James Webb is designed as a primarily IR telescope. You don't want to park it in low orbit because it would be subject to a lot of IR interference from the planet. So they're putting it at L2, which is a long way away from the planet. Unfortunately we can only really send astronauts to low orbit to fix things. So there's not really much choice about the repairability of the JW.
That sounds about right, for a general average among people who aren't trying to get pregnant. Pregnancy is less of a sure thing than some would have you believe.
That's too bad. I have a degree in computer science and have worked with machine learning for the last twenty or so years. The progress in the last five years has been incredible. Today a student can build a system on their own computer that easily solves problems that the my-brain-is-magic types thought were unsolvable ten years ago. That doesn't guarantee that the progress will continue, but it looks promising, and is already incredibly useful.
Your UID is low enough to remember the real Slashdot, when computers that could translate the world's languages in better than realtime, drive cars better than humans, and beat the best chess players would definitely have been AI.
We seem to have been invaded by irritable American political pundits in the meantime.
Female to male spread is much harder than the other direction. And male to female spread via vaginal intercourse is much harder than male to male or female spread via anal intercourse. Condom use for casual sex is also a lot more common when there's a risk of pregnancy.
1970s medical technology wasn't anything like what we've got now. Identifying and isolating a new virus is still a tricky undertaking. In the 1970s it was much more so. A Nobel prize was awarded for the discovery of HIV and its link to AIDS.
Medicine isn't nearly so scientific as you probably think. An average doctor might see a weird case once or twice that was actually AIDS but that's hard to separate from all the other weird cases they see on a daily basis (House: maybe it's lupus!). In the 70s there certainly weren't any good central databases for general medical records, and there still aren't, especially in the US, because of privacy and insurance concerns.
If you were a doctor in the 70s and you saw a malnourished person waste away and die, would you think "gee, it's horrible we let people starve on the street in America" or "OMG, this is the start of a plague that will sweep the world in twenty years"?
In the early 80s, when the number of patients increased, doctors, especially those who worked in gay communities, who were most at risk, DID notice unusual numbers of people dying and did report and track it.
Influenza outbreak modelling generally assumes that about 1/3 of transmission is from asymptomatic or presymptomatic carriers. Virus shedding starts around a day before symptoms appear, even in the symptomatic. There are estimates that 75% of common cold infections are asymptomatic.
Any generative model, and most of the modern systems are either generative or trivially easy to modify to be so, includes an internal model. Antagonistic training explicitly exploits this feature with two systems, one that tries to learn to spot real data from faked, and another that tries to learn to fool the first one.
It seems like changing the admin password to something random would work perfectly well. If the clueless user needed to change something they'd have to reset to factory defaults and in learning how to do that perhaps they'd learn about changing the password. Likely the vast majority would never even notice.
This seems unlikely, since there have been around 500 above ground nuclear detonations, some of which were far larger than anything the US or Russia would likely use in a modern war.
Anybody who knows anything about watches could have seen this coming. Everybody used to wear a watch. Then everybody started carrying a smartphone and now watches are rare items worn as decoration by a few people.
You have the ability to do "full class 5 driving." We're pretty familiar withe sensors included in homo sapiens sapiens, particularly those are are used while driving.
The idiots don't understand the difference between group tendencies and individual prediction. Picking people for jobs based on genes is stupid because you get a much better (although for many things still not very good) indication by giving an actual performance test.
And no dcblogs, nobody chooses basketball players by their genes. They're selected by their *height* and their ability to play basketball.
Mmm, that's productivity per worker. If that Frenchman, who manages to be within about 1.5% as productive as an American while working 20% less tells an American something about productivity, the American might want to grab a pen and take notes.
If you're not entering a proper address you don't know where you want to go and you DO need a search engine. You're just used to that search engine hitting the equivalent of "I'm feeling lucky" for you.
It's not different, except that this time there's nowhere to hide.
When the machines replaced unskilled labor the unskilled laborers either became "skilled" or servants (or both). Now that the machines are about to replace skilled labor we can continue the charade and all become servants to each other, or we can stop being stupid and enjoy real freedom.
No, he's a realist. The world is being reinvented right now, and our silly play money is going to have to be reinvented to match.
Last time this happened the world changed entirely. We call it the industrial revolution. This time it's going to happen faster, and the changes are going to be much more drastic.
To use your terminology, the capacity of a neural network to learn more complex functions is roughly governed by the number of parameters. This is true of most machine learning algorithms. More complex functions allow you to overfit simple data, or to learn reasonable models for more complex data. More complex data includes things like more image classes, more words, more relationships among elements, more states, etc. In other words, many of the things that we think of when we talk about "memory." Recurrent and analogous neural networks already have a memory of prior states but these are built into the network architecture as additional connections. Other people have experimented with more flexible memories, also built-in and composed of connections. These are limited as I said in my OP and so do not represent "the vast data storage of conventional computers."
A neural network normally uses it's own connection weights as "memory" or storage. There's a tradeoff between making a network with enough parameters to store lots of information and making one that's fast, efficient and doesn't overfit problems. In many cases you're practically limited by how much memory you've got on your video card. Having a neural net that can learn to store some information separately from its own processing apparatus is interesting.
Slashdot Mirror
Yeah. That's how math works. -200% of negative profit is a good thing.
This is Slashdot. Unlike accountants, many people here can handle integers, sometimes even real numbers, as opposed to just the naturals.
Good luck launching a single piece mirror that big.
James Webb is designed as a primarily IR telescope. You don't want to park it in low orbit because it would be subject to a lot of IR interference from the planet. So they're putting it at L2, which is a long way away from the planet. Unfortunately we can only really send astronauts to low orbit to fix things. So there's not really much choice about the repairability of the JW.
That sounds about right, for a general average among people who aren't trying to get pregnant. Pregnancy is less of a sure thing than some would have you believe.
Yeah, funny that hey? I wonder where that number-of-nines terminology originated.
That's too bad. I have a degree in computer science and have worked with machine learning for the last twenty or so years. The progress in the last five years has been incredible. Today a student can build a system on their own computer that easily solves problems that the my-brain-is-magic types thought were unsolvable ten years ago. That doesn't guarantee that the progress will continue, but it looks promising, and is already incredibly useful.
Your UID is low enough to remember the real Slashdot, when computers that could translate the world's languages in better than realtime, drive cars better than humans, and beat the best chess players would definitely have been AI.
We seem to have been invaded by irritable American political pundits in the meantime.
Female to male spread is much harder than the other direction. And male to female spread via vaginal intercourse is much harder than male to male or female spread via anal intercourse. Condom use for casual sex is also a lot more common when there's a risk of pregnancy.
1970s medical technology wasn't anything like what we've got now. Identifying and isolating a new virus is still a tricky undertaking. In the 1970s it was much more so. A Nobel prize was awarded for the discovery of HIV and its link to AIDS.
Medicine isn't nearly so scientific as you probably think. An average doctor might see a weird case once or twice that was actually AIDS but that's hard to separate from all the other weird cases they see on a daily basis (House: maybe it's lupus!). In the 70s there certainly weren't any good central databases for general medical records, and there still aren't, especially in the US, because of privacy and insurance concerns.
If you were a doctor in the 70s and you saw a malnourished person waste away and die, would you think "gee, it's horrible we let people starve on the street in America" or "OMG, this is the start of a plague that will sweep the world in twenty years"?
In the early 80s, when the number of patients increased, doctors, especially those who worked in gay communities, who were most at risk, DID notice unusual numbers of people dying and did report and track it.
Influenza outbreak modelling generally assumes that about 1/3 of transmission is from asymptomatic or presymptomatic carriers. Virus shedding starts around a day before symptoms appear, even in the symptomatic. There are estimates that 75% of common cold infections are asymptomatic.
Any generative model, and most of the modern systems are either generative or trivially easy to modify to be so, includes an internal model. Antagonistic training explicitly exploits this feature with two systems, one that tries to learn to spot real data from faked, and another that tries to learn to fool the first one.
It seems like changing the admin password to something random would work perfectly well. If the clueless user needed to change something they'd have to reset to factory defaults and in learning how to do that perhaps they'd learn about changing the password. Likely the vast majority would never even notice.
This seems unlikely, since there have been around 500 above ground nuclear detonations, some of which were far larger than anything the US or Russia would likely use in a modern war.
Anybody who knows anything about watches could have seen this coming. Everybody used to wear a watch. Then everybody started carrying a smartphone and now watches are rare items worn as decoration by a few people.
You have the ability to do "full class 5 driving." We're pretty familiar withe sensors included in homo sapiens sapiens, particularly those are are used while driving.
The idiots don't understand the difference between group tendencies and individual prediction. Picking people for jobs based on genes is stupid because you get a much better (although for many things still not very good) indication by giving an actual performance test.
And no dcblogs, nobody chooses basketball players by their genes. They're selected by their *height* and their ability to play basketball.
That's a fine definition of productivity, but it's a meaningless way to compare workers or work styles in different places.
Mmm, that's productivity per worker. If that Frenchman, who manages to be within about 1.5% as productive as an American while working 20% less tells an American something about productivity, the American might want to grab a pen and take notes.
If you're not entering a proper address you don't know where you want to go and you DO need a search engine. You're just used to that search engine hitting the equivalent of "I'm feeling lucky" for you.
It's not different, except that this time there's nowhere to hide.
When the machines replaced unskilled labor the unskilled laborers either became "skilled" or servants (or both). Now that the machines are about to replace skilled labor we can continue the charade and all become servants to each other, or we can stop being stupid and enjoy real freedom.
No, he's a realist. The world is being reinvented right now, and our silly play money is going to have to be reinvented to match.
Last time this happened the world changed entirely. We call it the industrial revolution. This time it's going to happen faster, and the changes are going to be much more drastic.
Not exactly.
To use your terminology, the capacity of a neural network to learn more complex functions is roughly governed by the number of parameters. This is true of most machine learning algorithms. More complex functions allow you to overfit simple data, or to learn reasonable models for more complex data. More complex data includes things like more image classes, more words, more relationships among elements, more states, etc. In other words, many of the things that we think of when we talk about "memory." Recurrent and analogous neural networks already have a memory of prior states but these are built into the network architecture as additional connections. Other people have experimented with more flexible memories, also built-in and composed of connections. These are limited as I said in my OP and so do not represent "the vast data storage of conventional computers."
A neural network normally uses it's own connection weights as "memory" or storage. There's a tradeoff between making a network with enough parameters to store lots of information and making one that's fast, efficient and doesn't overfit problems. In many cases you're practically limited by how much memory you've got on your video card. Having a neural net that can learn to store some information separately from its own processing apparatus is interesting.