Great debating technique. I can tell this is going to be good.
It absolutely does self-correct. When it encounters data that doesn't match its model it adjusts the model. If the AI is biased to say that a certain sex is more likely to have a certain trait, then if it encounters data that says otherwise the model is adjusted.
That's not correction. In order to self correct, it has to recognize that the output is wrong. You are talking about adding another data point to its statistical model.
You seem to think that the algorithm is beyond reproach here, but there are many obvious ways for it to be less than great. How does it handle historical data, is there some cut off age or is older data weighted differently, or does it just consider cases from the 1817 as valid as ones from the 2017? How is the data verified for accuracy and how does it integrate corrections? How is each data point weighted and what checks are done to ensure that the weighting is fair?
The AI can't determine causation vs. correlation, as you pointed out. Because the AI can't determine anything. It's all statistical. So when your real-world dataset exceeds the scope of your test dataset (or the human-driven classification of it), you have two choices: Accept the output and hope your AI is correct, or reject the output, retrain the AI, and hope you were correct in rejecting the output.
You call me an idiot, and then agree with me. Correlated data suggests you are an idiot too.
To whoever modded this "flamebait", you are part of the cancer of troll moderation that is killing Slashdot. Your ideological modding is polarizing this place.
Humans may infer causation, but that's not the fault of AI.
Correct, it's the fault of the developers. This isn't really AI, it's not a neural net, it's an algorithm designed by humans.
At the moment there seems to be little oversight of the design process or willingness to handle problems with the output, and that's the issue. It's the classic "computer says no", only instead of being denied a loan you get to spend an extra 5 years in jail.
In Europe even access to public data is regulated. For example, a credit reference agency can't bypass limits on reporting things like bankruptcy by pointing to publicly available news reports. Banks can't use such information, even if they have it, to make decisions (although it can be hard to prove).
There is a great incentive not to look at employees' social media profiles because if discovered (browser history, social media companies reporting it to users etc) it could be extremely bad for them.
What do those starts have to do with sentencing? Surely the sentence should be based on the nature of the crime and past behaviour, not income it race.
The data is incomplete. AI, like humans, makes mistakes like "correlation = causation". The problem is, like some humans, AI doesn't understand this and can't ask for additional information or self-correct.
My point was that you can't really relax and read the paper in traffic if you have to be constantly ready to throw it aside and grab the wheel, making sense of your situation to avoid an imminent crash.
I'm looking at the link but I can't see where it says "let's kill her instead", except for in the actual tweet you linked to. Who is supposed to have said that?
All I see is someone pointing out that people do get killed and their killers get away with it because of "trans panic". That's actually a thing, there are numerous cases of it happening and the perpetrators using the "trans panic" defence successfully in court. Even if they went the jail, the victim is still dead/beaten.
Laci Green is an interesting example. As you have noticed, the mainstream media doesn't cover random YouTube bloggers much, regardless of political affiliation. What is noteworthy is how progressive YouTubers have tried to be conciliatory and reasonable with her. The screenshot Ian Miles Cheong is actually an example of that - someone calmly explaining the problem to her.
By the way, what does this have to do with gamergate?
The continued population growth is because people are living longer, but it's levelling off. We are on target for about 10-11bn by the end of the century, which is sustainable with modern farming methods. The main issues now are all to do with the politics of handling the increase.
I tend to think of code in different ways depending on what it does and what my debugging requirements are.
Sometimes I think of it in terms of the assembler code it will produce. Sometimes I think of it in terms of the way it will modify binary data. Sometimes I think of it in mathematical terms. Sometimes I think about how the compiler will try to optimize it.
Mostly when I look at C code, my brain is working as a C code parser. C#/Java is a bit higher level so more in terms of relationships and OO, but still with a bit of parsing going on.
Something doesn't make sense here. If this person was overpowering an outside broadcast unit, why didn't the station just switch to some other signal (like the presenter in their offices) instead of playing that song?
I'm actually surprised that Amazon doesn't give away their hardware... Maybe they are happy with a slower roll out and some extra profit.
Eventually I expect they will be free, because they make so much money for Amazon. Not just extra purchases, but they also get to direct your purchases and control who else gets to make sales. If you say "order some AA batteries", they get to decide which brand of battery and how many and who from.
relying on the car to tell the driver when to pay attention take over
I wonder how long you have to take over. If you are reading a newspaper as they suggest, it might take you several seconds to finish the paragraph, fold it up and store it safely before taking in your environment and figuring out what is required of you.
The main problem I have with the Tesla system is that it suddenly starts beeping urgently at you, about 0.5 seconds before it hits something. If you were not paying full attention with hands on the wheel already, you are probably screwed.
They seem to have rules enforcing privacy in Star Trek. For example, when they ask the computer where someone is, it actually tells them where their comm badge is. If they take the badge off, they computer appears unable to locate them, even though the ship's sensors can apparently detect individuals on a planet full of other living creatures from orbit. There must be some kind of hard lock-out in place as they never even think to try to override it.
Presumably the computer would be programmed to ignore and never record anything it hears until it is address directly by name, which is why they always prefix any question or order for it with "Computer."
Talk about redefining harassment. You are claiming that saying "gamergate was about harassment" is somehow harassing everyone involved with gamergate.
All the ranting you do about "criticism is not harassment" and people needing to have thicker skins, yet when someone claims gamergate isn't about ethics at all, all that goes out the window and everyone's a victim of a sustained hate campaign.
In the UK it would be illegal to read a newspaper while "in control of a vehicle" (i.e. sat in the driver's seat with it in motion). In fact the current requirement for antonymous vehicles is to keep your hands on the wheel and pretend to be driving so as not to alarm other road users.
The law needs to change. I wonder what kind of liability Audi has here - presumably if there was an accident the insurance company would sue Audi on the driver's behalf if they had legal cover.
Oh, so you are saying I should have added other non-harassment items to my list?
No. If you could get past your assumption that everyone who disagrees with you is an "SJW" you might see a more reasonable interpretation, e.g. that the study is looking at a whole range of behaviours, some of which are quite serious and might even be illegal. Focusing on name calling and assuming that the report is trying to define harassment (it isn't) is silly.
Conflating terms is marxist tool
LOL, but yes, conflating a survey on frequency of incidents with an attempt to define harassment is deliberate misrepresentation, a common tactic in internet debates.
Slashdot Mirror
You're an idiot.
Great debating technique. I can tell this is going to be good.
It absolutely does self-correct. When it encounters data that doesn't match its model it adjusts the model. If the AI is biased to say that a certain sex is more likely to have a certain trait, then if it encounters data that says otherwise the model is adjusted.
That's not correction. In order to self correct, it has to recognize that the output is wrong. You are talking about adding another data point to its statistical model.
You seem to think that the algorithm is beyond reproach here, but there are many obvious ways for it to be less than great. How does it handle historical data, is there some cut off age or is older data weighted differently, or does it just consider cases from the 1817 as valid as ones from the 2017? How is the data verified for accuracy and how does it integrate corrections? How is each data point weighted and what checks are done to ensure that the weighting is fair?
The AI can't determine causation vs. correlation, as you pointed out. Because the AI can't determine anything. It's all statistical. So when your real-world dataset exceeds the scope of your test dataset (or the human-driven classification of it), you have two choices: Accept the output and hope your AI is correct, or reject the output, retrain the AI, and hope you were correct in rejecting the output.
You call me an idiot, and then agree with me. Correlated data suggests you are an idiot too.
To whoever modded this "flamebait", you are part of the cancer of troll moderation that is killing Slashdot. Your ideological modding is polarizing this place.
Humans may infer causation, but that's not the fault of AI.
Correct, it's the fault of the developers. This isn't really AI, it's not a neural net, it's an algorithm designed by humans.
At the moment there seems to be little oversight of the design process or willingness to handle problems with the output, and that's the issue. It's the classic "computer says no", only instead of being denied a loan you get to spend an extra 5 years in jail.
In Europe even access to public data is regulated. For example, a credit reference agency can't bypass limits on reporting things like bankruptcy by pointing to publicly available news reports. Banks can't use such information, even if they have it, to make decisions (although it can be hard to prove).
There is a great incentive not to look at employees' social media profiles because if discovered (browser history, social media companies reporting it to users etc) it could be extremely bad for them.
It's not any racial group doing it, black or white. It's institutional, for the most part.
The logical conclusion of your argument is that we shouldn't teach kids right and wrong because it's just SJW politically correct bullshit.
The problem is that AI needs to learn to ignore those biases like a good human does, not that there may be some statistical validity to them.
What do those starts have to do with sentencing? Surely the sentence should be based on the nature of the crime and past behaviour, not income it race.
That's a very bizarre interpretation of "give everyone a fair opportunity to obtain justice".
The data is incomplete. AI, like humans, makes mistakes like "correlation = causation". The problem is, like some humans, AI doesn't understand this and can't ask for additional information or self-correct.
Good point, I failed to account for the cheapness of the station adequately.
My point was that you can't really relax and read the paper in traffic if you have to be constantly ready to throw it aside and grab the wheel, making sense of your situation to avoid an imminent crash.
I'm looking at the link but I can't see where it says "let's kill her instead", except for in the actual tweet you linked to. Who is supposed to have said that?
All I see is someone pointing out that people do get killed and their killers get away with it because of "trans panic". That's actually a thing, there are numerous cases of it happening and the perpetrators using the "trans panic" defence successfully in court. Even if they went the jail, the victim is still dead/beaten.
Laci Green is an interesting example. As you have noticed, the mainstream media doesn't cover random YouTube bloggers much, regardless of political affiliation. What is noteworthy is how progressive YouTubers have tried to be conciliatory and reasonable with her. The screenshot Ian Miles Cheong is actually an example of that - someone calmly explaining the problem to her.
By the way, what does this have to do with gamergate?
Until we stop breeding as rabbits
We already did: http://data.worldbank.org/indi...
The continued population growth is because people are living longer, but it's levelling off. We are on target for about 10-11bn by the end of the century, which is sustainable with modern farming methods. The main issues now are all to do with the politics of handling the increase.
I prefer to use appropriate voices for comedic effect, having grown up with What the Papers Say.
I tend to think of code in different ways depending on what it does and what my debugging requirements are.
Sometimes I think of it in terms of the assembler code it will produce. Sometimes I think of it in terms of the way it will modify binary data. Sometimes I think of it in mathematical terms. Sometimes I think about how the compiler will try to optimize it.
Mostly when I look at C code, my brain is working as a C code parser. C#/Java is a bit higher level so more in terms of relationships and OO, but still with a bit of parsing going on.
Something doesn't make sense here. If this person was overpowering an outside broadcast unit, why didn't the station just switch to some other signal (like the presenter in their offices) instead of playing that song?
I'm actually surprised that Amazon doesn't give away their hardware... Maybe they are happy with a slower roll out and some extra profit.
Eventually I expect they will be free, because they make so much money for Amazon. Not just extra purchases, but they also get to direct your purchases and control who else gets to make sales. If you say "order some AA batteries", they get to decide which brand of battery and how many and who from.
relying on the car to tell the driver when to pay attention take over
I wonder how long you have to take over. If you are reading a newspaper as they suggest, it might take you several seconds to finish the paragraph, fold it up and store it safely before taking in your environment and figuring out what is required of you.
The main problem I have with the Tesla system is that it suddenly starts beeping urgently at you, about 0.5 seconds before it hits something. If you were not paying full attention with hands on the wheel already, you are probably screwed.
They seem to have rules enforcing privacy in Star Trek. For example, when they ask the computer where someone is, it actually tells them where their comm badge is. If they take the badge off, they computer appears unable to locate them, even though the ship's sensors can apparently detect individuals on a planet full of other living creatures from orbit. There must be some kind of hard lock-out in place as they never even think to try to override it.
Presumably the computer would be programmed to ignore and never record anything it hears until it is address directly by name, which is why they always prefix any question or order for it with "Computer."
Talk about redefining harassment. You are claiming that saying "gamergate was about harassment" is somehow harassing everyone involved with gamergate.
All the ranting you do about "criticism is not harassment" and people needing to have thicker skins, yet when someone claims gamergate isn't about ethics at all, all that goes out the window and everyone's a victim of a sustained hate campaign.
In the UK it would be illegal to read a newspaper while "in control of a vehicle" (i.e. sat in the driver's seat with it in motion). In fact the current requirement for antonymous vehicles is to keep your hands on the wheel and pretend to be driving so as not to alarm other road users.
The law needs to change. I wonder what kind of liability Audi has here - presumably if there was an accident the insurance company would sue Audi on the driver's behalf if they had legal cover.
Oh, so you are saying I should have added other non-harassment items to my list?
No. If you could get past your assumption that everyone who disagrees with you is an "SJW" you might see a more reasonable interpretation, e.g. that the study is looking at a whole range of behaviours, some of which are quite serious and might even be illegal. Focusing on name calling and assuming that the report is trying to define harassment (it isn't) is silly.
Conflating terms is marxist tool
LOL, but yes, conflating a survey on frequency of incidents with an attempt to define harassment is deliberate misrepresentation, a common tactic in internet debates.
Who is manufacturing victimhood now?
You know your argument is weak when you have to use Breitbart and YouTube videos as your sources.