Also more than have died from UFAI. Clearly that’s not worth worrying over either.
I’m not terrified of Ebola because it’s been demonstrated to be controllable in fairly developed counties, but as a general rule this quote seems incredibly out of place on less wrong. People here discuss the dangers of things which have literally never happened before almost every day.
How sure are you about that? Everyone on Air France flight 447 died as a direct result of insufficiently-friendly AI. Patient zero for ebola in the US was discharged from the emergency room with some painkillers and antibiotics because software looking through his medical records failed to recognize, and call the triage nurse’s attention to, obvious risk factors.
I’m saying the control systems on the plane constitute an artificial intelligence, advanced enough to do almost all the work of flying the plane but not general enough to do anything unrelated to flying the plane, and (at issue) not friendly enough to tell the pilots “this is likely just a little ice clog, in which case you don’t need to do anything” or, later in the catastrophe, “the single most important part of stall recovery is bringing the plane’s nose back down,” let alone override their panicky incompetence outright and act on such simple suggestions itself.
Also more than have died from UFAI. Clearly that’s not worth worrying over either.
I’m not terrified of Ebola because it’s been demonstrated to be controllable in fairly developed counties, but as a general rule this quote seems incredibly out of place on less wrong. People here discuss the dangers of things which have literally never happened before almost every day.
And for all the scaremongering over so-called ‘dinosaur killer’ asteroids, total casualties are very low (even including Chelyabinsk)!
How sure are you about that? Everyone on Air France flight 447 died as a direct result of insufficiently-friendly AI. Patient zero for ebola in the US was discharged from the emergency room with some painkillers and antibiotics because software looking through his medical records failed to recognize, and call the triage nurse’s attention to, obvious risk factors.
You could say they died of insufficient friendly AGI, but not from an AGI that was insufficiently friendly.
By that logic, every death everywhere can be attributed to insufficient friendly AGI.
Among the causes, the easiest to prevent was pilot error due to inadequate training .
I’m saying the control systems on the plane constitute an artificial intelligence, advanced enough to do almost all the work of flying the plane but not general enough to do anything unrelated to flying the plane, and (at issue) not friendly enough to tell the pilots “this is likely just a little ice clog, in which case you don’t need to do anything” or, later in the catastrophe, “the single most important part of stall recovery is bringing the plane’s nose back down,” let alone override their panicky incompetence outright and act on such simple suggestions itself.