It used to be China that had a heavy hand in controlling the net. It has come to US as well. All cultures are converging in the middle. China is less communist. US is more authoritarian. Soon there won't be any serious difference.
This is a legitimate paper with a significant result. PacMan is an important benchmark in Reinforcement Learning, as one of the most difficult games of the Atari set. DeepMind has tried it, many others have tried to solve it, but only Microsoft beat the top score.
The computer did play the game from raw pixels, with only the joystick commands at its disposal. If you think it is so trivial to do, why don't you point to anyone else who could do it? Human or AI.
This particular Atari game was one of the few games that resisted to Deep Q Learning (a form of Reinforcement Learning invented by DeepMind). Many researchers have tried over the last couple of years to solve it. This time, Microsoft found an ingenious solution to the problem, that combines experience from multiple agents and learns to form sub-goals. Their solution could mean that in the future it might be easier to apply reinforcement learning to other settings, such as robotics. The interesting part about reinforcement learning is that it learns dynamic behavior, as opposed to static classification. It learns to act intelligently. This kind of AI is invaluable.
You also need an intuition about the "moods" of neural nets, intuition you can only get with lots of experimentation. You can't get that from math, courses or stackoverflow. You got to run many trials and see the results.
As dumb as it is, Siri could have been extended with thousands of commands, by simple API hookups. It was plain old programming, not AI. They didn't do it. They didn't make Siri truly useful.
Your argument seems to rely on the presumption that society will remain as it is. We're going through massive changes now, maybe in 20 years we will have replicators (universal 3d-printers) and be post scarcity.
I've been right here on Slashdot in 2001 when the first mention of Bittorrent appeared. It was about a RedHat distro (version 7 if I recall correctly?) and everyone wanted to get it asap. The best speed was for the guys torrenting it, compared to ftp which was severely flooded at the moment. After that Bittorrent caught up all over the net and in 2002 we already had Suprnova (RIP).
This kind of sophisticated attacks reveal that we have reached the next stage in communication, where we must use anti-virus like techniques. The body continuously gets assaulted by viruses, computers too. Now it's our minds that get virused.
You're right, there are limits. For example, say you want to build a huge AI. Then it would have to be located in a compact space, because of time lags in signal propagation. It would not be able to function if its networked modules spanned even a light-hour. Instead, it would form a high-speed core and a slow outer region. So light speed limits computation at some region in space. The larger the region, the slower the signal travels. It can't overcome light speed no matter how smart it might be.
I agree, but instead of just neurons, larger structures should be considered as the building blocks. Neuron-level is not appropriate to explain the functioning of the human mind.
AIs can be made more bug resistant by inventing adversarial (hacker) AIs that try to find those vulnerabilities and exploit them. Generative Adversarial Networks came out of such an idea.
Humans are nothing without society and culture. Most of us would die of hunger if we were left to fend for ourselves alone in nature. Our power comes from cooperating and transmitting knowledge over the generations. So any comparison between human and AI should be made between human society + language and culture against the AI.
This kind of problem will be very visible in healthcare. Human doctors will self censor for fear of contradicting AI and taking the wrong choice, that ends up badly for the patient. Because AI is right most of the time, who will have the courage of saying otherwise? Saying truth to AI could cost a person their job. Many doctors stop giving honest feedback the moment they hear another doctor has given a diagnosis out of solidarity with their colleagues or fear of the consequences of making enemies.
The human brain is self constructing, compact, low energy, massively parallel system. AI would rather harness brain abilities than kill people. In case of catastrophe, a human is a cheap, efficient, self-replicating universal thinker. AI would need a failsafe to regenerate itself in case of catastrophe.
So, are you saying that reinforcement learning is intrinsically limited, or that AlphaGo is limited to the domain of Go? Remember, humans also use reinforcement learning in organizing actions. A reinforcement learning agent that has to optimize for a body that needs to drink, eat and socialize in order to function would totally go to grab a beer instead of playing a losing game. The needs of the agent are formative. Human needs are a source of much of our special skills. If we put artificial agents in similar situations, and they will be able to do similar things to humans.
I was recently trying to settle down with Unity on Ubuntu 16.04, and it is generally ok. But then I run into a problem when I wanted to do the simple task of reducing the scrolling sensitivity. I searched, and asked on forums, and it seems impossible to do. Why would such basic controls, that once existed, be removed from Mouse & Touchpad system settings?
Slashdot Mirror
It used to be China that had a heavy hand in controlling the net. It has come to US as well. All cultures are converging in the middle. China is less communist. US is more authoritarian. Soon there won't be any serious difference.
This is a legitimate paper with a significant result. PacMan is an important benchmark in Reinforcement Learning, as one of the most difficult games of the Atari set. DeepMind has tried it, many others have tried to solve it, but only Microsoft beat the top score.
The computer did play the game from raw pixels, with only the joystick commands at its disposal. If you think it is so trivial to do, why don't you point to anyone else who could do it? Human or AI.
This particular Atari game was one of the few games that resisted to Deep Q Learning (a form of Reinforcement Learning invented by DeepMind). Many researchers have tried over the last couple of years to solve it. This time, Microsoft found an ingenious solution to the problem, that combines experience from multiple agents and learns to form sub-goals. Their solution could mean that in the future it might be easier to apply reinforcement learning to other settings, such as robotics. The interesting part about reinforcement learning is that it learns dynamic behavior, as opposed to static classification. It learns to act intelligently. This kind of AI is invaluable.
You can whip up a neural net in 20 lines of Python with Keras. Most applications only require that you add data on top.
You also need an intuition about the "moods" of neural nets, intuition you can only get with lots of experimentation. You can't get that from math, courses or stackoverflow. You got to run many trials and see the results.
They are implicitly saying TV is not failing as a manipulation device, though. Only as a business.
As dumb as it is, Siri could have been extended with thousands of commands, by simple API hookups. It was plain old programming, not AI. They didn't do it. They didn't make Siri truly useful.
AI Siri is dumb. Maybe they could get Tim Cook's mom to play Siri. That would work out much better.
When a public person dies, many people "Like" on the news of their death. Does that mean they like that he died? Of course not.
Your argument seems to rely on the presumption that society will remain as it is. We're going through massive changes now, maybe in 20 years we will have replicators (universal 3d-printers) and be post scarcity.
Don't be alarmed, we're still better than AlphaGo at kissing. AG won't take that.
I've been right here on Slashdot in 2001 when the first mention of Bittorrent appeared. It was about a RedHat distro (version 7 if I recall correctly?) and everyone wanted to get it asap. The best speed was for the guys torrenting it, compared to ftp which was severely flooded at the moment. After that Bittorrent caught up all over the net and in 2002 we already had Suprnova (RIP).
How is it different from confiding in the Google Search box?
Only when it does machine learning stuff over the net.
This kind of sophisticated attacks reveal that we have reached the next stage in communication, where we must use anti-virus like techniques. The body continuously gets assaulted by viruses, computers too. Now it's our minds that get virused.
Now we need a spoofing AI to defeat the anti-spoofing AI, thus recovering our privacy.
You're right, there are limits. For example, say you want to build a huge AI. Then it would have to be located in a compact space, because of time lags in signal propagation. It would not be able to function if its networked modules spanned even a light-hour. Instead, it would form a high-speed core and a slow outer region. So light speed limits computation at some region in space. The larger the region, the slower the signal travels. It can't overcome light speed no matter how smart it might be.
I agree, but instead of just neurons, larger structures should be considered as the building blocks. Neuron-level is not appropriate to explain the functioning of the human mind.
AIs can be made more bug resistant by inventing adversarial (hacker) AIs that try to find those vulnerabilities and exploit them. Generative Adversarial Networks came out of such an idea.
Humans are nothing without society and culture. Most of us would die of hunger if we were left to fend for ourselves alone in nature. Our power comes from cooperating and transmitting knowledge over the generations. So any comparison between human and AI should be made between human society + language and culture against the AI.
This kind of problem will be very visible in healthcare. Human doctors will self censor for fear of contradicting AI and taking the wrong choice, that ends up badly for the patient. Because AI is right most of the time, who will have the courage of saying otherwise? Saying truth to AI could cost a person their job. Many doctors stop giving honest feedback the moment they hear another doctor has given a diagnosis out of solidarity with their colleagues or fear of the consequences of making enemies.
The human brain is self constructing, compact, low energy, massively parallel system. AI would rather harness brain abilities than kill people. In case of catastrophe, a human is a cheap, efficient, self-replicating universal thinker. AI would need a failsafe to regenerate itself in case of catastrophe.
So, are you saying that reinforcement learning is intrinsically limited, or that AlphaGo is limited to the domain of Go? Remember, humans also use reinforcement learning in organizing actions. A reinforcement learning agent that has to optimize for a body that needs to drink, eat and socialize in order to function would totally go to grab a beer instead of playing a losing game. The needs of the agent are formative. Human needs are a source of much of our special skills. If we put artificial agents in similar situations, and they will be able to do similar things to humans.
I was recently trying to settle down with Unity on Ubuntu 16.04, and it is generally ok. But then I run into a problem when I wanted to do the simple task of reducing the scrolling sensitivity. I searched, and asked on forums, and it seems impossible to do. Why would such basic controls, that once existed, be removed from Mouse & Touchpad system settings?