The Mindlessness Of AI
/“I have an idea, let’s teach machines how to think for themselves”.
————
Vat, now I must argue with my toaster about my toast shud be lighter or darker?
Oy veh! vat next? My whole life everybody tellin’ me vat to do. Now my toaster tells me vat my toast shud be! No, no, no. I throw that blabbermouth in der trasch and toast my bread over the fire. The fire don’t talk back to me.
Oh no, Bubbeh, AI’s not like that. Artificial Intelligence will only do what you want it to do. AI is all about making a better life for everyone.
Yah sure, like der Nazis put us on a train to a resettlement so we vud be happier. I make my own intelligence. I don’t need no artificial intelligence. I don’t need no artificial, nothing.
Bubbeh’s distrust of promise over delivery is common-sense, sensible. As Robert Burns wrote, “The best laid schemes o’ mice an’ men gang aft a-gley”. Beautifull promises often have ugly endings.
“I have an idea, let’s teach machines how to think for themselves”.
What could possibly go wrong?
Let me count the ways.
First to mind is the old computer axiom: Garbage in, garbage out. How will an artificially intelligent process know what’s true, and what isn’t? It won’t. It can’t be any better at that than the humans who program the information.
There are serious human disagreements about what is true and what is false, including disagreements between scientists, engineers and physicists.
The notion that Artificial Intelligence will be able to sort it all out is based on the stupid idea that the most referenced opinion is the true opinion.
In Medieval Europe the world was thought to be flat. The best minds of the time were in complete agreement on that point. If AI were possible in those years, it would concur.
By the time Medieval-AI got updated information many bold adventurers would have abandoned plans to sail to the edge of the world for fear of falling off the edge.
Well, so what, that’s what happened anyway. Medieval-AI wouldn’t have done any more damage than the human intelligence of the day. Quite so. That’s the point. Artificial Intelligence can never be any smarter than the information fed to it by humans.
“Intelligent” machines, like all calculators, are superb at counting and processing data. They’re flawed when analyzing data. Only humans can do that. Only humans can truly, “analyze” - because humans have agency for hunches.
No version of Artificial Intelligence will ever have a hunch.
However, Artificial Intelligence is capable of producing loopy ideas based on probability.
The greatest danger of AI is nerdish faith in the ability of process to replace wisdom.
If Artificial Intelligence can be used appropriately by Human intelligence it will be of great service. If not, I have the old-age consolation of likely being dead before it does too much damage.
I’ve a hopeful way of thinking about the future.
By K. L. Shipley
Website: https://www.eclecticessays