I agree that the world is messier than we think. I'd also extend that idea: AI's failure modes are so inhuman that almost any error rate is intolerable because error detection is very difficult. When humans genuinely don't know how to complete a task correctly, they hesitate, hedge, ask for clarifications, and engage in social self-checks that enable their collaborators to ratchet up the trust verification threshold. AI's "overconfidence" and AI's lack of social accountability signals make verification much harder. It's not that a non-zero error rate is intolerable per se, but the relatively more difficult error detection makes errors much costlier because they'll persist.
Good piece. I really liked Ok's post but agree with you on this: "Oks’ framing is skewed toward inefficiency and irrationality. The problem isn’t primarily that we’re poor at solving our problems – it’s that the problems are genuinely hard."
Example, physical therapy. It looks like people should do PT with various orthopedic issues. But most don't do and the superficial explanation is stupidity of lack of self control.
I've you get closer you see:
Many orthopedic issues aren't as easy to diagnose.
PT is intricate, and getting the proper Ilan, instructions and dynamic adjustments over time is really hard.
Many PT professionals are no good
This I've noticed with the experience of myself and friends....
I love the View of the World from 9th Avenue by Saul Steinberg drawing.
I think that it encapsulates AI / the world is messy amazingly well. Because, sure, AI could create that drawing, but knowing when to use it, it's deep meaning and how it fits into the article, that's spectacular and still requires a human touch.
I agree that the world is messier than we think. I'd also extend that idea: AI's failure modes are so inhuman that almost any error rate is intolerable because error detection is very difficult. When humans genuinely don't know how to complete a task correctly, they hesitate, hedge, ask for clarifications, and engage in social self-checks that enable their collaborators to ratchet up the trust verification threshold. AI's "overconfidence" and AI's lack of social accountability signals make verification much harder. It's not that a non-zero error rate is intolerable per se, but the relatively more difficult error detection makes errors much costlier because they'll persist.
Good piece. I really liked Ok's post but agree with you on this: "Oks’ framing is skewed toward inefficiency and irrationality. The problem isn’t primarily that we’re poor at solving our problems – it’s that the problems are genuinely hard."
Great on "the world is messy"
Example, physical therapy. It looks like people should do PT with various orthopedic issues. But most don't do and the superficial explanation is stupidity of lack of self control.
I've you get closer you see:
Many orthopedic issues aren't as easy to diagnose.
PT is intricate, and getting the proper Ilan, instructions and dynamic adjustments over time is really hard.
Many PT professionals are no good
This I've noticed with the experience of myself and friends....
For the chart of payroll employment, any idea why healthcare and private education are grouped together?
It's how BLS groups them. I do find it a bit odd.
Healthcare is the vast majority of that category.
This new piece of mine may also be of interest.
https://open.substack.com/pub/arachnemag/p/the-jevons-paradox-for-intelligence?r=18kjq3&utm_medium=ios
I love the View of the World from 9th Avenue by Saul Steinberg drawing.
I think that it encapsulates AI / the world is messy amazingly well. Because, sure, AI could create that drawing, but knowing when to use it, it's deep meaning and how it fits into the article, that's spectacular and still requires a human touch.
I mostly agree with this, but I think the frame is slightly off.
The question isn’t when AI can “do jobs,” it’s when execution no longer needs to pass through humans.
Most roles aren’t automated end-to-end, they’re hollowed out as coordination, reconciliation, and decision latency disappear.
That looks slow and boring from the outside, until suddenly the org chart doesn’t make sense anymore.