Seems Elon Musk thinks AI is the greatest menace to humanity.
This leads to:
So we need to be very careful with the artificial intelligence. There should be some regulatory oversight maybe at the national and international level, just to make sure that we don’t do something very foolish."
Which, on a moment's contemplation, makes no sense.
See, if I create a rogue AI, what can it do? Send spam e-mails? Start a blog? Surf the net for robot porn? Dream of electric sheep?
You really think some unregulated hacker is going to doom mankind?
But... if the government creates an AI... and puts it to use... that's an entirely different kettle of fish.
So perhaps the correct solution is to prevent any government involvement in AI?
Afterthoughts:
The classic "let's put it in charge of the nuclear arsenal" scenario isn't the only plausible big-government use for an infallible computer intelligence, to eliminate all possibility of human error. What else could be delegated to the Central Computer? Well, there's air traffic control. And economic planning. And Justice. Scared yet?
Meanwhile, consider what private industry might do. Self-driving cars! Wow! A rogue self-driving car could do as much damage as... as... as a rogue human with a regular car. Which is right up there with a rogue (minimally-trained) human with all the smallarms he can carry.
Now, if a whole bunch of self-driving cars went rogue all at once, that'd be more troublesome. Like, a really big deal. But, unless there's a shared, calendar-based bug, or the cars are networked and conspiring against us...?
On the economic front: suppose we put an AI in charge of monetary policy? Think it could screw up any worse than the clowns we've got now? But, to inspire confidence, how about a well-defined, non-intelligent algorithm? It'd have to be a little more complex than "adjust the money supply such that the price of gold remains at $35 per troy ounce", but not necessarily a lot more. Ideally, it'd be simple enough for everyone to understand what drove the money supply. Perhaps elegance and transparency would prove more important, in the real world, than conceptual optimality?
Comments