Predict the weather
Read an X-ray
Figure out the P&L of a large company
Pick a face out of a crowd
Fly a jet across the country
Maintain the temperature of your house
Book a flight
Create an index for a book
Weld a metal seam
Place online ads
Figure out what book to read next
Water a plant
Monitor a premature newborn
Detect a fire
Read documents in a lawsuit
If you've seen enough movies, you've probably bought into the homunculus model of AI–that it's in the future and it represents a little mechanical man in a box, as mysterious in his motivations as we are.
The future of AI is probably a lot like the past: it nibbles. Artificial intelligence does a job we weren't necessarily crazy about doing anyway, it does it quietly, and well, and then we take it for granted. No one complained when their thermostat took over the job of building a fire, opening the grate, opening a window, rebuilding a fire. And no one complained when the computer found 100 flights faster and better than we ever could.
But the system doesn't get tired, it keeps nibbling. Not with benign or mal intent, but with a focus on a clearly defined task.
This can't help but lead to unintended consequences, enormous when they happen to you, and mostly small in the universal scheme of things. Technology destroys the perfect and then it enables the impossible.
The question each of us has to ask is simple (but difficult): What can I become quite good at that's really difficult for a computer to do one day soon? How can I become so resilient, so human and such a linchpin that shifts in technology won't be able to catch up?
It was always important, but now it's urgent.