2026-09-02
Well, it's not always because it has decided that the humans are getting it all wrong and need to be removed.
Sometimes it's because of faulty programming.
But what is faulty programming? Maybe the program is running precisely according to specification, but still produces the wrong result.
How can this be? Usually (inevitably?) a faulty (and probably unstated) assumption upon which the program specification relied.
And who wrote that specification? A human being, imperfect ,and liable to unwitting errors of omission.
Or maybe the error was in the AI's training materials, which were (inevitably) sourced ultimately from human-written materials (even if also subsequently reprocessed by AI into relevant but still erroneous training materials).
Also subject to errors of all kinds.
