Page 2 of 8 First 123456 ... Last
  1. #11
    submessenger's Avatar
    Join Date
    Apr 2010
    Location
    KAUS
    Posts
    5,351
    Style
    BJJ/Judo/MT
    1
    This discussion has much broader implications than just 2A, as I'm sure you'll agree. When/if we can directly relate AI applications to the gun control debate, I'm sure Cassius will welcome them. In the interim, we can continue the discussion, here. Until then, it's just white noise to them, geeks talking about computers. Doesn't stop us talking, but it may serve to elevate the level of discussion.

    Now, if you'll excuse me for a bit, I'm having an argument with Watson, right now. Quite literally, the fucker is choosing to forget several facts when I'm telling it to add one new fact to the mix. Annoying as all hell.

  2. #12
    submessenger's Avatar
    Join Date
    Apr 2010
    Location
    KAUS
    Posts
    5,351
    Style
    BJJ/Judo/MT
    1
    So, here's how you fix it. You set up some temporary memory space. You record the facts into that temporary memory. You tell Big Blue your new fact. And, then you remind it of the other facts from temporary memory. WTF?

    Back in a few to continue the discussion.

  3. #13
    W. Rabbit's Avatar
    Join Date
    May 2010
    Location
    The Vale of Shadows
    Posts
    12,079
    Style
    Mixed
    Quote Originally Posted by submessenger View Post
    This discussion has much broader implications than just 2A, as I'm sure you'll agree. When/if we can directly relate AI applications to the gun control debate, I'm sure Cassius will welcome them. In the interim, we can continue the discussion, here. Until then, it's just white noise to them, geeks talking about computers. Doesn't stop us talking, but it may serve to elevate the level of discussion.

    Now, if you'll excuse me for a bit, I'm having an argument with Watson, right now. Quite literally, the fucker is choosing to forget several facts when I'm telling it to add one new fact to the mix. Annoying as all hell.
    Facts can change. They aren't immutable. Like your "fact" about neural networks needing to be trained by humans, which turned out to be false. It flipped over time. We could graph it as a step function.



    In fact, this ties right back to my point about stochastic signal theory and "random" mass shootings. Random signal theory is about the chaos of events and/or determinism in the time domain.

    However, I'm willing to posit that there's always a signal in the noise before these events, that they aren't completely non-deterministic, and that humans are just terrible at finding the signal in what they perceive to be "noise" but are in fact, filterable. There is advanced discrete SIG/SIGINT math that describes this exact sort of thing (finding a tiny, tiny signal in a vast ocean of random signals).
    Last edited by W. Rabbit; 11/09/2017 11:21pm at .

  4. #14
    W. Rabbit's Avatar
    Join Date
    May 2010
    Location
    The Vale of Shadows
    Posts
    12,079
    Style
    Mixed
    It's also very clear that the rate of attacks is increasing, unless you believe the mass shooting blip of 1999-2017 will end up being an anomaly of mass shootings. It's likely just the start of it.

    Laws and systems like NICS don't seem to solve the problem well enough, either. Why can't machines do a better job again? Humans make mistakes, we can't afford mistakes here.

  5. #15
    submessenger's Avatar
    Join Date
    Apr 2010
    Location
    KAUS
    Posts
    5,351
    Style
    BJJ/Judo/MT
    Quote Originally Posted by W. Rabbit View Post
    Facts can change. They aren't immutable. Like your "fact" about neural networks needing to be trained by humans, which turned out to be false. It flipped over time. We could graph it as a step function.



    In fact, this ties right back to my point about stochastic signal theory and "random" mass shootings. Random signal theory is about the chaos of events and/or determinism in the time domain.

    However, I'm willing to posit that there's always a signal in the noise before these events, that they aren't completely non-deterministic, and that humans are just terrible at finding the signal in what they perceive to be "noise" but are in fact, filterable. There is advanced discrete SIG/SIGINT math that describes this exact sort of thing (finding a tiny, tiny signal in a vast ocean of random signals).
    Super-secret Google **** is not available to the masses. I stand by my "nets need training," because that remains the status quo, until you can get me a TPU or 20.

    (edit) even if you did get me one tonight, there would still be an implementation delay, even if I threw every resource I had at the thing.
    (edit moar) the AGZ didn't just decide on its own to learn to play Go. AI remains dependent on meat.

  6. #16
    W. Rabbit's Avatar
    Join Date
    May 2010
    Location
    The Vale of Shadows
    Posts
    12,079
    Style
    Mixed
    Quote Originally Posted by submessenger View Post
    Super-secret Google **** is not available to the masses. I stand by my "nets need training," because that remains the status quo, until you can get me a TPU or 20.

    (edit) even if you did get me one tonight, there would still be an implementation delay, even if I threw every resource I had at the thing.
    (edit moar) the AGZ didn't just decide on its own to learn to play Go. AI remains dependent on meat.
    If you disagree with the findings published in Nature, publish your counterthesis. Either wise you can't determine the status quo.

    You're missing the point. The meat is everywhere. You don't need a human to tell a computer how to find it, you just need to give it access to the the game board. The machine will figure things out it all by itself.

  7. #17
    submessenger's Avatar
    Join Date
    Apr 2010
    Location
    KAUS
    Posts
    5,351
    Style
    BJJ/Judo/MT
    Quote Originally Posted by W. Rabbit View Post
    If you disagree with the findings published in Nature, publish your counterthesis. Either wise you can't determine the status quo.

    You're missing the point. The meat is everywhere. You don't need a human to tell a computer how to find it, you just need to give it access to the the game board. The machine will figure things out it all by itself.
    I'm not disputing the findings. I'm saying neither you nor I are in a position to even evaluate them, because you can't just pick up a TPU at Target on your way home from work. Repeatability is still a key tenet of the scientific method, no?

  8. #18
    W. Rabbit's Avatar
    Join Date
    May 2010
    Location
    The Vale of Shadows
    Posts
    12,079
    Style
    Mixed
    All AGZ needed was the rules:

    The neural network initially knew nothing about Go beyond the rules. The AI engaged in "unsupervised learning", playing against itself until it could anticipate its own moves and how those moves would affect the game's outcome.[6] In the first three days AlphaGo Zero played 4.9 million games against itself in quick succession.[7] It appeared to develop the skills required to beat top humans within just a few days, whereas the earlier AlphaGo took months of training to achieve the same level
    Do you think the rules of finding a potential mass killer are complex? I'll bet within one day, you could build a rule set for Kelley. It's just intrusion detection, human patterns can easily be learned by machines, or people. But computers do it faster, and don't make the kinds of mistakes humans do, like not reporting things to NICS, which is a (literally) dumb system.
    Last edited by W. Rabbit; 11/09/2017 11:59pm at .

  9. #19
    submessenger's Avatar
    Join Date
    Apr 2010
    Location
    KAUS
    Posts
    5,351
    Style
    BJJ/Judo/MT
    1
    Quote Originally Posted by W. Rabbit View Post
    All AGZ needed was the rules:



    Do you think the rules of finding a potential mass killer are complex? I'll bet within one day, you could build a rule set for Kelley. It's just intrusion detection, human patterns can easily be learned by machines, or people.
    All super-fucking awesome, but you're moving the goalposts, here. Can you let loose with this marvelous AI tomorrow?

    So, it sounds fucking brilliant, but just because it's published in a peer-reviewed almanac doesn't mean it's been peer-reviewed.

    (edit) the shortest version of what I'm trying to say is you're not going to stop any would-be mass murders with this tech in the near future

  10. #20
    W. Rabbit's Avatar
    Join Date
    May 2010
    Location
    The Vale of Shadows
    Posts
    12,079
    Style
    Mixed
    Quote Originally Posted by submessenger View Post
    I'm not disputing the findings. I'm saying neither you nor I are in a position to even evaluate them, because you can't just pick up a TPU at Target on your way home from work. Repeatability is still a key tenet of the scientific method, no?
    We don't need to evaluate them, I'm going by industry consensus on this one. Of course this is groundbreaking research, but it's pretty clear that if Go (a human game) can be conquered this way, many/any human game can be conquered this way.

    Remember, this thread is about hunting predators. Hunting is just a game, you know that. If a human can hunt, a computer can hunt, and probably do a much better job.

    AlphaGo Zero was widely regarded as a significant advance, even when compared with its groundbreaking predecessor, AlphaGo. Oren Etzioni of the Allen Institute for Artificial Intelligence called AlphaGo Zero "a very impressive technical result" in "both their ability to do it—and their ability to train the system in 40 days, on four TPUs".[6] The Guardian called it a "major breakthrough for artificial intelligence", citing Eleni Vasilaki of Sheffield University and Tom Mitchell of Carnegie Mellon University, who called it an impressive feat and an “outstanding engineering accomplishment" respectively.[11] Mark Pesce of the University of Sydney called AlphaGo Zero "a big technological advance" taking us into "undiscovered territory".[12]

Page 2 of 8 First 123456 ... Last

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •  

Log in

Log in