One semes­ter of grad­u­ate school is over. I am I sup­pose about 10% an engi­neer. A lower-​​case “p” of the Ph.D., perhaps.

The theme that has tied this first semes­ter together, its golden thread, is expe­ri­ence hurts.

Not the gain­ing of experience—that is in fact the fun, the deep­est plea­sure to be gained from an edu­ca­tion re-​​begun. But rather: hav­ing expe­ri­ence out­side the track expected by one’s instruc­tors is worse than hav­ing no expe­ri­ence at all.

The young green stu­dent is not expected to ask but why would any­body think to do it that way at all? for she has learned the tacit assump­tions of her field in her under­grad­u­ate classes [of which I have none]. The real inno­cent does not rail at the sys­tem­atic con­ceal­ment of mod­ern tech­niques for the sake of tried-​​and-​​true prim­i­tive ones, or the inten­tional mask­ing of dif­fi­cult prob­lems and cru­cially pre­car­i­ous assump­tions [noise? ran­dom­ness? uncer­tainty? for­get them now; they will only con­fuse the issue]. Nor at the paucity of cru­cial (even help­ful) insights from other dis­ci­plines, from other silos, from far-​​off weirdos work­ing in other build­ings [We do not learn from data. We do not know here the rev­o­lu­tion that has come upon sta­tis­tics in the last 20 years.]. “You will all remem­ber from your 373 classes that…” or “Cre­ate an AMPL model (you all know how to run AMPL, right?) that…” are state­ments of pre­sump­tive track­ing [Such a shock to an instruc­tor, to know the cul­ture is so very local].

This is the way we do it here can­not be met suc­cess­fully with This is the way we did it else­where. There is no facil­ity for such a thing in higher education.

And: that is a good thing. For what bet­ter role is there for a grad­u­ate degree, than to make you want to push the System’s but­tons right back? How else can it grow, than by instill­ing rage rather than com­pla­cency in its own con­stituents? Who will come to its aid, if not those who are made most inti­mately aware of its self-​​destructive shortcomings?

A pro­fes­sor whose class I did not take said to me the other day, “There are only two ways to learn a sub­ject well: Take a good class in it, or teach a good class in it.”

There are at least two oth­ers he missed. One is: build a machine that does it. But that is beyond most of my peers, still, though I think I will be chang­ing that quickly enough.

The other one he missed, well… it’s bet­ter demon­strated, I think.