Any comments?
https://medium.com/basic-income/deep-learning-is-going-to-teach-us-all-the-lesson-of-our-lives-jobs-are-for-machines-7c6442e37a49#.hwkoweo6e
might be intresting
EU fantasies then? Along with the techno-babble of Silicon Valley? Society functions because of social reasons, not because of technical reasons. Asymmetrical wealth will not result in a species of Eloi ruling over robots ... but Morlocks in mansions ruling over Eloi in slums. The Morlocks will definitely not live underground.
A fantastically productive industrial sector, where one worker can produce a billion shoes in one day ... is no help for a billion unemployed people, who can't afford to buy a pair of shoes. The society anyone lives in, will not be giving away free middle class lives to the proletariat.
Quote from: Baruch on March 16, 2016, 07:05:28 PM
EU fantasies then? Along with the techno-babble of Silicon Valley? Society functions because of social reasons, not because of technical reasons. Asymmetrical wealth will not result in a species of Eloi ruling over robots ... but Morlocks in mansions ruling over Eloi in slums. The Morlocks will definitely not live underground. A fantastically productive industrial sector, where one worker can produce a billion shoes in one day ... is no help for a billion unemployed people, who can't afford to buy a pair of shoes. The society anyone lives in, will not be giving away free middle class lives to the proletariat.
You may very well be right. Even mass, armed, violent, and revolutionary revolt becomes impossible once this reaches a certain stage. And even before then, the changes will likely be incremental enough to avoid setting off alarms until its too late. My interest here though is to see opposing viewpoints hash it out. Optimists, like Ray Kurzweil, and realists like Stephen Hawking have differing views on the role of AI. I think it will be interesting to watch both sides hash it out.
Quote from: _Xenu_ on March 16, 2016, 07:19:31 PM
You may very well be right. Even mass, armed, violent, and revolutionary revolt becomes impossible once this reaches a certain stage. And even before then, the changes will likely be incremental enough to avoid setting off alarms until its too late. My interest here though is to see opposing viewpoints hash it out. Optimists, like Ray Kurzweil, and realists like Stephen Hawking have differing views on the role of AI. I think it will be interesting to watch both sides hash it out.
This is simply hyper-Marxism updated for the iPod generation. No industrialization (aside from asymmetrical advantage for the US and its allies after WW II) produces prosperity for the multitude. And revolution only replaces one oppressor with another (Marx's end-of-history is bullshit).
I skimmed it a bit but did not read. I'm assuming it says there has been some sort of breakthrough that will pave the way for the day when all jobs are done by machines?
Quote from: Johan on March 17, 2016, 12:33:56 AM
I skimmed it a bit but did not read. I'm assuming it says there has been some sort of breakthrough that will pave the way for the day when all jobs are done by machines?
A computer can play Go ... so now we can replace all the White collar workers, who can join all the Blue collar workers under the viaduct.
And this turn of events is surprising to anyone because?
Quote from: Johan on March 17, 2016, 06:30:15 AM
And this turn of events is surprising to anyone because?
Not to me ... a computer has been able to play good chess for sometime now. Since all White collar workers only play games all day, it will be easy to replace all of them with an X-box on steroids ;-)
Some jobs I can think of would be difficult to replace simply because human hands can articulate better than a robot's. Most intricate welding processes in constructing complicated shapes where a machine hand can't easily get into come to mind. On the other hand, engineers and robotic experts have solved many such issues with positioning equipment.
Simple line assembly in the U.S. is becoming extinct because a robot can do it better, more accurately and cheaply. Mid level assembly methods are also becoming extinct because of positioning equipment and improved articulation of robotic "hands". When robotic machines can solve problems to the extent they can do decision making in prototyping or other areas that are developmental, then you have a problem; if an Engineering function can be replaced by a machine, then the purpose of an Engineering degree in industry becomes rather pointless.
Get a robot to do geological field work, for example, and the need for humans is obviated. Now you have a mass society of billions of unemployable humans. Could be a problem.
Yes, work is for robots. They're metal. Computers "thinking" for our progress as a nation is another story. Human glitches are pretty definable; early onset dementia, sociopathy, and power lust. We know why people in control make bad decisions for the rest of us. A neural network could develop a glitch that would lead us slowly into building a planetwide amusement park. Did the programmer not define humanity properly? Did the AI misinterpret humanity? Was there a ghost in the machine?
Quote from: stromboli on March 17, 2016, 08:31:38 AM
Some jobs I can think of would be difficult to replace simply because human hands can articulate better than a robot's. Most intricate welding processes in constructing complicated shapes where a machine hand can't easily get into come to mind. On the other hand, engineers and robotic experts have solved many such issues with positioning equipment.
Simple line assembly in the U.S. is becoming extinct because a robot can do it better, more accurately and cheaply. Mid level assembly methods are also becoming extinct because of positioning equipment and improved articulation of robotic "hands". When robotic machines can solve problems to the extent they can do decision making in prototyping or other areas that are developmental, then you have a problem; if an Engineering function can be replaced by a machine, then the purpose of an Engineering degree in industry becomes rather pointless.
Get a robot to do geological field work, for example, and the need for humans is obviated. Now you have a mass society of billions of unemployable humans. Could be a problem.
Cuts both ways ... if a robot can't match the quality of a human's work, then lower the expected quality until it can ;-( No win situation with the sociopaths in charge.
One of the reasons vehicles last longer and work better is robotics. A robot does repetitive work tirelessly and accurately around the clock. The only failure in robotics is still human error, either in design or application. A machine can't fail except by wear and tear or improper setup and programming. Any failure of quality would come from misapplication or human mistakes in setup, not the device.
What I'm talking about primarily is assembly or repair functions that involve the ability to reach and articulate in tight spaces, where the function required can vary somewhat. To do a specific job or set of jobs can be programmed. But if there are variables that require decision making, a robot would have to be very sophisticated and have the same tactility and delicacy as the human hand, besides brain/hand/eye coordination. That would require not only AI but the ability to completely mimic human skills.
https://www.quora.com/How-does-the-complexity-of-Go-compare-with-Chess
QuoteSomething else to consider: these games are typically played with fixed-size boards, so we can actually bound the size of the game tree and compare the number of possible states. For example, chess on a 9x9 board has a game tree complexity of 1012310123 , while go on a 19x19 board has a game tree complexity of 1036010360 . So completely solving go is a more complex task than completely solving chess.
Finally, it may be the case that one could find an optimal strategy for playing either of these games without needing to fully "solve" them. In that case, the complexity of solving the game doesn't really matter for the purpose of winning it, but really we have no way of "proving" that a strategy is truly optimal without understanding the game tree.
How many of you have read the entire article? I have to admit that Watson caught be off guard a few years ago, and deep learning really is looking alot like intuition to me. Deep Blue mastered chess in the mid 90s, but that was a matter of pure number crunching which just isn't possible with Go due to near infinite complexity.. Instead, it shows an ability to learn a task by trial and error, which could and possibly will lead to something like Skynet. This stuff is becoming less science fiction all the time.
Don't hold your breath for real AI. It doesn't exist in humans either ;-) There is a difference between emulation and simulation. With simulation, you solve the same problem, but using a different method (Watson). With emulation, you have to solve the same problem, using the same method. Metal-ware and meat-ware are not the same method. If you want to generate an emulation to play chess, you have a child, raise it for 20 years, encourage Jr to be a Chess master ... only to have him give you the middle finger ... so you go back to work creating a new child ...
The presumption is that consciousness is understood ... therefore we can make metalware do the same thing that meatware is doing. Most people are unconscious of the fact that consciousness is not understood. In fact most scientists consider the brain to be a simple stimulus/response network of vast complexity ... but they are unable to confirm that. Description is not the same as explanation. Mastering vast combinatoric problems, via pattern matching, is not the same thing as consciousness ... it is just chess on steroids.
Nay say all you like, but the ability to learn new skills with correction and use intuition is what makes humans humans. The idea of machines being able to do this both scares and fascinates me.Whether they do this exactly how we do doesn't matter, so long as it gets the same results. If you read the article, a machine recently passed a sort of Turing test in regards to image recognition.
Quote from: _Xenu_ on March 17, 2016, 07:32:58 PM
Nay say all you like, but the ability to learn new skills with correction and use intuition is what makes humans humans. The idea of machines being able to do this both scares and fascinates me.Whether they do this exactly how we do doesn't matter, so long as it gets the same results. If you read the article, a machine recently passed a sort of Turing test in regards to image recognition.
The Turing Test has always been a straw man argument. Humans learn (sort of) ... software is programmed (by humans). Now you can partly drive the software with external data ... like sensation in humans ... and you can partly five the software with internal data ... like memory in humans. Both internal and external data can be changing. But that in fact is not what learning is, in humans. That is a very simplified model, that we can use to poorly model what humans are doing ... but it is no more actual learning ... than a ship in a bottle can be sailed by sailors across the ocean. A model is not the thing itself, the plastic Panzer IV that I built as a child, is not a real Panzer IV. It is a toy for children.