A Chessboard of Feedback Loops

I always felt that the best jobs, the well-paid and desirable ones, the ones parents hope for their children, consist of telling other people what to do.

It seemed to me the people doing the actual work, the ones pouring concrete, stocking shelves and scrubbing floors, occupied the bottom of the arrangement, while above them rose whole enormous cities of office towers home to people that, as far as I could tell, held a lot of meetings, and wrote various kinds of documents about the work rather than doing the work itself.

I could not see why any of it was necessary. It had the look of a trick that had been played on everyone, long ago, and that everyone had since agreed not to mention. This perspective was reinforced by reading things like The New Industrial State and with some familiarity with concepts like Bullshit Jobs.

It took years of trial and error, the majority of it error, for me to see what I was missing. There are a nearly limitless number of ways to do anything. Any job can be done in a thousand different sequences, at a dozen scales, with this crew or that one, starting here or starting there, and nearly all of these ways are bad. The difference between an endeavor where somebody has sat down beforehand and worked out the how, the what, the where, and the why, and an endeavor run scattershot, is not a matter of ten percent efficieny. It is often the difference between finishing and not finishing, between usefulness and disaster.

So I came to understand gradually, through my own experience, and being able to get over my native intuitive hangups, colored by my well-meaning, sympathetic prejudices, that knowledge work is, yes, actually useful work.

And now it's going to change its character and we don't know what it will look like or what will happen to it.

In my minds eye, I go through an office and watch it closely. What I see is documents generated, sent, read, and responded to. Forms filled out, which cause other forms to be created. Databases assembled, then checked against other databases. Information moved from one container to another, container after container. At the end of the line a human being looks at the assembled information and its various streams and renders a decision. This decision then goes back through the system and a new chain of thinking and reading and writing and deciding is set off. Eventually, in my mind's eye, I see what looks like a vast array of feedback loops interacting with one another on an immense chess board.

Large segments of this maelstrom of planning and communication activity is now going to be merged into the whirlpool of the computer systems, with AI as the glue that holds it together: the transmission mechanism, the thing that speaks to the person requesting at one end and to the human decision maker at the other, and absorbs the churn of paper-shaped effort in between. The chessboard with the feedback loops has always been analog: a physical system of buildings, offices and people. But in retrospect, thinking of it in those terms, it seems crazy that the whole thing wasn't digitized sooner, and that we actually spent years getting permits for things like parking lots due to the immense actual physical and biological effort required in manually operating these cobbled-together systems of steel, concrete, paper, soundwaves, vocal cords, shoe leather, business suits, haircuts, glass, handshaking, printing ink, pencils, gasoline, eyeglasses and lovely, brilliant people trying to do their best.