It's not as if this hasn't actually been studied. Solar with batteries is viable for some locations but not others, based on the maximum length storm. Nor is it necessarily either/or. PV gets you a lot more watts/dollar, so even if you had one of these you might well decide to use both.
You'd need significant industrial capability -- which translates into literally tons of mass -- to bootstrap that scenario. Mining equipment is not light.
It's one of those scenarios that's easy to imagine working once you got it up and running, but is hard to image how to get up and running.
1 KW is about what you'd need to run a popup toaster or a blow dryer. This is, from a NASA engineer's perspective, a huge amount of power, but you couldn't run your neighborhood hair salon on it.
It's an engineering problem. You surely could get some combination of solar and battery to work on the Martian surface, but it would impose design and operational constraints -- constraints which could be mitigated with money.
Presumably they crunched the numbers and developing an entirely novel compact reactor looks like it could be a win. However lets imagine this "Kilopower" project is a total failure; that doesn't mean that a Mars habitation mission couldn't proceed, it'd just cost more to get a certain amount done.
It tried to be fair and actually failed, because it uses a methodology that clearly wasn't designed by a statistician.
The program uses over a hundred factors in its classification scheme, but statisticians and data scientists make a point of pruning factors because long experience has shown that introducing many irrelevant factors actually reduces predictive accuracy. And just because race is not an explicit factor doesn't mean that the algorithm is race blind either. It's entirely feasible to given the huge number of factors involve to recover the subject's race with a better-than-chance reliabilty, whether explicitly or implicitly; intentionally or even by accident.
Now the program's score is equally correlated with reoffending rates whether the subject happens to be white or black, which sounds impressive and color-blind -- to a layman. To a mathematician not so much. It's actually quite easy to produce this result by tweaking your model, implicitly recovering race in the manner suggested above and forcing it to produce a result that looks right -- in aggregate.
But what a statistician wants to know is about conditional probabilities, and it turns out that when applied to retrospective data the program is twice as likely to commit a type 1 error (falsely predicting reoffending) for black subjects as white. If this makes the whole process of achieving fairness sound hard, that's because it is. Color-blindness in aggregate isn't the same as color blindness on a case-by-case basis, and that's the thing that actually matters.
Ultimately you want criminal justice decisions to be based on reason, and mathematics is the purest form of reason there is. And because you want those decisions to be based on reason, they have to be transparent. Secret methods for arriving at decision-making are fundamentally antithetical to our concept of justice.
Alternatively: vendor oversells effectiveness of its proprietary, secret sauce methodology and doesn't like any independent evaluation of its products unless it's favorable. Customers, having a naive faith in technology, buy anyways, which produces exactly the results you mention: programs will be forever terrible at this task. Why should anyone bother to make a program good when customers will shell out good money for mediocre?
Well, there's bad (i.e., stupid) clients too. They're responsible for a lot of bad software.
If a customer wants to buy magic software without an understanding of what it does or proof that it even works, what are the programmers supposed to do about that? They just report to work and build what their boss tells them to build, and he tells them to build what the customer will buy.
I don't think it for the most part blows more than it ever did. It's all about creating a product that will sell, and occasionally someone makes something that's good by accident. That's why it's called "pop" -- it's like soda pop: it's supposed to be effortlessly consumable.
The big difference today is Autotune. It's not that as a tool it's inherently bad, it's that it allows the music industry to do two things: produce a more predictably uniform product and focus on packaging -- and by packaging I mean looking hot. Would Janis Joplin have a chance in the industry today? She didn't look like a fashion model, didn't dance, and didn't have a big rack.
The system should _never_ _ever_ send "false alarm".
Sure, but how do you design a system where that never happens?
Adding ad hoc messages is an even worse idea. At some point a politician will get the bright idea of using the system for trivial notifications (like "air quality alerts") that are better handled via other channels.
And how do you design a system that cannot deliberately be misused?
The problem with some notions of equality is that they don't take into account the fact that money, like everything else has diminishing returns.
About ten years ago researchers looking at this question discovered that while perceptions of personal success do continue to rise with income, income above $70,000 ($98,000 in current dollars) has essentially zero impact on emotional well-being -- the actual quality of life experienced by the individual on a day to day basis. In other words while our wants are flexible, our actual needs are quite modest.
This suggests to me that if you look at $98,000 as a consumption level, and factor in technologically mediated productivity changes, practical equality is something achievable in the historically speaking near future. A few countries are close to achieving this; if you look at the countries with the highest reports of well-being they're all wealthy countries with a low gini coefficients, which means that they have the largest proportion of people with solidly middle-class incomes.
In general I don't see any reason to care if someone wants to become the next Elon Musk, except so far as such people are able to buy politicians with their money. When politicians are working for the super-wealthy they're putting their efforts into things that will make literally nobody happier.
This does not surprise me at all, especially as the cost is arrived at by taking the total campaign cost divided by the number of soldiers.
Everything the US military does costs an eye-popping amount. The V22 Osprey costs $64000/hour to operate. The Bradley AFV cost over $50 for every mile driven. Recoilless rifle ammunition runs between $500 and $3000 per round. Every time an A10 opens up its mighty 3900/ round/minute cannon, each of those rounds costs $150.
The current administration's plans for increases in troop levels in Afghanistan are expected to cost the US taxpayer over a trillion dollars when all the downstream costs are included. In return they hope to secure access to about a trillion dollars in mineral reserves for US companies.
First of all, Agile doesn't work in every situation unless you stretch the definition to include non-agile practices where warranted. Second, the distinction between users and testers isn't as clean as you suggest. Users *are* testers until they become habituated to the system.
Give that some systems are worse than others in inviting operator error, you can't just assume it's not the tech because operator error was involved. However even if the tech is as good as humans can possibly make it, that still wouldn't prevent operator error.
This kind of fault is hard to test for, because it's a non-functional requirement. You can't simply do a functional test and check off "prevent accidental message from being sent". At best you can simulate various scenarios, but those simulations are unreliable because you're dealing with testers, not people who are habituated to the system and who thus use it differently.
Clearly there were several kinds of operational faults here that may have been compounded by design flaws. But one of the operational mistakes was purely a matter of planning: not programming in a "false alarm" message to be sent after the inevitable operator error. This also suggests a design shortcoming in the system in that designers didn't anticipate the need to ever issue an ad hoc message on short notice.
Slashdot Mirror
It's not as if this hasn't actually been studied. Solar with batteries is viable for some locations but not others, based on the maximum length storm. Nor is it necessarily either/or. PV gets you a lot more watts/dollar, so even if you had one of these you might well decide to use both.
Wouldn't be very exciting in a 1.3 horsepower car.
The reason for all those "robot runs amok" stories is that "human runs amok" isn't a story premise; it's more like a chronic condition.
You'd need significant industrial capability -- which translates into literally tons of mass -- to bootstrap that scenario. Mining equipment is not light.
It's one of those scenarios that's easy to imagine working once you got it up and running, but is hard to image how to get up and running.
You could load it in your bass boat and then circumnavigate the world on your trolling motor -- if it were a small trolling motor.
1 KW is about what you'd need to run a popup toaster or a blow dryer. This is, from a NASA engineer's perspective, a huge amount of power, but you couldn't run your neighborhood hair salon on it.
It's an engineering problem. You surely could get some combination of solar and battery to work on the Martian surface, but it would impose design and operational constraints -- constraints which could be mitigated with money.
Presumably they crunched the numbers and developing an entirely novel compact reactor looks like it could be a win. However lets imagine this "Kilopower" project is a total failure; that doesn't mean that a Mars habitation mission couldn't proceed, it'd just cost more to get a certain amount done.
Kinda jumping the gun on Texas there. Still a few more years before that flips.
It brings in jobs, and the workers pay taxes. At least that's the theory.
I think they're fine if they provide a paper audit trail.
And of course the way you know tomfoolery went into the the computer in this case is that you don't like the answer that came out.
It tried to be fair and actually failed, because it uses a methodology that clearly wasn't designed by a statistician.
The program uses over a hundred factors in its classification scheme, but statisticians and data scientists make a point of pruning factors because long experience has shown that introducing many irrelevant factors actually reduces predictive accuracy. And just because race is not an explicit factor doesn't mean that the algorithm is race blind either. It's entirely feasible to given the huge number of factors involve to recover the subject's race with a better-than-chance reliabilty, whether explicitly or implicitly; intentionally or even by accident.
Now the program's score is equally correlated with reoffending rates whether the subject happens to be white or black, which sounds impressive and color-blind -- to a layman. To a mathematician not so much. It's actually quite easy to produce this result by tweaking your model, implicitly recovering race in the manner suggested above and forcing it to produce a result that looks right -- in aggregate.
But what a statistician wants to know is about conditional probabilities, and it turns out that when applied to retrospective data the program is twice as likely to commit a type 1 error (falsely predicting reoffending) for black subjects as white. If this makes the whole process of achieving fairness sound hard, that's because it is. Color-blindness in aggregate isn't the same as color blindness on a case-by-case basis, and that's the thing that actually matters.
Ultimately you want criminal justice decisions to be based on reason, and mathematics is the purest form of reason there is. And because you want those decisions to be based on reason, they have to be transparent. Secret methods for arriving at decision-making are fundamentally antithetical to our concept of justice.
Alternatively: vendor oversells effectiveness of its proprietary, secret sauce methodology and doesn't like any independent evaluation of its products unless it's favorable. Customers, having a naive faith in technology, buy anyways, which produces exactly the results you mention: programs will be forever terrible at this task. Why should anyone bother to make a program good when customers will shell out good money for mediocre?
Well, there's bad (i.e., stupid) clients too. They're responsible for a lot of bad software.
If a customer wants to buy magic software without an understanding of what it does or proof that it even works, what are the programmers supposed to do about that? They just report to work and build what their boss tells them to build, and he tells them to build what the customer will buy.
I'm not saying looks didn't matter. But as record label you couldn't have as much of everything you wanted.
I don't think it for the most part blows more than it ever did. It's all about creating a product that will sell, and occasionally someone makes something that's good by accident. That's why it's called "pop" -- it's like soda pop: it's supposed to be effortlessly consumable.
The big difference today is Autotune. It's not that as a tool it's inherently bad, it's that it allows the music industry to do two things: produce a more predictably uniform product and focus on packaging -- and by packaging I mean looking hot. Would Janis Joplin have a chance in the industry today? She didn't look like a fashion model, didn't dance, and didn't have a big rack.
Lockheed Constellation.
The system should _never_ _ever_ send "false alarm".
Sure, but how do you design a system where that never happens?
Adding ad hoc messages is an even worse idea. At some point a politician will get the bright idea of using the system for trivial notifications (like "air quality alerts") that are better handled via other channels.
And how do you design a system that cannot deliberately be misused?
Sure, it varies; about $100,000 is an average figure. But again the point isn't income, it's consumption levels.
I should have been clearer: it will bring the total cost for the war into the trillion dollar range, counting all downstream costs.
The problem with some notions of equality is that they don't take into account the fact that money, like everything else has diminishing returns.
About ten years ago researchers looking at this question discovered that while perceptions of personal success do continue to rise with income, income above $70,000 ($98,000 in current dollars) has essentially zero impact on emotional well-being -- the actual quality of life experienced by the individual on a day to day basis. In other words while our wants are flexible, our actual needs are quite modest.
This suggests to me that if you look at $98,000 as a consumption level, and factor in technologically mediated productivity changes, practical equality is something achievable in the historically speaking near future. A few countries are close to achieving this; if you look at the countries with the highest reports of well-being they're all wealthy countries with a low gini coefficients, which means that they have the largest proportion of people with solidly middle-class incomes.
In general I don't see any reason to care if someone wants to become the next Elon Musk, except so far as such people are able to buy politicians with their money. When politicians are working for the super-wealthy they're putting their efforts into things that will make literally nobody happier.
This does not surprise me at all, especially as the cost is arrived at by taking the total campaign cost divided by the number of soldiers.
Everything the US military does costs an eye-popping amount. The V22 Osprey costs $64000/hour to operate. The Bradley AFV cost over $50 for every mile driven. Recoilless rifle ammunition runs between $500 and $3000 per round. Every time an A10 opens up its mighty 3900/ round/minute cannon, each of those rounds costs $150.
The current administration's plans for increases in troop levels in Afghanistan are expected to cost the US taxpayer over a trillion dollars when all the downstream costs are included. In return they hope to secure access to about a trillion dollars in mineral reserves for US companies.
First of all, Agile doesn't work in every situation unless you stretch the definition to include non-agile practices where warranted. Second, the distinction between users and testers isn't as clean as you suggest. Users *are* testers until they become habituated to the system.
And what material. Dice rolls could probably outperform slashdot readers on article summaries.
Give that some systems are worse than others in inviting operator error, you can't just assume it's not the tech because operator error was involved. However even if the tech is as good as humans can possibly make it, that still wouldn't prevent operator error.
This kind of fault is hard to test for, because it's a non-functional requirement. You can't simply do a functional test and check off "prevent accidental message from being sent". At best you can simulate various scenarios, but those simulations are unreliable because you're dealing with testers, not people who are habituated to the system and who thus use it differently.
Clearly there were several kinds of operational faults here that may have been compounded by design flaws. But one of the operational mistakes was purely a matter of planning: not programming in a "false alarm" message to be sent after the inevitable operator error. This also suggests a design shortcoming in the system in that designers didn't anticipate the need to ever issue an ad hoc message on short notice.