EPTUS Politics Mega-thread - Page 4748
| Forum Index > Closed |
Read the rules in the OP before posting, please. In order to ensure that this thread continues to meet TL standards and follows the proper guidelines, we will be enforcing the rules in the OP more strictly. Be sure to give them a re-read to refresh your memory! The vast majority of you are contributing in a healthy way, keep it up! NOTE: When providing a source, explain why you feel it is relevant and what purpose it adds to the discussion if it's not obvious. Also take note that unsubstantiated tweets/posts meant only to rekindle old arguments can result in a mod action. | ||
KwarK
United States41991 Posts
| ||
|
Mohdoo
United States15398 Posts
On August 16 2016 05:09 DarkPlasmaBall wrote: He's basically denying that 9/11 was a terrorist attack... even though he was literally the mayor of NYC at the time. Sigh. It wasn't terrorism. It was Saddam Hussein declaring war on freedom. | ||
|
mahrgell
Germany3942 Posts
On August 16 2016 05:11 RoomOfMush wrote: An AI can certainly not be written in a day. parameters can be changed in seconds. And for evaluation parameters are everything  | ||
|
RoomOfMush
1296 Posts
On August 16 2016 05:13 mahrgell wrote: parameters can be changed in seconds. And for evaluation parameters are everything You can change them in a second but you can not predict their effects accurately. Well, unless it is a very simple kind of algorithm. | ||
|
KwarK
United States41991 Posts
On August 16 2016 05:11 RoomOfMush wrote: Because thats how computers work. The computer can only do the things you teach it to do. If the human programmer does not know how to check pieces of news for proper procedure then its impossible for the AI to do it too. And the AI can not use gut feeling. You have to define a step-by-step guide of atomic actions to take to determine whether a piece of news is trustworthy, correct, relevant, etc to teach your news-AI how to work. I dont know how you could evaluate all news in such a way but if you know it then please explain it to me. You could approach it from the other end. Instead of trying to define quality journalism to a machine such that it can compare anything to the definition you created and say whether it is quality journalism you could instead feed hundreds of thousands of articles of varying quality into a machine with each one rated by a trustworthy human for accuracy, accessibility, bias, loaded language etc. The AI could then come up with its own criteria for judging the worthiness of any given piece based upon what it learned from the examples you gave it. That's how google are teaching machines to recognize a picture of a cat. Rather than define cat attributes in a way a machine can understand they're feeding it millions of cat pictures and millions more non cat pictures and seeing if it can work out what all the cat ones have in common. | ||
|
RoomOfMush
1296 Posts
On August 16 2016 05:17 KwarK wrote: You could approach it from the other end. Instead of trying to define quality journalism to a machine such that it can compare anything to the definition you created and say whether it is quality journalism you could instead feed hundreds of thousands of articles of varying quality into a machine with each one rated by a trustworthy human for accuracy, accessibility, bias, loaded language etc. The AI could then come up with its own criteria for judging the worthiness of any given piece based upon what it learned from the examples you gave it. That's how google are teaching machines to recognize a picture of a cat. Rather than define cat attributes in a way a machine can understand they're feeding it millions of cat pictures and millions more non cat pictures and seeing if it can work out what all the cat ones have in common. Not quite. The AI can not "come up with its own criteria". A computer can not come up with anything. The algorithm is always fixed, the only things that are changed are certain parameters and values based on pure statistics and heuristics. Its very easy to manipulate if you know how it works and it needs a lot of support from humans to initialize it. These AI's are not magic. Most of them are incredibly simple and boring and do very low level stuff but on a very large scale. | ||
|
GreenHorizons
United States22724 Posts
+ Show Spoiler + 1 Are you the actual transferee/buyer of the firearm(s)?s Yes No 2 Are you under indictment or information in any court for a felony, or any other crime, for which the judge could imprison you for more than one year? Yes No 3 Have you ever been convicted in any court of a felony, or any other crime, for which the judge could have you imprisoned for more than one year, even if you received a shorter sentence including probation? Yes No 4 Are you a fugitive from justice? Yes No 5 Are you an unlawful user of, or addicted to, marijuana or any depressant, stimulant, narcotic drug, or any other controlled substance? Yes No 6 Have you ever been adjudicated mentally defective (which includes a determination by a court, board, commission, or other lawful authority that you are a danger to yourself or to others or are incompetent to manage your own affairs) or have you ever been committed to a mental institution? Yes No 7 Have you been discharged from the Armed Forces under dishonorable conditions? Yes No 8 Are you subject to a court order restraining you from harassing, stalking, or threatening your child or an intimate partner or child of such partner? Yes No 9 Have you ever been convicted in any court of a misdemeanor crime of domestic violence? Yes No 10 Have you ever renounced your United States citizenship? Yes No 11 Are you an alien illegally in the United States? Yes No 12 Are you an alien admitted to the United States under a nonimmigrant visa? Yes No 13 Do you have a valid government-issued photo identification, such as a driver’s license? Yes No | ||
|
Plansix
United States60190 Posts
On August 16 2016 05:16 RoomOfMush wrote: You can change them in a second but you can not predict their effects accurately. Well, unless it is a very simple kind of algorithm. I don’t think that is going to make people trust them more, which is the root of the discussion. That this automated system built by humans cannot be free of bias or engender trust more than a human doing the same thing. Or that people won’t be more willing to trust the news feed AI more than a news network long term. | ||
|
Evotroid
Hungary176 Posts
On August 16 2016 05:11 RoomOfMush wrote: Because thats how computers work. The computer can only do the things you teach it to do. If the human programmer does not know how to check pieces of news for proper procedure then its impossible for the AI to do it too. And the AI can not use gut feeling. You have to define a step-by-step guide of atomic actions to take to determine whether a piece of news is trustworthy, correct, relevant, etc to teach your news-AI how to work. I dont know how you could evaluate all news in such a way but if you know it then please explain it to me. (...) That is not actually how computers work. There is no inherent limitation of that kind, and in fact, there are ways to show a program the desired outcome, a bunch of problems, and let it figure out how to solve the problems even if the programmer himself does not know how to solve them. (though it is not necessarily true in this case). I also think that the trust issue could be semi solved with an open source AI. Though I do think, if we ever reach the level of AI required for a working and really unbiased press, the media will be the least of our worries or achievements. | ||
|
RoomOfMush
1296 Posts
On August 16 2016 05:29 Evotroid wrote: That is not actually how computers work. There is no inherent limitation of that kind, and in fact, there are ways to show a program the desired outcome, a bunch of problems, and let it figure out how to solve the problems even if the programmer himself does not know how to solve them. (though it is not necessarily true in this case). No. How do you come up with an idea like that? Computers are very very simple tools, they perform one action at a time and the actions are incredibly simple. They can not come up with anything and they can not expand themselfs. Every piece of code in software has to come from a programmer. The software can never do anything that was not put there by a programmer. That is the fundamental idea behind all computers of today. Of course there is theoretical stuff of computers that actually use true randomness to produce results which are not deterministic (at least to our current understanding of physics) but these are not useable in real life at this point in time. All computers we have a fully deterministic and work sequentially. | ||
|
Plansix
United States60190 Posts
AI and computer software is not magically immune to human influence. They are just like books, written by humans. They have all the same problems that any other piece of writing has. The tech industry has tried to play up computers as being unbiased, cold thinking machines. Like sci-fi always shows, the creation always inherits the flaws of its creator. We have every ability to create a racist AI. In fact, that might be an AI’s default state. | ||
|
zlefin
United States7689 Posts
They most certainly can come up with things (though depending on your definition of that you could argue otherwise) And they're not very simple, they're hideously complex. or maybe you're just using a really weird and irregular definition. | ||
|
KwarK
United States41991 Posts
On August 16 2016 05:35 RoomOfMush wrote: No. How do you come up with an idea like that? Computers are very very simple tools, they perform one action at a time and the actions are incredibly simple. They can not come up with anything. You can program a computer to read a string of numbers, come up with a common denominator for those numbers and then determine whether any other given number is a product of that. This is the same principle. With sufficient examples of news and not news to work with a computer can identify what makes news news. | ||
|
Evotroid
Hungary176 Posts
On August 16 2016 05:35 RoomOfMush wrote: No. How do you come up with an idea like that? Computers are very very simple tools, they perform one action at a time and the actions are incredibly simple. They can not come up with anything and they can not expand themselfs. Every piece of code in software has to come from a programmer. The software can never do anything that was not put there by a programmer. That is the fundamental idea behind all computers of today. Of course there is theoretical stuff of computers that actually use true randomness to produce results which are not deterministic (at least to our current understanding of physics) but these are not useable in real life at this point in time. All computers we have a fully deterministic and work sequentially. I don't "come up" with stuff like that, unlike you. Being fully deterministic and working sequentially has nothing to do with the ability of self programming or the like. And to show that I am not talking out of my ass: Link to pdf from caltech, second on google after a paywalled article Also, easiest example: completely simulate a human brain with computer, do you accept that a human brain can learn on it's own? bam then a computer can as well. | ||
|
TheYango
United States47024 Posts
On August 16 2016 05:39 KwarK wrote: This is the same principle. With sufficient examples of news and not news to work with a computer can identify what makes news news. But this is limited by the need for an objective standard of what is and isn't news, and without that is still subject to bias in the training dataset. | ||
|
LegalLord
United Kingdom13775 Posts
On August 16 2016 05:37 zlefin wrote: Mush, you don't seem too familiar with how learning ais work; and the capabilities of adaptation that exist. They most certainly can come up with things (though depending on your definition of that you could argue otherwise) And they're not very simple, they're hideously complex. or maybe you're just using a really weird and irregular definition. They can learn, but only according to pre-specified criteria and pre-specified methods of learning. They don't have general intelligence and cannot, broadly speaking, do something they aren't programmed to do. | ||
|
RoomOfMush
1296 Posts
On August 16 2016 05:37 zlefin wrote: Mush, you don't seem too familiar with how learning ais work; and the capabilities of adaptation that exist. They most certainly can come up with things (though depending on your definition of that you could argue otherwise) And they're not very simple, they're hideously complex. or maybe you're just using a really weird and irregular definition. I have attended many many lectures on AI's and computer learning. I know exactly what kinds of limits these things have and their limits are very real. The thing is that the way they operate works great for many problems. The things google does for example can very easily be done by an AI. But there are many tasks for which our current AI's are no solution. On August 16 2016 05:39 KwarK wrote: You can program a computer to read a string of numbers, come up with a common denominator for those numbers and then determine whether any other given number is a product of that. This is the same principle. With sufficient examples of news and not news to work with a computer can identify what makes news news. Thats not "coming up with something", thats following a very strict definition and doing calculations which are known beforehand. The computer does not understand what the denominator is. The computer is told which bits to manipulate in which way. The outcome happens to be what we understand as a denominator. To the machine it doesnt matter. It just follows orders. On August 16 2016 05:40 Evotroid wrote: Also, easiest example: completely simulate a human brain with computer, do you accept that a human brain can learn on it's own? bam then a computer can as well. But nobody ever managed to simulate a human brain. We dont even know how human brains work. We dont know if human brains are deterministic or not. Our computers are. If brains are not then our computer can not simulate brains. | ||
|
zlefin
United States7689 Posts
On August 16 2016 05:43 LegalLord wrote: They can learn, but only according to pre-specified criteria and pre-specified methods of learning. They don't have general intelligence and cannot, broadly speaking, do something they aren't programmed to do. i'm very well aware of that; and I don't see how it contradicts my points which were objections to the specifics of what mush said. and since mush ninja'd: you're just using very different meanings of the word than what other people are using mush. | ||
|
Gorsameth
Netherlands21368 Posts
On August 16 2016 05:41 TheYango wrote: But this is limited by the need for an objective standard of what is and isn't news, and without that is still subject to bias in the training dataset. What is or is not news is not the problem with current media bias. It is much more a truth vs false and shades of grey argument. It is theoretically possible (tho perhaps not yet practically so) to have an AI news that reports events based solely on verifiable facts and without interpretation. | ||
|
KwarK
United States41991 Posts
On August 16 2016 05:41 TheYango wrote: But this is limited by the need for an objective standard of what is and isn't news, and without that is still subject to bias in the training dataset. Sure, but only because there isn't an objective standard that everyone can agree on. But even if we had that there would still be randoms insisting that the news is biased against their favourite form of reality. Even if we had an objective way of measuring music we'd still have people insisting that their music is the best. Hell, we have an objective way of measuring soccer teams, they can play each other in a game of soccer and see who wins. We still have people disagreeing about that. But for the most part most people understand what makes news news and what detracts from news. Authoritative sources and verifiability makes news news. Avoiding subjective or weighted language makes news news etc. These things can mostly be agreed upon. | ||
| ||
