Brian Hayes is a treasure, and so is American Scientist. American Scientist has taken over the role that Scientific American had in the 70s and 80s, when it was good, and Brian Hayes is almost the A. K. Dewdney (different, but better in some ways) of our current era.
Cultures of Code
21–26 of 26 posts
Re: Cultures of Code
#22There's at least a fourth even more removed culture, the plc programmer who is even more pragmatic than the software developer. He cares about the program about as much as the data scientist. It's just there to do the job.
Re: Cultures of Code
#23I'm not as worried about the author's conclusions as he seems to be. I think there will always be more engineers than physicists, more financial experts than economists, and more security professionals than cryptographers. There will always be more people who want to take advantage of the fruits of basic research than there are people who want to do that basic research. And moreover, the basic researchers are general…
"I think we could do a lot better publicizing our work to encourage more people to think about our problems" One step would be to tear down the paywall that keeps decades of super-interesting computer science research locked up. I'm looking at you, ACM. (Paywalls really limit access from people non-affiliated to any subscribing institutions, and even for affiliated people, it might be that slightly annoying barrier t…
Interestingly, some of the more theoretical conferences have started to break off from ACM/IEEE and become independent. See, for example, the computational complexity conference[1] which is now completely independent.
Re: Cultures of Code
#24Earlier quoted context omitted.
In a hypothetical situation I could be developing a piece of software and have an algorithm that is astronomically more efficient than its closest alternative, BUT if the efficient algorithm is good at doing exactly one thing and nothing else, while its alternative can be easily adopted to a variety of tasks and furthermore can be more easily understood by the guy who comes along after me, I am going to use the one t…
Another issue is that from a theoretical perspective, in a lot of cases, an algorithm that terminates , even if it takes longer than the heat-death of the universe, is an immense improvement over the current state. And the problem will be solved. And most of the field upheaved. But is the problem really solved solved? Nah.
What you mean is it's not solved from your perspective of the word solved. Again, that's the central point, that theoretical research is asking different questions (and answering them).
Re: Cultures of Code
#25I'm not as worried about the author's conclusions as he seems to be. I think there will always be more engineers than physicists, more financial experts than economists, and more security professionals than cryptographers. There will always be more people who want to take advantage of the fruits of basic research than there are people who want to do that basic research. And moreover, the basic researchers are general…
"I think we could do a lot better publicizing our work to encourage more people to think about our problems" One step would be to tear down the paywall that keeps decades of super-interesting computer science research locked up. I'm looking at you, ACM. (Paywalls really limit access from people non-affiliated to any subscribing institutions, and even for affiliated people, it might be that slightly annoying barrier t…
ACM is discussing removing the paywall, but they don't seem to have the final solution figured out yet.
http://cacm.acm.org/magazines/2014/11/179834-dealing-with-th...
Re: Cultures of Code
#26Earlier quoted context omitted.
In a hypothetical situation I could be developing a piece of software and have an algorithm that is astronomically more efficient than its closest alternative, BUT if the efficient algorithm is good at doing exactly one thing and nothing else, while its alternative can be easily adopted to a variety of tasks and furthermore can be more easily understood by the guy who comes along after me, I am going to use the one t…
Unless of course, you realize you need to optimize that particular use case... then you implement that ideality as a special case. ;)
Exactly, but it then the just-good-enough case is the one that is really really efficient.