Skip to main content

View Diary: Computerized Brain To Be Completed by 2020 (180 comments)

Comment Preferences

  •  I disagree (0+ / 0-)

    If you break down things we think of as distinctly human, like emotions, or humor, you find that to a large degree, the most important factor is how our bodies respond, not how the thought was formed.

    Take the infamous "Get a brain morans" photo. It's not that hard to derive a basic template for what's funny about it.

    Dude + attempt to insult intelligence + misspelled word = stupid dude => funny

    A computer could easily use this rule to generate dozens of genuinely funny examples. Where the magic happens is how our bodies respond to humor. Crinkling the corners of our eyes, contracting the muscles of our face into smiles, emitting repetitive barking sounds, etc. All of these responses serve as positive feedback.

    It wouldn't be that hard to create a few dozen humor templates and have volunteers rate computer generated evaluations about their humor content. A learning algorithm could record which jokes generated the most laughter, compare those with jokes that generated the least laughter and distill fairly reliable humor patterns.

    If a computer could consistently come up with jokes that make people laugh, then in a very real sense, it understands humor because it can distinguish between funny and not funny.

    It's not necessary to make a computer exactly mimic human brain function to produce humor, art, or music all that's required is a few simple goals and a trial and error process that records and analyzes human responses.

    Acquiring the ability to filter humor from non humor, art from non art, music from noise is half the equation, understanding. The other half, positive feedback, is a bit trickier, but I can envision some sort of meta process that evaluates how these understandings affect overall system functioning and with understanding plus feedback, you would have something functionally indistinguishable from intelligence.

    •  If you were right (0+ / 0-)

      then we wouldn't be having the difficulties we're having.

      A computer could easily use this rule to generate dozens of genuinely funny examples.

      This is what's been said for decades and decades and is why everyone keeps thinking AI is close.  But what you say is simple isn't, or it would have been done already.

      The revolution will not be televised. But it will be blogged, a lot. Probably more so than is necessary.

      by AoT on Thu Dec 06, 2012 at 11:05:08 AM PST

      [ Parent ]

Subscribe or Donate to support Daily Kos.

Click here for the mobile view of the site