You know, I've wondered about this. As I've understood it -- just listened to Jack Valenti talking about it on the news today -- they claim that there's been a drop of something like 30 percent in sales, which they ascribe to file sharing.
But, I've also seen it said that book sales were down about 30 percent.
It seems as if the least hypothesis is that file sharing, rather than costs the RIAA members zillions, is actually costing them statistically nothing.
Sadly, if there's anything like a PLA or EEPROM inside, you're hosed. I went through this for just *months* some time ago, when trying to figure out why a simple little box would work in the lab at my company, but not in the field. The answer was that the box was getting x-rayed in transit, causing the EEPROM to get random extra bits....
Well, yes I am. Since the commodity is one that's delivered by wire over a complicated infrastructure, it seems like a fair comparison.
But go ahead, do the figures: adjusted for inflation, gas prices are about the same as 1970; a $1.50 Coke adjusted for inflation is comparable to the dime coke of my extreme youth -- and you get to keep the bottle; cable TV is more expensive but hell, 20 years ago you got 10 channels. Thinking about it, TV per channel is probably two orders of magnitude cheaper than it once was.
Economics strikes again: if it really would cost trillions and trillions, you've got the real problem that we don't have trillions and trillions to spend. The whole GDP of the US is in the neighborhood of $6 trillion, and the world overall certainly doesn't exceed $20 trillion. So what you're proposing is tantamount to suggesting we put everyone on Earth into working on power sats, dropping everything else from food production to programming video games.
I doan' theeeeenk so, Cisco.
Solar power is certainly a fine idea, but out here at the Earth's orbit, it's still only 1 kW/m^2, and the best conversion efficiencies are still less that 20 percent -- in other words, we'd really get about 5 m^2 per kilowatt. It'd take a big satellite to replace one of those 10 GW reactors.
"Essentially for free" is flawed too, but I'll leave the reason for that as an exercise.
crmartin 98227 points out below that if you can't be bothered to learn something about economics, you could at least develop a bit of historical perspective. Given that we have seen multiple order of magnitude price drops in other similar commodities in the past, your notion that prices for energy wouldn't drop seems empirically unsound.
Christ, if you can't learn some economics, at least get a historical perspective of more than the time between the morning and evening news.
Consider, eg, the cost of long distance telephone calls: when I was a kid, back at the Dawn of Time (electromechanical trunk switching, that kind of thing) a three minute long-distance call from Colorado to Oklahoma cost (inflation adjusted) about $20. Five years ago, the cost was around 20 cents a minute. And now I get my phone service -- including US long distance -- for $40 a month flat, or I could buy long distance for 3-5 cents a minute a la carte. In other words, the price has dropped by a couple of orders of magnitude.
If your idiotic model were correct, then it'd still cost $7 a minute and AT&T would be making immense profits, instead of being on the ragged edge of making no profit whatsoever.
You can get really close to breakeven on food using hydroponics and a greenhouse even now, so cheap power would make it even better. It'd help a lot with the Moon, too -- it takes about a kilowatt per square meter to grow food, (modulo lighting efficiency etc) so underground growing would be driven largely by the cost of power.
The thing I find interesting is that -- at least on the Moon -- this might make fission more viable. It's easier to build "small" -- say 10 to 100 megawatts -- and we know there are decent amounts of thorium available on the Moon. The thorium would be a byproduct of 3He production.
Long term, you're probably right -- the cost of steel, copper, aluminum for the transmission lines goes down. That'd be competing with the obvious pressure to move power generation closer to the users, which would be balanced out by capital cost and capacity limits -- you tell me what they are and we can make a guess where the breakeven would be. When I was thinking about it, it was with Bussard's notion of a "Farnsworth fusor" (see, eg, here, here, or here, or the Google search here.)
This leads to a notional reactor that's 5 meters across, and yields 10 gigaWatts (6600 Amps at about 1.5 megavolts DC, and be damned to Tesla!) using proton-boron fusion.
(Note: I'm not a physicist, and I'm not a power engineer, so don't come after me if you don't like these ideas.)
The whole thing is basically a big empty conductive sphere with some accessories, so it shouldn't cost more than about $1 million, so we're definitely in the neighborhood of pennies to mils per megawatt-hour. But it's almost an embarrassment of riches: how to you deal with a city of, say, 5000? A million bucks is a feasible investment for a city that size, but what do you do with the 9.75 gigawatts left over?
I've done a lot of thinking about this myself, and it turns out to have some interesting implications.
First, it turns out that the cost of electric at the wall-socket is not dominated by the cost of production, but by the cost of the power grid. If the power were completely free, cost/kW-h at the home would only go down by about 50 percent.
On the other hand, cost of electricity does dominate the cost to make aluminum, steel, and many chemicals: profits would immediately go up, and costs would quickly drop precipitously for everything from cars to Tylenol.
Free electric power wouldn't in itself make space travel cheaper, but if you have cheap fusion you can either make fusion rockets, or extend VASIMR. If you can get thrust high and exhaust velocity very high -- say tens to hundreds of km/sec -- then you can quickly start doing things like going to the Moon with constant acceleration. In other words, a trip from Earth to Moon could be quicker than a trip from New York to Boston today.
Waste disposal would change radically -- give me enough power and I'll just do mass spectroscopy on a plasma made from the wastes. Call it 'mass mass spectroscopy' -- out the end comes pure (isotopically pure, if you care to do it) oxygen, hydrogen, carbon, and so on. This will be very handy for Lunar exploration, as it makes possible the easy separation of 3He from 4He; 3He makes for good fusion fuel, and 4He ("depleted helium"?) makes for cheap reaction mass or lots of other things. On the other hand, it makes uranium enrichment much easier as well -- throw in yellowcake, and out the other end comes O2 235U and 238U.
If lunar 3He production is economic, so is production of hydrogen (either from fossil water or as a byproduct of 3He production) as well as oxygen, nitrogen, argon, potassium, thorium, and so on. (See KREEP.) Add O2, N2, and lights to a lunar lava tube, and you've got living space and farms -- with cheap power.
I'm not quite sure why you read like we're arguing, since I don't think we are -- except to the extent that you don't thing the lengthy period of Markovian behavior is interesting, while I think that's the most interesting part.
The point is this: the MTBF is computed for the Markovian part of the total life. The value for MTBF is computed as 1 / failure-rate IF AND ONLY IF the time distribution of failures is Markovian -- otherwise it's a more complicated function. The useful life is the length of time over which the failure rate is exponential: when you get to the upward inflection, that's the end of the useful life. Thus if a drive has a useful life of 5 years, and a failure rate of 1/50 years (that is, a MTBF of 50 years) then the chances are one in ten that the drive will fail during that five years.
The warranty issue is something else: a warranty is exactly like insurance -- it's a bet that you won't have to pay off on the warranty too often. The amount of money you have to charge for the warranty is the amount at risk times the chances of a failure over the lifetime of the warranty. Thus if you by a $100 CD player, and they offer you a $20 extended warranty for 3 years, they're saying they think the odds are less than on in five that they'll have to pay off during the year. Given the number of people who promptly lose the paperwork, that's usually an excellent bet; as you say, that makes the extended warranty a profit center all in itself.
Anthony, that's not really true. First of all, for a long part of the lifetime, exponential is a very good approximation. I can't cite it offhand, but I read a paper that showed it was better than 95% accurate. Secondly, the Bayesian method will give measures that you can advertise as correct (and people like Telcordia are very stubborn about it) with a relatively short and inexpensive experiment.
That's the difference between "statistical inference" and what people who have had a probability class tend to do. If you've got 500,000 drives, test for an hour, and get one failure, you've got some evidence that the failure rate is 1 per 500,000 hours (== MTBF of 500,000 hours in the exponential case.) But the confidence in that estimate is very low. If you then replace the failed drive and do another 1 hour test, and get another single failure, then you still can estimate the MTBF as 500,000 hours -- but your confidence level is higher.
This trick is the key to the Bayesian estimation methods I mentioned above. Check out the CiteSeer and Google links for more papers than you ever wanted to read.
Now, consider if you get >1 failure in that second hour: there are several possible explanations:
the failure rate is increasing over time
the failure rate is 1/500,000 but it just happened that you had two random failures in that one hour
The failure rate is really more than 1/500,000 and it just happened randomly that the first hour had only one failure.
In general, you've got to test a lot of items for a fair chunk of the expected lifetime to get the confidence level up.
Work it out; IF it's an exponential, THEN for failure rate parameter lambda, mean time between failures is 1/lambda. It's too painful to try to show this in HTML, but Trivedi's book cited above ("favorite text") shows the derivation.
Yes. Strictly -- as someone else pointed out -- the failures tend to have a so-called "bathtub" distribution. That is, there's a high failur rate at first ("infant mortality") followed by a long Poisson/Markovian/exponential (you pick your term) stretch, followed by a higher failure rate as it get old. In general, the "lifetime" of a component is the time to the inflection at the end of the exponential portion.
"There are lots of planes and lots of years" is the right answer: I don't recall the exact figures right now, but at the time of the TWA 800 crash I predicted that it would turn out to be an airframe failure (on the heuristic of preferring failure to malice) because when I worked the numbers it turned out that MTBF of 747s was about that same 1e10.
By the way, your supposition about answer (1) is correct. It's just a definitional thing: we're really talking about proximate and root causes. You don't count it as an airframe failure if someone throws a cigarette butt into the trash and smoke and fire causes the crash, and while the airframe certainly fails if someone flies the plane into a mountain, that's still not the immediate cause.
Well, I write books and screenplays, but I do that on a computer so I don't think that counts. I'm a docent at the new Space Odyssey exhibit, but that's all computer-controlled....
I work out... no, the treadmill is computer controlled.
Slashdot Mirror
You know, I've wondered about this. As I've understood it -- just listened to Jack Valenti talking about it on the news today -- they claim that there's been a drop of something like 30 percent in sales, which they ascribe to file sharing.
But, I've also seen it said that book sales were down about 30 percent.
It seems as if the least hypothesis is that file sharing, rather than costs the RIAA members zillions, is actually costing them statistically nothing.
Sadly, if there's anything like a PLA or EEPROM inside, you're hosed. I went through this for just *months* some time ago, when trying to figure out why a simple little box would work in the lab at my company, but not in the field. The answer was that the box was getting x-rayed in transit, causing the EEPROM to get random extra bits....
Well, yes I am. Since the commodity is one that's delivered by wire over a complicated infrastructure, it seems like a fair comparison.
But go ahead, do the figures: adjusted for inflation, gas prices are about the same as 1970; a $1.50 Coke adjusted for inflation is comparable to the dime coke of my extreme youth -- and you get to keep the bottle; cable TV is more expensive but hell, 20 years ago you got 10 channels. Thinking about it, TV per channel is probably two orders of magnitude cheaper than it once was.
Economics strikes again: if it really would cost trillions and trillions, you've got the real problem that we don't have trillions and trillions to spend. The whole GDP of the US is in the neighborhood of $6 trillion, and the world overall certainly doesn't exceed $20 trillion. So what you're proposing is tantamount to suggesting we put everyone on Earth into working on power sats, dropping everything else from food production to programming video games.
I doan' theeeeenk so, Cisco.
Solar power is certainly a fine idea, but out here at the Earth's orbit, it's still only 1 kW/m^2, and the best conversion efficiencies are still less that 20 percent -- in other words, we'd really get about 5 m^2 per kilowatt. It'd take a big satellite to replace one of those 10 GW reactors.
"Essentially for free" is flawed too, but I'll leave the reason for that as an exercise.
It'd take a helluva extension cord.
crmartin 98227 points out below that if you can't be bothered to learn something about economics, you could at least develop a bit of historical perspective. Given that we have seen multiple order of magnitude price drops in other similar commodities in the past, your notion that prices for energy wouldn't drop seems empirically unsound.
Christ, if you can't learn some economics, at least get a historical perspective of more than the time between the morning and evening news.
Consider, eg, the cost of long distance telephone calls: when I was a kid, back at the Dawn of Time (electromechanical trunk switching, that kind of thing) a three minute long-distance call from Colorado to Oklahoma cost (inflation adjusted) about $20. Five years ago, the cost was around 20 cents a minute. And now I get my phone service -- including US long distance -- for $40 a month flat, or I could buy long distance for 3-5 cents a minute a la carte. In other words, the price has dropped by a couple of orders of magnitude.
If your idiotic model were correct, then it'd still cost $7 a minute and AT&T would be making immense profits, instead of being on the ragged edge of making no profit whatsoever.
Your point being that if the extremely cheap energy were really expensive, it wouldn't be cheap after all?
You can get really close to breakeven on food using hydroponics and a greenhouse even now, so cheap power would make it even better. It'd help a lot with the Moon, too -- it takes about a kilowatt per square meter to grow food, (modulo lighting efficiency etc) so underground growing would be driven largely by the cost of power.
The thing I find interesting is that -- at least on the Moon -- this might make fission more viable. It's easier to build "small" -- say 10 to 100 megawatts -- and we know there are decent amounts of thorium available on the Moon. The thorium would be a byproduct of 3He production.
Long term, you're probably right -- the cost of steel, copper, aluminum for the transmission lines goes down. That'd be competing with the obvious pressure to move power generation closer to the users, which would be balanced out by capital cost and capacity limits -- you tell me what they are and we can make a guess where the breakeven would be. When I was thinking about it, it was with Bussard's notion of a "Farnsworth fusor" (see, eg, here, here, or here, or the Google search here.)
This leads to a notional reactor that's 5 meters across, and yields 10 gigaWatts (6600 Amps at about 1.5 megavolts DC, and be damned to Tesla!) using proton-boron fusion.
(Note: I'm not a physicist, and I'm not a power engineer, so don't come after me if you don't like these ideas.)
The whole thing is basically a big empty conductive sphere with some accessories, so it shouldn't cost more than about $1 million, so we're definitely in the neighborhood of pennies to mils per megawatt-hour. But it's almost an embarrassment of riches: how to you deal with a city of, say, 5000? A million bucks is a feasible investment for a city that size, but what do you do with the 9.75 gigawatts left over?
I've done a lot of thinking about this myself, and it turns out to have some interesting implications.
First, it turns out that the cost of electric at the wall-socket is not dominated by the cost of production, but by the cost of the power grid. If the power were completely free, cost/kW-h at the home would only go down by about 50 percent.
On the other hand, cost of electricity does dominate the cost to make aluminum, steel, and many chemicals: profits would immediately go up, and costs would quickly drop precipitously for everything from cars to Tylenol.
Free electric power wouldn't in itself make space travel cheaper, but if you have cheap fusion you can either make fusion rockets, or extend VASIMR. If you can get thrust high and exhaust velocity very high -- say tens to hundreds of km/sec -- then you can quickly start doing things like going to the Moon with constant acceleration. In other words, a trip from Earth to Moon could be quicker than a trip from New York to Boston today.
Waste disposal would change radically -- give me enough power and I'll just do mass spectroscopy on a plasma made from the wastes. Call it 'mass mass spectroscopy' -- out the end comes pure (isotopically pure, if you care to do it) oxygen, hydrogen, carbon, and so on. This will be very handy for Lunar exploration, as it makes possible the easy separation of 3He from 4He; 3He makes for good fusion fuel, and 4He ("depleted helium"?) makes for cheap reaction mass or lots of other things. On the other hand, it makes uranium enrichment much easier as well -- throw in yellowcake, and out the other end comes O2 235U and 238U.
If lunar 3He production is economic, so is production of hydrogen (either from fossil water or as a byproduct of 3He production) as well as oxygen, nitrogen, argon, potassium, thorium, and so on. (See KREEP.) Add O2, N2, and lights to a lunar lava tube, and you've got living space and farms -- with cheap power.
"am" does not agree with "you"....
Did it hurt when they removed your sense of humor?
No, you am just American. The Brits consider a collective noun (like the name of a company) to be plural, so "Activision are" would be correct.
Okay, guys, how the hell do you get them naked?
...it would be Rob Schneider. Schneider has made a career of pretending to be things he's not.
....
Yeah. Pretending to be an actor, pretending to be funny
I'm not quite sure why you read like we're arguing, since I don't think we are -- except to the extent that you don't thing the lengthy period of Markovian behavior is interesting, while I think that's the most interesting part.
The point is this: the MTBF is computed for the Markovian part of the total life. The value for MTBF is computed as 1 / failure-rate IF AND ONLY IF the time distribution of failures is Markovian -- otherwise it's a more complicated function. The useful life is the length of time over which the failure rate is exponential: when you get to the upward inflection, that's the end of the useful life. Thus if a drive has a useful life of 5 years, and a failure rate of 1/50 years (that is, a MTBF of 50 years) then the chances are one in ten that the drive will fail during that five years.
The warranty issue is something else: a warranty is exactly like insurance -- it's a bet that you won't have to pay off on the warranty too often. The amount of money you have to charge for the warranty is the amount at risk times the chances of a failure over the lifetime of the warranty. Thus if you by a $100 CD player, and they offer you a $20 extended warranty for 3 years, they're saying they think the odds are less than on in five that they'll have to pay off during the year. Given the number of people who promptly lose the paperwork, that's usually an excellent bet; as you say, that makes the extended warranty a profit center all in itself.
Anthony, that's not really true. First of all, for a long part of the lifetime, exponential is a very good approximation. I can't cite it offhand, but I read a paper that showed it was better than 95% accurate. Secondly, the Bayesian method will give measures that you can advertise as correct (and people like Telcordia are very stubborn about it) with a relatively short and inexpensive experiment.
Right. Just an ambiguity in what you were saying: 'after time t'. You're giving failure in time (t-t0), PMF instead of PDF.
This trick is the key to the Bayesian estimation methods I mentioned above. Check out the CiteSeer and Google links for more papers than you ever wanted to read.
Now, consider if you get >1 failure in that second hour: there are several possible explanations:
In general, you've got to test a lot of items for a fair chunk of the expected lifetime to get the confidence level up.
Thanks you. I will try to reign ineffectively from home, as all good geeks would prefer.
Sadly, on some online geek test recently, I not only scored high, I scored #4 among everyone who had ever taken the test.
Then I couldn't decide whether to be embarrassed at being that much of a geek, or at only scoring fourth.
Work it out; IF it's an exponential, THEN for failure rate parameter lambda, mean time between failures is 1/lambda. It's too painful to try to show this in HTML, but Trivedi's book cited above ("favorite text") shows the derivation.
Yes. Strictly -- as someone else pointed out -- the failures tend to have a so-called "bathtub" distribution. That is, there's a high failur rate at first ("infant mortality") followed by a long Poisson/Markovian/exponential (you pick your term) stretch, followed by a higher failure rate as it get old. In general, the "lifetime" of a component is the time to the inflection at the end of the exponential portion.
Sure, but in a "bathtub" distribution it's approximately exponential over most of the lifetime, so it's a decent approximation.
"There are lots of planes and lots of years" is the right answer: I don't recall the exact figures right now, but at the time of the TWA 800 crash I predicted that it would turn out to be an airframe failure (on the heuristic of preferring failure to malice) because when I worked the numbers it turned out that MTBF of 747s was about that same 1e10.
By the way, your supposition about answer (1) is correct. It's just a definitional thing: we're really talking about proximate and root causes. You don't count it as an airframe failure if someone throws a cigarette butt into the trash and smoke and fire causes the crash, and while the airframe certainly fails if someone flies the plane into a mountain, that's still not the immediate cause.
Well, I write books and screenplays, but I do that on a computer so I don't think that counts. I'm a docent at the new Space Odyssey exhibit, but that's all computer-controlled....
... no, the treadmill is computer controlled.
I work out
Damn, I dunno: sleep?