Definitely agree that power/performance are still important, and these factors were lacking in the analysis in TFA.
I used to think modern computers were all fast enough that learning C could be left to the minority, until I realized some of my complicated database queries that were taking hours could be done in seconds by adding C routines (thank god for open source).
C is fast because it's optimized for hardware, and hardware is optimized for C. There used to be groups optimizing hardware for Java and Lisp, but those never took off because other hardware became faster and made the efforts obsolete. But "other hardware" is still optimized for C, so it will be a long time until C goes into obsolescence.
I'm disappointed to see that the paper doesn't discuss evolution over time of the conversational roles. You can't expect a social science paper to do everything known to mankind, and this paper does a heck of a lot of things. In fact, it might take this paper to be cited a few times before some other social scientist picks up on the historical data.
I was just at the conference where Marc Smith (incidentally, an author of the paper and author of the/. story) presented this paper, and what I thought was most interesting was not just the computational tools they used to visualize thread data (apparently they existed for a long time but I wasn't aware of them), but the team's ability to use these tools to characterize users to a high degree of accuracy. What disappoints me about Smith's post is that he only emphasizes ``answer people'', while in their study they could accurately identify many other types of users including spammers and flame warriors, without looking at message content.
What makes this cool is that traditional social network analysis has not done very much in differentiating types of relations between users. They just draw lines between users, and the resulting network diagram is an incomprehensible mess. These people differentiate between incoming and outgoing messages, initiations and replies, first visits and returns. Maybe social scientists should have figured this out sooner, but better now than never.
it's late and i might be saying something really obvious... but i've convinced myself that slashdot is better because it has been around for so long. the user base has mostly been around for very long and is familiar with the system as well as what possibilities exist to exploit and troll it. ie, it is stable and i always know what i'm getting.
i don't think digg will forever be a forum for immature posts, but it is still young and what we see now may not be its equilibrium state. though, i sure wouldn't mind if its homepage were always as hilarious as it is right now.
Should there be a race distribution? Does it help us out to always take everything and split it up and see how the blacks are different, how the hispanics are different, how the whites are different, etc? Perhaps not always, but it is just as unhelpful to never take everything and just split it up by race. I see your point about imposing a priori assumptions on people, which is harmful when the assumptions are wrong. However, sometimes you really do want to know how things break down by race, especially if it's something like Internet usage. I have two major reasons to think this.
The better reason is that you're interested in the distributional impacts of something, for example Internet use. If you look at the data and see that, all else equal, half as many people in group X are using the Internet as group Y, you might be concerned that group X is not being given equal access for some reason. This is not stereotyping, but possibly to helps to motivate activities that actually mitigate stereotypes.
The more controversial reason is that certain types of stereotyping can be rational. Using your example of guessing what songs people like based purely on their looks with no further info, chances are if you guess consistent with stereotypes, you will get it right with higher probability than someone who just guessed at random. I hate to say it, but if this were the case in hiring, and a firm had no information whatsoever about any applicant other than race, it may actually be in their best interest to hire more of the group that statistically has done better in the firm's area of specialty.
In any case, I believe that statistically discriminating is acceptable to the point where it does not cause people more general harm than the alternative. The only catch is you have to be aware of when you are doing it. Note that this is different from taste-based discrimination, which is irrational as people just dislike group X for no justifiable reason, homophobia for example. But for something as innocuous as guessing music tastes, I don't see the harm in getting many of them wrong, and I can't think of a way that would generate more accurate guesses anyway.
Sorry, I can't help myself here...
gdp = C+I+G+NX = (income - savings)+I+G+NX
profits = costs - revenue = (wages + other costs) - (wages + other income such as capital gains)
when you kill wages/income, you kill your own profits as well as us gdp. Doesn't investment equal savings? Then your first equation says that
gdp = income + G + NX
which... I guess you could put it that way, but it doesn't really advance your point.
One of the three definitions of GDP is total income, so by reducing income you mechanically reduce GDP.
There's no need to show this with macro.
The second equation, even after you reverse it, you still get wages to cancel out with each other, leaving
profits = other income such as capital gains - other costs
So you've just made profit independent of wages by killing wages mathematically. At first I thought you meant to say
profit = revenue - (wages + other costs)
but this shows a reduction in wages actually increases profits, which seems to be the opposite of your argument. You might (or might not - I don't know) have the right conclusion, but you'll need to show it through some other means.
From the summary:
The article also quotes an economist who is skeptical that this report's outsized claims for productivity gains have been proven. I only glanced at the report, but there's good reason to be skeptical. It cites a lot of literature but does scant empirical work - if you look at the graphs they basically just take a bunch of time series datasets and line them up with each other. There's no good reason to believe that they show a causal relationship between the IT figures and the economic figures. Although they pose many plausible qualitative explanations, I don't really see any original research that tests any untested theories or devises any new ones.
I just don't know what to take away from the fact that this study has gained so much press coverage.
I've only used Google Docs (which I still prefer to call Writely because it has fewer syllables) though I'm sure the others are similar.
The one feature that I find novel is the way it does revisions, especially for multi-author documents. The Writely revisions interface is looking more and more like MS Word's "track changes" but with the additional benefits that (a) many people can edit at the same time, and (b) you can see all revisions ever made, whereas in Word, once an author overwrites his own revisions you can't see the old version anymore.
What's worked pretty well with a lot of my colleagues these days is to do everything on Writely until the text is complete, then one person downloads it and does the formatting in Word. (Then submit the paper and get rejected, but that happens with or without Writely.)
That said, I still think LaTeX looks the best but I don't know as many people who use it.
This isn't much of a response to the previous post, but I didn't want to start a thread with the same title...
The ITU has way too many things on its agenda to do a good job improving security in any area. Consider the ICANN, whose focus on Internet addresses is rather one-dimensional. IMO, it isn't even clear that ICANN does enough for security and efficiency. For example, every time the ICANN releases a new domain extension, the majority of memorable names get taken up almost immediately by domain squatters who proceed to make big bucks off of nothing. This is inefficient in the sense that monopoly is inefficient (not to mention annoying and responsible for the proliferation of firms with nonsense names that make you want to punch them). This is also a security issue because of the information people could inadvertently release by accidentally going to domain squatters' web pages.
I guess there are good things to be said about ICANN's being hands-off. But my point is if such gaping issues can exist within the single area of assigning addresses, imagine how many things the ITU would have to deal with if they were serious about security...
Actually, it isn't quite the poster's fault; this was exactly how some of the news sources wrote their headlines. The BBC version said "Asia communications hit by quake," and the CNN version (as of the time I post this) makes no mention of casualties.
Furthermore, severe earthquakes in Taiwan that damaged semiconductor manufacturing plants have costed billions (I guess in USD that would be tens of millions) in losses per day. It isn't that casualties are not significant losses as well, but in terms of impact, many more people - there are thousands whose daily living depends on these plants - are affected by damages to technological equipment.
Slashdot Mirror
Definitely agree that power/performance are still important, and these factors were lacking in the analysis in TFA. I used to think modern computers were all fast enough that learning C could be left to the minority, until I realized some of my complicated database queries that were taking hours could be done in seconds by adding C routines (thank god for open source). C is fast because it's optimized for hardware, and hardware is optimized for C. There used to be groups optimizing hardware for Java and Lisp, but those never took off because other hardware became faster and made the efforts obsolete. But "other hardware" is still optimized for C, so it will be a long time until C goes into obsolescence.
I was just at the conference where Marc Smith (incidentally, an author of the paper and author of the
What makes this cool is that traditional social network analysis has not done very much in differentiating types of relations between users. They just draw lines between users, and the resulting network diagram is an incomprehensible mess. These people differentiate between incoming and outgoing messages, initiations and replies, first visits and returns. Maybe social scientists should have figured this out sooner, but better now than never.
indeed. [...] When I get bored, I check my email. That's exactly what I do, and that's why I check email so often.
When I check and don't have new messages I visit slashdot.
it's late and i might be saying something really obvious... but i've convinced myself that slashdot is better because it has been around for so long. the user base has mostly been around for very long and is familiar with the system as well as what possibilities exist to exploit and troll it. ie, it is stable and i always know what i'm getting.
i don't think digg will forever be a forum for immature posts, but it is still young and what we see now may not be its equilibrium state. though, i sure wouldn't mind if its homepage were always as hilarious as it is right now.
The better reason is that you're interested in the distributional impacts of something, for example Internet use. If you look at the data and see that, all else equal, half as many people in group X are using the Internet as group Y, you might be concerned that group X is not being given equal access for some reason. This is not stereotyping, but possibly to helps to motivate activities that actually mitigate stereotypes.
The more controversial reason is that certain types of stereotyping can be rational. Using your example of guessing what songs people like based purely on their looks with no further info, chances are if you guess consistent with stereotypes, you will get it right with higher probability than someone who just guessed at random. I hate to say it, but if this were the case in hiring, and a firm had no information whatsoever about any applicant other than race, it may actually be in their best interest to hire more of the group that statistically has done better in the firm's area of specialty.
In any case, I believe that statistically discriminating is acceptable to the point where it does not cause people more general harm than the alternative. The only catch is you have to be aware of when you are doing it. Note that this is different from taste-based discrimination, which is irrational as people just dislike group X for no justifiable reason, homophobia for example. But for something as innocuous as guessing music tastes, I don't see the harm in getting many of them wrong, and I can't think of a way that would generate more accurate guesses anyway.
profits = costs - revenue = (wages + other costs) - (wages + other income such as capital gains)
when you kill wages/income, you kill your own profits as well as us gdp.
Doesn't investment equal savings? Then your first equation says that
gdp = income + G + NX
which... I guess you could put it that way, but it doesn't really advance your point. One of the three definitions of GDP is total income, so by reducing income you mechanically reduce GDP. There's no need to show this with macro.
The second equation, even after you reverse it, you still get wages to cancel out with each other, leaving
profits = other income such as capital gains - other costs
So you've just made profit independent of wages by killing wages mathematically. At first I thought you meant to say
profit = revenue - (wages + other costs)
but this shows a reduction in wages actually increases profits, which seems to be the opposite of your argument. You might (or might not - I don't know) have the right conclusion, but you'll need to show it through some other means.
From the summary: The article also quotes an economist who is skeptical that this report's outsized claims for productivity gains have been proven. I only glanced at the report, but there's good reason to be skeptical. It cites a lot of literature but does scant empirical work - if you look at the graphs they basically just take a bunch of time series datasets and line them up with each other. There's no good reason to believe that they show a causal relationship between the IT figures and the economic figures. Although they pose many plausible qualitative explanations, I don't really see any original research that tests any untested theories or devises any new ones.
I just don't know what to take away from the fact that this study has gained so much press coverage.
I've only used Google Docs (which I still prefer to call Writely because it has fewer syllables) though I'm sure the others are similar.
The one feature that I find novel is the way it does revisions, especially for multi-author documents. The Writely revisions interface is looking more and more like MS Word's "track changes" but with the additional benefits that (a) many people can edit at the same time, and (b) you can see all revisions ever made, whereas in Word, once an author overwrites his own revisions you can't see the old version anymore.
What's worked pretty well with a lot of my colleagues these days is to do everything on Writely until the text is complete, then one person downloads it and does the formatting in Word. (Then submit the paper and get rejected, but that happens with or without Writely.)
That said, I still think LaTeX looks the best but I don't know as many people who use it.
This isn't much of a response to the previous post, but I didn't want to start a thread with the same title...
The ITU has way too many things on its agenda to do a good job improving security in any area. Consider the ICANN, whose focus on Internet addresses is rather one-dimensional. IMO, it isn't even clear that ICANN does enough for security and efficiency. For example, every time the ICANN releases a new domain extension, the majority of memorable names get taken up almost immediately by domain squatters who proceed to make big bucks off of nothing. This is inefficient in the sense that monopoly is inefficient (not to mention annoying and responsible for the proliferation of firms with nonsense names that make you want to punch them). This is also a security issue because of the information people could inadvertently release by accidentally going to domain squatters' web pages.
I guess there are good things to be said about ICANN's being hands-off. But my point is if such gaping issues can exist within the single area of assigning addresses, imagine how many things the ITU would have to deal with if they were serious about security...
Actually, it isn't quite the poster's fault; this was exactly how some of the news sources wrote their headlines. The BBC version said "Asia communications hit by quake," and the CNN version (as of the time I post this) makes no mention of casualties.
Furthermore, severe earthquakes in Taiwan that damaged semiconductor manufacturing plants have costed billions (I guess in USD that would be tens of millions) in losses per day. It isn't that casualties are not significant losses as well, but in terms of impact, many more people - there are thousands whose daily living depends on these plants - are affected by damages to technological equipment.