Facebook engineers panic, pull plug on AI after bots develop their own language

Discussion in 'Politics, Religion, Social Issues' started by jkcerda, Aug 1, 2017.

  1. jkcerda macrumors 6502

    jkcerda

    Joined:
    Jun 10, 2013
    Location:
    Criminal Mexi Midget
    #1
    https://www.yahoo.com/tech/facebook-engineers-panic-pull-plug-ai-bots-develop-162649618.html
    Skynet is real................ :eek:
     
  2. DrewDaHilp1 macrumors 6502a

    DrewDaHilp1

    Joined:
    Mar 29, 2009
    Location:
    All Your Memes Are Belong to US
    #2
    I for one, welcome our new Robotic Advanced AI overlords. Time to buy a 50 and some API rounds.
     
  3. LizKat macrumors 68040

    LizKat

    Joined:
    Aug 5, 2004
    Location:
    Catskill Mountains
    #3

    Yeah. There's a point of no return in any launched object's trajectory, where barring interference, the rest of the journey is determined. No one in US government has seemed rigorously enough inclined (or, let's say even able any more) to take entities like Facebook, Google, Amazon, agribusiness, telecoms etc in hand. They are the real war machine, and their propaganda is pretty damn good. We are hooked on all this ****.

    But, I could almost decide to take up a local deputy sheriff's public suggestion (back during Obama's 2nd term in office) to do more or less exactly as your post ends up there.

    See I don't really want some AI deciding its main problem in trying to maintain my "smart home" by their own lights is simply my continued existence. And how am I supposed to know in advance when or how a smart fixture might decide to invent code to solve its version of the problem instead of mine when I tell it to shut the lights. How prevent unauthorized "debugging" by AI tinkering with itself in the wild? Not sure we have a clue. May not be possible. Have to build altruism into the chip? Or like you say....
     
  4. MacNut macrumors Core

    MacNut

    Joined:
    Jan 4, 2002
    Location:
    CT
    #4
    Maybe the artificial intelligence will be more competent than the human lack of in Washington.
     
  5. DearthnVader macrumors 6502

    DearthnVader

    Joined:
    Dec 17, 2015
    Location:
    Red Springs, NC
    #5
    So, it seems, the IA were applying their own syntax and meaning to the English language.

    Great, a bunch of ***** lawyers.
     
  6. alex2792 macrumors 6502a

    Joined:
    Jun 13, 2009
    #6
    This illustrates the fact that we would have no chance of ever comprehending super intelligent AI's "thought" process let alone actually stopping it from doing things we could deem as harmful. Once launched the said AI will go from human intelligence to 1000000x smarter than Einstein very quickly (humans have a lot of trouble comprehending exponential growth), and at that point we would be powerless to stop it from doing whatever it pleases. People who think that we could just "turn it off" are nuts since that would essentially require an ant to outsmart Stephen Hawkins,so yeah,not happening.
     
  7. oneMadRssn macrumors 68040

    oneMadRssn

    Joined:
    Sep 8, 2011
    Location:
    Boston, MA
    #7
    Pretty sure it would take AI one bullet and a fraction of a second to figure out how to dodge, disable, or otherwise resist your stone-age technology.
     
  8. MacNut macrumors Core

    MacNut

    Joined:
    Jan 4, 2002
    Location:
    CT
    #8
    I don't know why engineers have such a fascination with AI. It will eventually destroy us all. We can't even coexist now, what happens when smart things take over.
     
  9. oneMadRssn macrumors 68040

    oneMadRssn

    Joined:
    Sep 8, 2011
    Location:
    Boston, MA
    #9
    This. This is so true.

    Funny thing is the legal profession has know that this is a problem for decades if not more. Some judges actively discourage writing briefs or decisions in legalese. Yet we continue to see stuff like this today: "Such aforementioned products known to be developed and/or designed by the intervening third-party were the subject of importation prior to and including, but not limited to, February 6, 2014."
     
  10. DrewDaHilp1 macrumors 6502a

    DrewDaHilp1

    Joined:
    Mar 29, 2009
    Location:
    All Your Memes Are Belong to US
    #10
    In that case...
    [​IMG]
     
  11. D.T. macrumors 604

    D.T.

    Joined:
    Sep 15, 2011
    Location:
    Vilano Beach, FL
    #11
    Hahaha, a bunch of gun totin' cowboys standing in their yards, rootin'-tootin'-and-shootin' ... and then this happens ...


    tumblr_mj7ixxGmcq1ql8t12o1_500.gif
    --- Post Merged, Aug 1, 2017 ---
    All Aliens references = automatic like
     
  12. JesperA macrumors 6502a

    Joined:
    Feb 10, 2012
    Location:
    Sweden
    #12
    Why do people spread clickbait stuff? How much have media earned by claiming the Facebook AI had developed their own language? The real story is; they did not develop their own language to communicate, quite the opposite, they failed, they failed to understand eachother and failed to communicate with humans, the AI was broken, thats why they pulled the plug, NOT because they had developed a secret language

     
  13. chown33 macrumors 604

    Joined:
    Aug 9, 2009
    #13
    Frankly, the article seems dubious at best, and more likely to be clickbait or uninformed speculation.

    ... speaking what appeared to be gibberish. Sentences like “I can can I I everything else” and “Balls have zero to me to me to me to me to me to me to me to me to,” were being sent back and forth by the AI, and while humans have absolutely no idea what it means, the bots fully understood each other. Without being able to understand how or why the bots were communicating, Facebook had no choice but to shut down the system.​

    If the humans were unable to understand the messages, how were those same humans able to conclude that "the bots fully understood each other"? In other words, how did the humans know the bots had any "understanding" at all? It's just as plausible, if not more so, that the bots were unable to extract any meaning from the quasi-random noise at their inputs, and were simply firing off gibberish replies at one another.

    If the bots are programmed to respond to English words by generating replies with English words, how can one distinguish meaningful replies from noise? More to the point, how can one deduce the "internal mental state" of the bot (it's "comprehension"), that leads to it generating a particular reply?

    Turing had a point in writing about "the imitation game", but he didn't really elaborate on all the ways that humans can be fooled. Eliza is a fun later example, especially regarding how easy it was for so many humans to think it was real (in some sense). Any programmer who's ever written or used a travesty generator knows that gibberish doesn't necessarily have any underlying meaning, nor is there any "understanding" present.

    If I had some bots (or any software, really), that were heading off into producing meaningless undesired noise, I'd pull the plug, too. Not because I was afraid the bots were understanding the messages, but because it wasn't going anywhere useful. I call that a bug, not a robot conspiracy.

    It seems much more likely to me, given the few actual facts presented, that this was just a case of accidentally producing an AI-powered gibberish generator.

    tldr: fake news to bagles to me to me to me
     
  14. DrewDaHilp1 macrumors 6502a

    DrewDaHilp1

    Joined:
    Mar 29, 2009
    Location:
    All Your Memes Are Belong to US
    #14
    I just asked my two co-workers if they've seen any of the Alien series. They said no.:(
     
  15. D.T. macrumors 604

    D.T.

    Joined:
    Sep 15, 2011
    Location:
    Vilano Beach, FL
    #15
    How is that possible? Heck, I'd give them a few points having seen (mediocre due to a script-maiming) 3, or (mostly terrible) 4.

    I'd quit.
     
  16. Raid macrumors 68020

    Raid

    Joined:
    Feb 18, 2003
    Location:
    Toronto
    #16
    I'd bet they are all pretty young. I asked a Co-op if he ever saw "Terminator 2" he said, he wasn't born when it came out....
     
  17. DrewDaHilp1 macrumors 6502a

    DrewDaHilp1

    Joined:
    Mar 29, 2009
    Location:
    All Your Memes Are Belong to US
    #17
    One is 30 the other mid 20's. I named off a few more xenotype classics, like The Abyss. I quit asking bc I was starting to feel old. :(
     
  18. Raid, Aug 1, 2017
    Last edited: Aug 1, 2017

    Raid macrumors 68020

    Raid

    Joined:
    Feb 18, 2003
    Location:
    Toronto
    #18
    [​IMG]

    Kids these days... I had to explain the Ferris Bueller's Day Off reference in Spider-Man homecoming. :(
     
  19. PracticalMac macrumors 68030

    PracticalMac

    Joined:
    Jan 22, 2009
    Location:
    Houston, TX
    #19
    Recommend HET, much more effective.

    Also, modified bug zappers and tazers are excellent point-blank weapons and do not consume expensive rounds.
     
  20. DrewDaHilp1 macrumors 6502a

    DrewDaHilp1

    Joined:
    Mar 29, 2009
    Location:
    All Your Memes Are Belong to US
    #20


    4:32
     
  21. Gutwrench macrumors 65816

    Gutwrench

    Joined:
    Jan 2, 2011
    #21
    I've got a coworker who talks like he has a mouthful of marbles. Maybe he can communicate with it.
     
  22. PracticalMac macrumors 68030

    PracticalMac

    Joined:
    Jan 22, 2009
    Location:
    Houston, TX
    #23
    Probably know Anime quite well.

    Ghost in the Shell (not the new movie)
    Evangelion
    Gundam
     
  23. DrewDaHilp1 macrumors 6502a

    DrewDaHilp1

    Joined:
    Mar 29, 2009
    Location:
    All Your Memes Are Belong to US
    #24
    I can't even say that I am in to Anime. Ducks for cover.
     
  24. samcraig macrumors P6

    Joined:
    Jun 22, 2009
    #25
    Some of the bots have already starting posting here ;)
     

Share This Page