1K cpus to our derivatives pricing compute farms. We'd love robust multi-threading. We'd love a ton of other things too. At the end of the day, the markup needed to sell exotic derivative X is strongly related to cost to simulate X's behavior in millions of possible worlds. I've had discussions were I tell the sales/trading team a single sale is going to cost them $100K/year in compute power just to track risk. If I could say $50K/year through multi-threading, they would get much happier.
I assumed that "banking app" was shorthand for "high-performance, complex app." Maybe the OP just meant an end-of-day retail banking job, and, if so, your comment about it being trivial to multi-thread applies.
I also assumed it referred to derivatives trading because that's where banking runs into the issues under discussion. Maybe I'm just naive and somehow missed a giant area of compute complexity in banking, but I'm pretty sure that is not the case.
My multi-threaded banking app blew up this week: Euros magically turned into an instance of some random user class.
A big banking app (I'm assuming derivatives trading rather than retail batch processing) probably has a 20 million LOC code base from 3 or 4 source languages and several hundred developers. Pure java, etc, is not an option.
There are probably only a few hundred programmers in the entire word that I would trust to multi-thread these types of apps, and even then, it's usually cheaper to just spin up another compute farm and have people refactor the apps to work nicely in a distributed fashion.
"Programmer" covers a wide range of skills, just like "lawyer." It is absurd to think the B-grade in-house counsel is suddenly going to morph into a expert free-speech supreme court advocate, and it's absurd to think a code-monkey struggling with C++ syntax really thinks about micro-kernels in her spare time.
Hey, if you have good math and logic skills, I don't much care what languages you actually know, and I trust you'll be up to speed on what I use within a few weeks.
If, on the other hand, you spent n years in school to learn a specific programming language, it doesn't really matter which one you learned, because I'm not going to hire you.
Actually it does. If he deserves it, then him not receiving it is wrong. The winner of a race deserves the gold medal - for him not to receive it would be wrong. An infinitely glorious being is infinitely worthy of worship - to not worship would be wrong, infinitely so.
That is an absurd position to take. A winner of a race deserves a medal because that was a question as to the outcome. An infinitely fast being can obviously win any race: so we yawn and move on.
What does "glorious" mean? Rational people accept what you did and say it was cool? "Infinitely glorious" is just silly: we accept what you did and now we have to say it was cool for ever and ever?
If God was real, I'd thank him for creating the universe, and even tell him that it was a neat idea. Then I'd go back to doing something useful.
"Never attribute to malice what can be accounted for by stupidity" goes the old phrase.
We think maintaining old hardware is easy, but it's actually pretty much impossible. Even if the doc had been kept up-to-date, it would be distributed in a thousand different places (I remember getting hundred page stacks to insert/replace into multiple three ring binders for small projects - you think I did it?) Add security/secrecy/compartmentalization/field workarounds, and good luck even figuring out what the spec is. A project is unrecreatable once the original designers/manufacturers/users have retired. Oh, and all that testing/debugging/validation machinery? You get to build that from scratch too:(
Heck, in the modern world, you just rewrite a six-month old software module if it's non-obvious and you can't find the intellectual owners.
Thanks for ruining my day. I enjoyed that bj from my wife, but now I'm thinking about those 100 million 50% contenders who got digested rather than getting converted into happy babies.
Yep - I spent a lot of time in the kitchen and knew how things were done:)
Had a similar experience with 5-8, I said -3, and the teacher declared it wrong because "we haven't learned negative numbers yet."
Luckily, I later moved to a different school district, and they just moved me from the 3rd grade to the 8th grade classroom when it was time for math. Didn't affect my rank in class, but at least I wasn't completely bored:)
Agreed. There is a talent modifier, possibly because certain skills are acquired and fixed early or are basically innate, e.g. proprioception, curiosity, observation, abstract thought, moral sense, sociability.
Heck, I remember a first grade math question from the teacher: if it takes 8 minutes to boil an egg, how long does it take to boil two eggs? I was stunned that all the other kids got the answer wrong. Then again, I sucked at a lot of other things they found easy.
Gladwell is way over-simplifying. 10,000 hours is not something magic. Many people have noted that practice must be directed (e.g. I know a number of people who spent 10K hours practicing martial arts: the good ones were the ones who thought about every kick, the mediocre just got fit.) It's like Tiger Woods describing how he thinks about his golf swing: deep, repetitive self analysis. Or George Burns at 80 discussing his comedy routine: he was still listening to the audience and working on getting the words and pauses exactly right.
As Paul Graham notes, it takes 10 years to become expert. 10K hours lets you get your foot in the door. Then spend another 20K hours to get to that expert level. And if you want to be a true master, expect another 90K hours working on your craft. As Chaucer said: the life so short, the craft so long to learn.
Reminds me of the last time I was interviewing candidates at MIT. One rattled off his achievements and skills, and asked when he could start fixing our business and how much he could expect to be paid for doing it. Trying not to laugh, I explained we considered him promising, and with another ten years of education in the business he might achieve the position to which he was aspiring.
Yeah, multi-core is clearly the future, but the future is too expensive for a single game-dev shop. You have to provide tools to abstract away the DMA transfers, the SIMD stuff, the chunk-size tuning, etc. At a minimum, you wind up with a functional language type framework, perhaps with classic imperative code at the bottom layer. Ideally, you also want a general parallel algorithm framework with good application tuning tools.
Even then, only 20% of current game developers are going to be able to use it effectively.
Total cost to create? Probably around $100M/year for 5 years, starting when the cell concept was first given the green light. Sadly, it's a bit late now.
A business bids to get exclusive access to a public good (a band) and pays us all in exchange. Then it provides a service to paying customers and recoups its costs.
The whole thing seems so sensible that we need some republicans to swoop in and explain why it's unfair to businesses.
Pretty much, but private with full audit trail, etc. As always, many in-house groups sought to kill it and replace it with something less anarchistic, but those efforts failed because the managers saw that a problem-queue approach would destroy the sharing of institutional knowledge.
We handle that traffic level with a few simple many-to-many chatrooms. All askers and answerers can see all messages, with highlighting of messages aimed at them. Bad answers are corrected quickly, and stupid questioners tend to get told to STFU: you quickly learn who is competent and who is not to be trusted. New users get up to speed quickly because they can watch the text stream and learn the expected style of communication.
Dear God, that site feels as if it were written by a tenth grader who plans on becoming a lawyer: unfocused drive-by posts spiced with pointless emphasis. What plays well as a soundbite on the network news just doesn't work on the internet: compared to, say, TPM or Firedoglake, it just lacks coherence. The republicans' best bet at this point is to just get the internet shut down: it's a medium that favors the intelligent and educated.
MS senior management probably wasn't even aware the firm had sent out these letters. Screwups like these happen when you start believing you can automate business to the point that no one is in control. Sure, it works for customer support for Verizon wireless, but is it a good idea for employee layoffs? These people are going to be getting drunk and telling anecdotes about the experience for the next 50 years: does your firm want the happy tale or the bitter tale?
Sheesh, last time I quit, the firm went out of its way to ask if there was anything they could have done better, and if there was anything they could do to encourage me to return in a few years. Yes, they were assholes, but they were very self-aware assholes and understood how important reputation is if you're around for the long term.
If it's less than $100K, the company will not take it to court. A good lawyer will merely hint that she plans to discover information on overtime, H1Bs, and discretionary bonuses. When the company realizes it has to explain to a jury that one senior developer earns $100K/yr, while another gets $400K, it STFU pretty fast, and eats the few thou it paid by mistake.
The whole reason managers exist is to maintain sanity: if a company reaches the point that it is terminating an employee without the employee's direct manager reviewing the documents involved, it deserves everything it gets.
You can chose to be either an ongoing, profitable business, or a bizarre digital art-form, but if you chose the later, do not be surprised when you lose money in the real world.
As I said, "absurd" numbers are easy to fix, but even $18K is tough to clawback from, for example, an employee with a five year work history. Heck, when I quit my last job, a nice lady from HR emailed me asking about unused vacation days: I picked a number, and it was close to $18K pre-tax.
Dollar numbers seem so important in entry level positions, but, at the higher levels, $BIGNUM amounts are used to smooth over the issues that arise in terminations.
Well, they're probably SOL regardless (unless the sum involved was absurd.) They chose to terminate the employee, they chose the terms, they asked for a promise to not sue (for overtime, unused vacation, discrimination, etc) in exchange for some cash. They paid, and now think they paid the wrong amount: good luck winning this one in court.
No attempt at humor, no attempt to do anything other than lash out at the administration because she was angry. And she was angry b/c she wasn't getting her way.
So she's stuck in a situation she can't walk away from. I'd be inclined to give her infinite free-speech points, and let the administration sever the relationship if they can't deal.
I'm glad that her behavior was ultimately punished (although I do agree that prison time is overkill) and I'm also glad that corrupt officials are also being punished. One has nothing to do with the other.
Yeah, expel her if you can't deal. For the last two thousand years, this has been the traditional punishment for students who wage non-violent campaigns against their tutors (beatings also factor in, but they are rarer nowadays.)
That said, I hope the judges involved enjoy their time in federal rape-me-in-the-ass prison because understanding what it feels like to be powerless in the face of arbitrary authority is something a judge would be better off for understanding.
Slashdot Mirror
Hmm, you seem to have problems understanding "if" statements. I guess UNC only teaches those in post-grad.
You want to actually quote a statement I made about multi-threading that was incorrect?
Feel free to assert my lack of understanding about multi-threading. Feel free to explain how you are the expert on exotics. Guess it's good to keep your mind occupied between making the customers' lattes.
You sound like an intelligent person opining on a topic you know nothing about. I'm sure you wow the crowd at the local Olive Garden.
Hmm, citing Cato on derivatives is like citing my 4 year old on cats: yes, you'll get a strong opinion, but it won't be very useful to you.
Last week we added >1K cpus to our derivatives pricing compute farms. We'd love robust multi-threading. We'd love a ton of other things too. At the end of the day, the markup needed to sell exotic derivative X is strongly related to cost to simulate X's behavior in millions of possible worlds. I've had discussions were I tell the sales/trading team a single sale is going to cost them $100K/year in compute power just to track risk. If I could say $50K/year through multi-threading, they would get much happier.
I assumed that "banking app" was shorthand for "high-performance, complex app." Maybe the OP just meant an end-of-day retail banking job, and, if so, your comment about it being trivial to multi-thread applies.
I also assumed it referred to derivatives trading because that's where banking runs into the issues under discussion. Maybe I'm just naive and somehow missed a giant area of compute complexity in banking, but I'm pretty sure that is not the case.
My multi-threaded banking app blew up this week: Euros magically turned into an instance of some random user class.
A big banking app (I'm assuming derivatives trading rather than retail batch processing) probably has a 20 million LOC code base from 3 or 4 source languages and several hundred developers. Pure java, etc, is not an option.
There are probably only a few hundred programmers in the entire word that I would trust to multi-thread these types of apps, and even then, it's usually cheaper to just spin up another compute farm and have people refactor the apps to work nicely in a distributed fashion.
It is depressing, and also very true.
"Programmer" covers a wide range of skills, just like "lawyer." It is absurd to think the B-grade in-house counsel is suddenly going to morph into a expert free-speech supreme court advocate, and it's absurd to think a code-monkey struggling with C++ syntax really thinks about micro-kernels in her spare time.
Hey, if you have good math and logic skills, I don't much care what languages you actually know, and I trust you'll be up to speed on what I use within a few weeks.
If, on the other hand, you spent n years in school to learn a specific programming language, it doesn't really matter which one you learned, because I'm not going to hire you.
Actually it does. If he deserves it, then him not receiving it is wrong. The winner of a race deserves the gold medal - for him not to receive it would be wrong. An infinitely glorious being is infinitely worthy of worship - to not worship would be wrong, infinitely so.
That is an absurd position to take. A winner of a race deserves a medal because that was a question as to the outcome. An infinitely fast being can obviously win any race: so we yawn and move on.
What does "glorious" mean? Rational people accept what you did and say it was cool? "Infinitely glorious" is just silly: we accept what you did and now we have to say it was cool for ever and ever?
If God was real, I'd thank him for creating the universe, and even tell him that it was a neat idea. Then I'd go back to doing something useful.
"Never attribute to malice what can be accounted for by stupidity" goes the old phrase.
We think maintaining old hardware is easy, but it's actually pretty much impossible. Even if the doc had been kept up-to-date, it would be distributed in a thousand different places (I remember getting hundred page stacks to insert/replace into multiple three ring binders for small projects - you think I did it?) Add security/secrecy/compartmentalization/field workarounds, and good luck even figuring out what the spec is. A project is unrecreatable once the original designers/manufacturers/users have retired. Oh, and all that testing/debugging/validation machinery? You get to build that from scratch too :(
Heck, in the modern world, you just rewrite a six-month old software module if it's non-obvious and you can't find the intellectual owners.
Thanks for ruining my day. I enjoyed that bj from my wife, but now I'm thinking about those 100 million 50% contenders who got digested rather than getting converted into happy babies.
Yep - I spent a lot of time in the kitchen and knew how things were done :)
Had a similar experience with 5-8, I said -3, and the teacher declared it wrong because "we haven't learned negative numbers yet."
Luckily, I later moved to a different school district, and they just moved me from the 3rd grade to the 8th grade classroom when it was time for math. Didn't affect my rank in class, but at least I wasn't completely bored :)
Agreed. There is a talent modifier, possibly because certain skills are acquired and fixed early or are basically innate, e.g. proprioception, curiosity, observation, abstract thought, moral sense, sociability.
Heck, I remember a first grade math question from the teacher: if it takes 8 minutes to boil an egg, how long does it take to boil two eggs? I was stunned that all the other kids got the answer wrong. Then again, I sucked at a lot of other things they found easy.
Good point.
Gladwell is way over-simplifying. 10,000 hours is not something magic. Many people have noted that practice must be directed (e.g. I know a number of people who spent 10K hours practicing martial arts: the good ones were the ones who thought about every kick, the mediocre just got fit.) It's like Tiger Woods describing how he thinks about his golf swing: deep, repetitive self analysis. Or George Burns at 80 discussing his comedy routine: he was still listening to the audience and working on getting the words and pauses exactly right.
As Paul Graham notes, it takes 10 years to become expert. 10K hours lets you get your foot in the door. Then spend another 20K hours to get to that expert level. And if you want to be a true master, expect another 90K hours working on your craft. As Chaucer said: the life so short, the craft so long to learn.
Reminds me of the last time I was interviewing candidates at MIT. One rattled off his achievements and skills, and asked when he could start fixing our business and how much he could expect to be paid for doing it. Trying not to laugh, I explained we considered him promising, and with another ten years of education in the business he might achieve the position to which he was aspiring.
Yeah, multi-core is clearly the future, but the future is too expensive for a single game-dev shop. You have to provide tools to abstract away the DMA transfers, the SIMD stuff, the chunk-size tuning, etc. At a minimum, you wind up with a functional language type framework, perhaps with classic imperative code at the bottom layer. Ideally, you also want a general parallel algorithm framework with good application tuning tools.
Even then, only 20% of current game developers are going to be able to use it effectively.
Total cost to create? Probably around $100M/year for 5 years, starting when the cell concept was first given the green light. Sadly, it's a bit late now.
Exactly, good capitalism in action!
A business bids to get exclusive access to a public good (a band) and pays us all in exchange. Then it provides a service to paying customers and recoups its costs.
The whole thing seems so sensible that we need some republicans to swoop in and explain why it's unfair to businesses.
Pretty much, but private with full audit trail, etc. As always, many in-house groups sought to kill it and replace it with something less anarchistic, but those efforts failed because the managers saw that a problem-queue approach would destroy the sharing of institutional knowledge.
We handle that traffic level with a few simple many-to-many chatrooms. All askers and answerers can see all messages, with highlighting of messages aimed at them. Bad answers are corrected quickly, and stupid questioners tend to get told to STFU: you quickly learn who is competent and who is not to be trusted. New users get up to speed quickly because they can watch the text stream and learn the expected style of communication.
Dear God, that site feels as if it were written by a tenth grader who plans on becoming a lawyer: unfocused drive-by posts spiced with pointless emphasis. What plays well as a soundbite on the network news just doesn't work on the internet: compared to, say, TPM or Firedoglake, it just lacks coherence. The republicans' best bet at this point is to just get the internet shut down: it's a medium that favors the intelligent and educated.
MS senior management probably wasn't even aware the firm had sent out these letters. Screwups like these happen when you start believing you can automate business to the point that no one is in control. Sure, it works for customer support for Verizon wireless, but is it a good idea for employee layoffs? These people are going to be getting drunk and telling anecdotes about the experience for the next 50 years: does your firm want the happy tale or the bitter tale?
Sheesh, last time I quit, the firm went out of its way to ask if there was anything they could have done better, and if there was anything they could do to encourage me to return in a few years. Yes, they were assholes, but they were very self-aware assholes and understood how important reputation is if you're around for the long term.
If it's less than $100K, the company will not take it to court. A good lawyer will merely hint that she plans to discover information on overtime, H1Bs, and discretionary bonuses. When the company realizes it has to explain to a jury that one senior developer earns $100K/yr, while another gets $400K, it STFU pretty fast, and eats the few thou it paid by mistake.
The whole reason managers exist is to maintain sanity: if a company reaches the point that it is terminating an employee without the employee's direct manager reviewing the documents involved, it deserves everything it gets.
You can chose to be either an ongoing, profitable business, or a bizarre digital art-form, but if you chose the later, do not be surprised when you lose money in the real world.
As I said, "absurd" numbers are easy to fix, but even $18K is tough to clawback from, for example, an employee with a five year work history. Heck, when I quit my last job, a nice lady from HR emailed me asking about unused vacation days: I picked a number, and it was close to $18K pre-tax.
Dollar numbers seem so important in entry level positions, but, at the higher levels, $BIGNUM amounts are used to smooth over the issues that arise in terminations.
Well, they're probably SOL regardless (unless the sum involved was absurd.) They chose to terminate the employee, they chose the terms, they asked for a promise to not sue (for overtime, unused vacation, discrimination, etc) in exchange for some cash. They paid, and now think they paid the wrong amount: good luck winning this one in court.
No attempt at humor, no attempt to do anything other than lash out at the administration because she was angry. And she was angry b/c she wasn't getting her way.
So she's stuck in a situation she can't walk away from. I'd be inclined to give her infinite free-speech points, and let the administration sever the relationship if they can't deal.
I'm glad that her behavior was ultimately punished (although I do agree that prison time is overkill) and I'm also glad that corrupt officials are also being punished. One has nothing to do with the other.
Yeah, expel her if you can't deal. For the last two thousand years, this has been the traditional punishment for students who wage non-violent campaigns against their tutors (beatings also factor in, but they are rarer nowadays.)
That said, I hope the judges involved enjoy their time in federal rape-me-in-the-ass prison because understanding what it feels like to be powerless in the face of arbitrary authority is something a judge would be better off for understanding.