The rules didn't change because blogs started making money. Rather, now bloggers have something to lose, and they don't want to lose it. And worse than losing something would be throwing it away by pointing out your own problems to the world. People's first instinct is hide, not voice, their own problems. Unlike traditional news sources, blogs haven't had the longevity to know that transparency is the best policy.
Do you feel that e-voting companies are honestly addressing the complaints raised against their products or are relying on lobbying and politics to override concern? Please give a concrete example of a company working with detractors to address their issues.
Google is going to return what people think censorship is. This is like the case of commonly misused words--the general population is not the group to trust. Try the dictionary. Or I can make it easy on you: censorship is the act deletion by an official for the purpose of suppressing parts deemed objectionable on moral, political, military, or other grounds.
A perennial example is that the Slashdot mod system is "censorship." Other examples are spattered throughout nearly every free speech issue, generally due to errors in thinking.
I looked at the recurrent neural network page, and although the page was somewhat terse, I think it is at least close to the strategy we followed. We did not model time within the system. A picture is really the best way to show how the nodes were structured, but I'll do my best to explain. I'm going to use arbitrary numbers, but varying the counts affects how the network behaves in sometimes interesting ways. It was a couple years ago now, so I may not accurately remember the exact construction anymore, but hopefully I get close.
Simple setup:
A "set" is a collection of five TLU nodes. Initially, some number of inputs are "dead"; in this example it will be a single node for 20%. The output of five sets of nodes are used as inputs into a child set. The output of this child set is then fed back into the "dead" nodes (will retaining the original inputs for the other nodes.) I tied a bunch of these systems together to create the full network. The network runs until a) it stabalizes or b) all the outputs are either 0 or 1.
Less stupid setup:
Five sets are fed inputs from some source (I used images and mathematical functions). Their output is taken and fed back into them as input. I realize that sentence wasn't very clear, so I'll try again. Let us have sets A, B, C, D, and E. They produce output Ao, Bo,..., Eo. After getting this output, Ao,..., Eo is fed into set A as input. This is done for sets B, C, D, and E as well. Naturally this produces infinite recursion, but it stabalizes pretty fast. A naive approach is to allow N recursions, but we actually used a parallel network and swapped the outputs waiting for the output to either be equal or not equal depending on boring academic expectations.
Summary:
Hopefully I didn't confuse the issue too much. These configurations were just the first couple of homework assignments and were quickly replaced by self-training networks, which are simultaneously more useful and more interesting. For the self-training networks, I used a fixed training length and a fixed number of trainings.
This seems like the sort of thing that could happen if the brain's speech area's neurons somehow became trained to stop delivering impulses for "normal" speech. In this case, it would be theoretically possible to train the network back to normal levels. Of course, it could be something completely different.
I feel that there are a lot of softeners in there. I'd hoped that it was clear that I wasn't trying to advance a hypothesis but rather pointing out a possibility. Since three people misinterpreted my intent, odds are good I failed on that point.
The power would return to the circuit... I hope. I don't think this is probably worth arguing about anymore. We both seem fairly entrenched. I'm afraid that I'm in the Twin Cities. I was in Milwalkee last week though. What a chance we missed!
The context that I studied in could use feedback to strengthen inputs causing feedback loops or weaken inputs to simulate network damage (or both on different nodes). Clearly we used a different paradigm than whatever you are talking about.
"doesn't mean we know enough to make claims about the root causes of neurological diseases because we've taken an AI course"
Yeah, I wasn't trying to say that. I noticed the similarity between what I studied and what Scott alluded to in how he tried to get his voice back.
And I defer to your superior knowledge of neural networks. Like I said, it was an undergraduate class, and I wasn't trying to misrepresent my knowledge. As far as weights and noise tolerance go, that was the exact focus of the neural network section of my AI class, so I actually did study that part. Again, this is because it was the area the professor was most interested in. He uses a neural network to categorize nerve impulses and to categorize cat scan brain "slices."
That is simply not true. The brain is composed of neurons, and the way that neurons work is exactly what neural networks model. Current technology can't even come close to the complexity of human brains, so studies are naturally simplified. Current technology can do a pretty good job of modeling the brains of simpler creatures, such as lobsters. I haven't heard of any studies that have found anything particularly useful from modeling an entire brain.
Frankly, the wikipedia article can claim that it is "much debated," but that doesn't really change the facts. I've read a dozen articles in refereed journals that support my claims. Nature has published at least three. (Most articles published on neural networks do not draw parallels to biological systems; the reason that I know about the biological side is that my AI professor is the lead programmer on a pain research project with Mayo Clinic in Rochester, Minnesota.)
I asked a physics friend about this, and he figured out the part I was missing. The efficiency of transformers is measured by how much energy they transform for an amount of power drawn. If the transformer is connected, it draws X power whether that power is used or not. A percent of that power is transformed, and the percent that is not transformed becomes heat from resistance, etc. The waste heat is the untransformed power lost to resistance. The successfully transformed power heads back into the circuit and is never used.
On the other hand, a space-heater's efficiency is measured in how much of its drawn power is converted to heat. Most transformers have ~80% efficiency. Let's say we've got a poor space heater with only 80% efficency as well. Both draw 100 watts. In this case, the transformer turns 20 watts to heat and the space heater turns 80 watts to heat.
"I sit down and I play the game and I have a blast with it... co-op's especially bad."
He means that co-op is especially good. A typical (ab)use of bad in this way is like the following example: "I like video games. I've been wasting hours on them. Videogame X is especially bad." In this context, it is clear that Videogame X is especially time wasteful and, by implication, is the most-liked of the speaker's videogames.
Wikipedia has a nice article on Spasmodic Dysphonia.
As the blog indicates, this is thought to be a neurological condition. When I was studying AI as an undergrad, we learned a lot about neural networks. This seems like the sort of thing that could happen if the brain's speech area's neurons somehow became trained to stop delivering impulses for "normal" speech. In this case, it would be theoretically possible to train the network back to normal levels. Of course, it could be something completely different.
They aren't open for debate, they are specifically retained by the people. Read your own link. For example, the Constitution doesn't address the right to jumping. That doesn't mean that my right to jump up and down is open for debate. It means that my right to jump up and down is specifically protected from Federal laws.
Let's take the case of a 40 watt light bulb and a 40 watt space heater. The 40 watt bulb consumes electricity and produces light and heat. On the other hand, the space heater consumes electricity and produces only heat. With the same amount of consumed power, the space heater will produce more heat because none of the power is being used to produce non-heat outputs (in this case light.) This is the conservation of energy--light and heat are both energy, and the same amount of electricity can't magically produce extra energy just because it happens to be used by a light bulb.
Now let me address your exact question, because it doesn't have the benefit of the light to simplify the example. Let's say that we have a 20 watt transformer hooked onto nothing; it's just an open circuit with a transformer slapped in there. The transformer produces heat for several reasons, including the resistance it produces in the circuit. On the other hand, a space heater doesn't have nearly as many non-heat losses as a transformer. (It doesn't have magnetostriction, mechanical losses, or hysteresis losses for instance.)
I don't see how saving two hundred dollars a year by using electricity rather than propane, affects which electrical source produces heat most efficiently. An electric space heater designed for the task is a more efficient heat source than a hodge-podge of random electronic gadgets spewing waste heat. In other words, both may save you money over propane, but an electric heater will save you money over waste heat.
Your analogy is like saying that going down a hill saves gas because you can coast--it ignores that fact that you had to use more gas to climb the hill in the first place.
While your devices may produce waste heat, they don't produce it as efficiently as something designed for the purpose. A space heater would save you more money. Suggesting that devices continue to be designed in a deliberately inefficient matter strikes me as somewhat foolish.
I can think of a dozen bars within 20 minutes of me (Saint Paul, MN, USA) that have regular poker tournaments. Some have entry fees, but most do not. The prizes differ from location to location, but most have cash prizes. The bars make money from their expensive drinks, while the contestants get a fun evening of poker. I think you are not looking hard enough for a game.
It is larger than a boomerang. The videos are of a prototype. The website lists dimensions of "two to ten feet in length" and "two to six inches in height".
What was the last war the USA fought against anyone with smart weapons? My guess is that this is being marketed toward the needs of warfar against low-tech enemies using guerilla techniques. I could also anticipate use as a "look over this next hill" tool where you only need 30 seconds of flight. If a smart missile is only 30 seconds out I think you may have bigger problems.
I think there is great future in bionics. In addition to limbs as discussed in this submission, scientists have various approaches to bionic sight as well. This subject is truly fascinating. Here is a BBC article on a different project.
Interestingly and unfortunately, much advanced and successful bionics research is being done in South America because of restrictive laws in more typical countries. While I understand the need to protect patients, research for a paper I wrote two years ago indicates that the most successful scientists are pragmatically drawn away from first-class research institutes.
The problem with the commentary for me was that there was no way to fast-forward, skip, or rewind. Some parts I found tremendously boring, while others were fascinating. I only want to hear the fascinating parts. Going slower by watching physics is not something that I'm interested in. I'm good enough at FPS games that I can typically just do something a second time if I want to see it again.
I played a bit of Episode 1 with the commentary on and found it interesting, though not enough to finish the entire episode a second time. I'm looking forward to Episode 2, which looks better in every way.
Slashdot Mirror
Who cares? No one gets karma from +Funny.
The rules didn't change because blogs started making money. Rather, now bloggers have something to lose, and they don't want to lose it. And worse than losing something would be throwing it away by pointing out your own problems to the world. People's first instinct is hide, not voice, their own problems. Unlike traditional news sources, blogs haven't had the longevity to know that transparency is the best policy.
Do you feel that e-voting companies are honestly addressing the complaints raised against their products or are relying on lobbying and politics to override concern? Please give a concrete example of a company working with detractors to address their issues.
Google is going to return what people think censorship is. This is like the case of commonly misused words--the general population is not the group to trust. Try the dictionary. Or I can make it easy on you: censorship is the act deletion by an official for the purpose of suppressing parts deemed objectionable on moral, political, military, or other grounds.
A perennial example is that the Slashdot mod system is "censorship." Other examples are spattered throughout nearly every free speech issue, generally due to errors in thinking.
I looked at the recurrent neural network page, and although the page was somewhat terse, I think it is at least close to the strategy we followed. We did not model time within the system. A picture is really the best way to show how the nodes were structured, but I'll do my best to explain. I'm going to use arbitrary numbers, but varying the counts affects how the network behaves in sometimes interesting ways. It was a couple years ago now, so I may not accurately remember the exact construction anymore, but hopefully I get close.
..., Eo. After getting this output, Ao, ..., Eo is fed into set A as input. This is done for sets B, C, D, and E as well. Naturally this produces infinite recursion, but it stabalizes pretty fast. A naive approach is to allow N recursions, but we actually used a parallel network and swapped the outputs waiting for the output to either be equal or not equal depending on boring academic expectations.
Simple setup:
A "set" is a collection of five TLU nodes. Initially, some number of inputs are "dead"; in this example it will be a single node for 20%. The output of five sets of nodes are used as inputs into a child set. The output of this child set is then fed back into the "dead" nodes (will retaining the original inputs for the other nodes.) I tied a bunch of these systems together to create the full network. The network runs until a) it stabalizes or b) all the outputs are either 0 or 1.
Less stupid setup:
Five sets are fed inputs from some source (I used images and mathematical functions). Their output is taken and fed back into them as input. I realize that sentence wasn't very clear, so I'll try again. Let us have sets A, B, C, D, and E. They produce output Ao, Bo,
Summary:
Hopefully I didn't confuse the issue too much. These configurations were just the first couple of homework assignments and were quickly replaced by self-training networks, which are simultaneously more useful and more interesting. For the self-training networks, I used a fixed training length and a fixed number of trainings.
That isn't exactly what I meant. If you take a battery and touch a wire to each end it will drain the whole battery without generating very much heat.
The power would return to the circuit... I hope. I don't think this is probably worth arguing about anymore. We both seem fairly entrenched. I'm afraid that I'm in the Twin Cities. I was in Milwalkee last week though. What a chance we missed!
I meant to say what that context was: chained sets of threshold logic units modeled in Mathematica.
The context that I studied in could use feedback to strengthen inputs causing feedback loops or weaken inputs to simulate network damage (or both on different nodes). Clearly we used a different paradigm than whatever you are talking about.
"doesn't mean we know enough to make claims about the root causes of neurological diseases because we've taken an AI course"
Yeah, I wasn't trying to say that. I noticed the similarity between what I studied and what Scott alluded to in how he tried to get his voice back.
And I defer to your superior knowledge of neural networks. Like I said, it was an undergraduate class, and I wasn't trying to misrepresent my knowledge. As far as weights and noise tolerance go, that was the exact focus of the neural network section of my AI class, so I actually did study that part. Again, this is because it was the area the professor was most interested in. He uses a neural network to categorize nerve impulses and to categorize cat scan brain "slices."
That is simply not true. The brain is composed of neurons, and the way that neurons work is exactly what neural networks model. Current technology can't even come close to the complexity of human brains, so studies are naturally simplified. Current technology can do a pretty good job of modeling the brains of simpler creatures, such as lobsters. I haven't heard of any studies that have found anything particularly useful from modeling an entire brain.
Frankly, the wikipedia article can claim that it is "much debated," but that doesn't really change the facts. I've read a dozen articles in refereed journals that support my claims. Nature has published at least three. (Most articles published on neural networks do not draw parallels to biological systems; the reason that I know about the biological side is that my AI professor is the lead programmer on a pain research project with Mayo Clinic in Rochester, Minnesota.)
I asked a physics friend about this, and he figured out the part I was missing. The efficiency of transformers is measured by how much energy they transform for an amount of power drawn. If the transformer is connected, it draws X power whether that power is used or not. A percent of that power is transformed, and the percent that is not transformed becomes heat from resistance, etc. The waste heat is the untransformed power lost to resistance. The successfully transformed power heads back into the circuit and is never used.
On the other hand, a space-heater's efficiency is measured in how much of its drawn power is converted to heat. Most transformers have ~80% efficiency. Let's say we've got a poor space heater with only 80% efficency as well. Both draw 100 watts. In this case, the transformer turns 20 watts to heat and the space heater turns 80 watts to heat.
QED.
"I sit down and I play the game and I have a blast with it... co-op's especially bad." He means that co-op is especially good. A typical (ab)use of bad in this way is like the following example: "I like video games. I've been wasting hours on them. Videogame X is especially bad." In this context, it is clear that Videogame X is especially time wasteful and, by implication, is the most-liked of the speaker's videogames.
Wikipedia has a nice article on Spasmodic Dysphonia.
As the blog indicates, this is thought to be a neurological condition. When I was studying AI as an undergrad, we learned a lot about neural networks. This seems like the sort of thing that could happen if the brain's speech area's neurons somehow became trained to stop delivering impulses for "normal" speech. In this case, it would be theoretically possible to train the network back to normal levels. Of course, it could be something completely different.
Here's wishing Scott the best.
They aren't open for debate, they are specifically retained by the people. Read your own link. For example, the Constitution doesn't address the right to jumping. That doesn't mean that my right to jump up and down is open for debate. It means that my right to jump up and down is specifically protected from Federal laws.
Let's take the case of a 40 watt light bulb and a 40 watt space heater. The 40 watt bulb consumes electricity and produces light and heat. On the other hand, the space heater consumes electricity and produces only heat. With the same amount of consumed power, the space heater will produce more heat because none of the power is being used to produce non-heat outputs (in this case light.) This is the conservation of energy--light and heat are both energy, and the same amount of electricity can't magically produce extra energy just because it happens to be used by a light bulb.
Now let me address your exact question, because it doesn't have the benefit of the light to simplify the example. Let's say that we have a 20 watt transformer hooked onto nothing; it's just an open circuit with a transformer slapped in there. The transformer produces heat for several reasons, including the resistance it produces in the circuit. On the other hand, a space heater doesn't have nearly as many non-heat losses as a transformer. (It doesn't have magnetostriction, mechanical losses, or hysteresis losses for instance.)
I don't see how saving two hundred dollars a year by using electricity rather than propane, affects which electrical source produces heat most efficiently. An electric space heater designed for the task is a more efficient heat source than a hodge-podge of random electronic gadgets spewing waste heat. In other words, both may save you money over propane, but an electric heater will save you money over waste heat.
Your analogy is like saying that going down a hill saves gas because you can coast--it ignores that fact that you had to use more gas to climb the hill in the first place.
While your devices may produce waste heat, they don't produce it as efficiently as something designed for the purpose. A space heater would save you more money. Suggesting that devices continue to be designed in a deliberately inefficient matter strikes me as somewhat foolish.
I can think of a dozen bars within 20 minutes of me (Saint Paul, MN, USA) that have regular poker tournaments. Some have entry fees, but most do not. The prizes differ from location to location, but most have cash prizes. The bars make money from their expensive drinks, while the contestants get a fun evening of poker. I think you are not looking hard enough for a game.
It is larger than a boomerang. The videos are of a prototype. The website lists dimensions of "two to ten feet in length" and "two to six inches in height".
What was the last war the USA fought against anyone with smart weapons? My guess is that this is being marketed toward the needs of warfar against low-tech enemies using guerilla techniques. I could also anticipate use as a "look over this next hill" tool where you only need 30 seconds of flight. If a smart missile is only 30 seconds out I think you may have bigger problems.
I think there is great future in bionics. In addition to limbs as discussed in this submission, scientists have various approaches to bionic sight as well. This subject is truly fascinating. Here is a BBC article on a different project.
Interestingly and unfortunately, much advanced and successful bionics research is being done in South America because of restrictive laws in more typical countries. While I understand the need to protect patients, research for a paper I wrote two years ago indicates that the most successful scientists are pragmatically drawn away from first-class research institutes.
The problem with the commentary for me was that there was no way to fast-forward, skip, or rewind. Some parts I found tremendously boring, while others were fascinating. I only want to hear the fascinating parts. Going slower by watching physics is not something that I'm interested in. I'm good enough at FPS games that I can typically just do something a second time if I want to see it again.
I played a bit of Episode 1 with the commentary on and found it interesting, though not enough to finish the entire episode a second time. I'm looking forward to Episode 2, which looks better in every way.