AI might never achieve true human awareness (called “AGI”) but it has already shown that much of what we think and do can be accomplished by a machine.
And it’s scary.
What got me thinking about it this past weekend was a long interview with AI skeptic Emily Bender (sorry, it’s behind a paywall), who convincingly points out that LLMs are fancy puppets and that AI will never be pissed off enough to destroy the world.
The tech bros selling AI — she calls it “automation in a shiny wrapper” — are lying and making zillions doing so, and will end up blowing up social and environmental well-being in order to ultimately fail. She’s written a book called The AI Con that makes a very compelling case that AI is, in so many words, glorified search.
I think that she’s absolutely right. And she totally misses the point.
The idea that consciousness will one day pop into a silicon brain is preposterous; there’s no reason to believe that awareness of self, presence in physical space, or experience of emotions can emerge from evermore complicated and data-rich machines.
We can’t deconstruct how human minds do it, so the premise that we could build an AI that possessed it is a misinformed hope, at best, and more likely intentional marketing hyperbole, as Ms. Bender (and I) believe.
But relentlessly bombarding us with it focuses us on the fantasy that AIs will one day think and act as we do, when the technology is already revealing that we think and act like them.
We’ve got it backwards.
It turns out that much of what we humans do in business, governing, science, and the arts can be done by machines. They may not be brilliantly good at doing any of the tasks but they’re mostly no worse than we are at doing them.
Analyzing insurance claims. Diagnosing illnesses. Summarizing meeting reports. Picking stocks. Writing poems and songs. Researching school book reports. Assessing potential life partners.
The list goes on and on.
AIs don’t have to become exactly like us to do most things mostly like us. And we may never get an artificial Buddha or Beethoven, but the 99.99999% of us who aren’t particularly gifted are on the road to discovering that we’re not terribly special.
Most of what we do will be outsourced to AIs, whether or not we want it to happen, while we’ll chose to let AI augment and manage the rest of it for us. And we have no public discourse on what kind of world that will be, let alone if we’ll want to live in it.
While we’re fantasizing about AGI and an AI Apocalypse, our lives are being irrevocably changed by an imperfect (and imperfectly understood) technology.
Only we’re not talking about it. And that’s the real con.
[This essay appeared originally at You And Your Machine]