Thursday, November 2, 2017

AI's failures ... kind of

If I were to take my pick, I guess I'd guess that the single most shuddering aspect of Artificial Intelligence is its cold-bloodedness -- the inability to accede to the frailties of its current 'master,' i.e. (wo)man. AI doesn't make mistakes and has no sympathy for a world that does.

So perhaps AI would be less upsetting if failure were built into its constitution. Failure, I might argue over a lot of beer, is (wo)man's greatest accomplishment. Creators like to claim a closeness to their creations, dontcha know.

But it would be tricky:

If failure were programmed into AI, that would be a success ... which is not a failure.

On the other hand, failure to program failure in would be a failure ... which is not what AI is programmed for.

Oh fuck it! The dog ate my homework!

1 comment:

  1. Recall: “A Persian Flaw”

    “Long ago Persian tribe members worked together to weave carpets. Just ones large, high end Persian carpet would take years to complete. It is said these Persians believed only ‘God’ was perfect in all aspects and to show this carpet makers would intentionally place flaws or mistakes in the carpet.”

    You present an interesting assertion: that correctly functioning robots are “cold blooded.”
    As though, somehow, functioning well is without compassion, even somehow cruel.

    And yet I doubt you’d pay full price for a robot, a machine, that intentionally malfunctions at random intervals.

    However in some distant or not so distant future when androids, emulated humans, act as teachers, aids or companions, we probably won’t want our Androids to be smug know-it-all’s, or worse, Trump-like braggers. Will we simply want emulated humility? Or will we really want artificial sentient beings who make fake mistakes to allow us to feel better about ourselves?

    You may like one type and I another type.