Bill de hÓra wrote,
> We're always optimizing, just not always optimizing for performance.
> For the hillclimbers among us:
>
> http://citeseer.nj.nec.com/wolpert95no.html
Fascinating. Go back twenty years to the late 70's and you'll find a
flurry of papers debunking Popper's "truthlikeness" metric on more or
less exactly the same grounds: given a metric which gives you a ranking
of theories by accuracy, you can always derive (in a reasonably
non-perverse way) other metrics which give you arbitrary reorderings of
that ranking.
Google for "Graham Oddie truthlikeness".
Cheers,
Miles
